JuliaPro v0.6.4.

1 API Manual

July 13, 2018

Julia Computing Inc.

[email protected]
Contents

1 Distributions 1

2 StatsBase 46

3 PyPlot 84

4 IndexedTables 112

5 Images 146

6 DataFrames 165

7 Knet 183

8 BinaryProvider 199

9 DataStructures 213

10 JDBC 226

11 ImageCore 237

12 Reactive 246

13 Combinatorics 254

14 JuliaDB 262

15 HypothesisTests 269

16 PyCall 275

17 CategoricalArrays 282

18 GLM 288

19 Roots 295

I
II CONTENTS

20 Documenter 302

21 ColorTypes 307

22 FileIO 312

23 DataArrays 317

24 Primes 324

25 ImageTransformations 330

26 StatsFuns 334

27 Gadfly 339

28 IterTools 344

29 Polynomials 352

30 Colors 356

31 Interact 360

32 FFTW 364

33 WebSockets 368

34 AxisArrays 371

35 BusinessDays 375

36 QuadGK 378

37 Dagger 381

38 NNlib 384

39 ShowItLikeYouBuildIt 387

40 CoordinateTransformations 390

41 Graphics 392

42 MacroTools 395

43 ImageMorphology 398

44 NLSolversBase 400
CONTENTS III

45 Contour 402

46 Compose 404

47 CoupledFields 407

48 AxisAlgorithms 409

49 Libz 411

50 ImageMetadata 414

51 LegacyStrings 416

52 BufferedStreams 418

53 MbedTLS 420

54 DataValues 422

55 NearestNeighbors 424

56 IJulia 425

57 HDF5 427

58 CUDAapi 429

59 AutoGrad 431

60 ComputationalResources 433

61 MicroLogging 435

62 Clustering 437

63 JuliaWebAPI 439

64 DecisionTree 441

65 Parameters 443

66 Blosc 445

67 Missings 447

68 ProgressMeter 449

69 HttpServer 451
IV CONTENTS

70 MappedArrays 453

71 TextParse 454

72 LossFunctions 456

73 LearnBase 457

74 Juno 459

75 StatsModels 461

76 InternedStrings 462

77 HttpParser 463

78 ImageAxes 464

79 Flux 465

80 OnlineStats 466

81 IntervalSets 467

82 Media 468

83 Rotations 469

84 Mustache 470

85 TiledIteration 471

86 Distances 472

87 VersionParsing 473

88 HttpCommon 474

89 StaticArrays 475

90 SweepOperator 476

91 PaddedViews 477

92 Pidfile 478

93 SpecialFunctions 479

94 NamedTuples 480
CONTENTS V

95 OnlineStatsBase 481

96 Loess 482

97 WoodburyMatrices 483

98 Lazy 484

99 Requests 485

100 LineSearches 486

101 MemPool 487

102 SimpleTraits 488

103 BinDeps 489

104 LaTeXStrings 490

VI CONTENTS
Chapter 1

Distributions

1.1 Base.LinAlg.scale!
Base.LinAlg.scale! — Method.

scale!{D<:AbstractMvLogNormal}(::Type{D},s::Symbol,m::AbstractVector,S::AbstractMatrix,::Abstract

Calculate the scale parameter, as defined for the location parameter above
and store the result in .
source

1.2 Base.mean
Base.mean — Method.

mean(d::Union{UnivariateMixture, MultivariateMixture})

Compute the overall mean (expectation).

source

1.3 Base.mean
Base.mean — Method.

mean(d::MatrixDistribution)

Return the mean matrix of d.

source

1
2 CHAPTER 1. DISTRIBUTIONS

1.4 Base.mean
Base.mean — Method.

mean(d::MultivariateDistribution)

Compute the mean vector of distribution d.

source

1.5 Base.mean
Base.mean — Method.

mean(d::UnivariateDistribution)

Compute the expectation.

source

1.6 Base.median
Base.median — Method.

median(d::UnivariateDistribution)

Return the median value of distribution d.

source

1.7 Base.median
Base.median — Method.

median(d::MvLogNormal)

Return the median vector of the lognormal distribution. which is strictly

smaller than the mean.
source

1.8 Base.quantile
Base.quantile — Method.

quantile(d::UnivariateDistribution, q::Real)

Evaluate the inverse cumulative distribution function at q.

See also: cquantile, invlogcdf, and invlogccdf.
source
1.9. BASE.STD 3

1.9 Base.std
Base.std — Method.
std(d::UnivariateDistribution)
Return the standard deviation of distribution d, i.e. sqrt(var(d)).
source

1.10 Base.var
Base.var — Method.
var(d::UnivariateMixture)
Compute the overall variance (only for U nivariateM ixture).
source

1.11 Base.var
Base.var — Method.
var(d::MultivariateDistribution)
Compute the vector of element-wise variances for distribution d.
source

1.12 Base.var
Base.var — Method.
var(d::UnivariateDistribution)
Compute the variance. (A generic std is provided as std(d) = sqrt(var(d)))
source

1.13 Distributions.TruncatedNormal
Distributions.TruncatedNormal — Method.
TruncatedNormal(mu, sigma, l, u)
The truncated normal distribution is a particularly important one in the fam-
ily of truncated distributions. We provide additional support for this type with
TruncatedNormal which calls Truncated(Normal(mu, sigma), l, u). Unlike
the general case, truncated normal distributions support mean, mode, modes,
var, std, and entropy.
source
4 CHAPTER 1. DISTRIBUTIONS

1.14 Distributions.ccdf
Distributions.ccdf — Method.

ccdf(d::UnivariateDistribution, x::Real)

The complementary cumulative function evaluated at x, i.e. 1 - cdf(d,

x).
source

1.15 Distributions.cdf
Distributions.cdf — Method.

cdf(d::UnivariateDistribution, x::Real)

Evaluate the cumulative probability at x.

See also ccdf, logcdf, and logccdf.
source

1.16 Distributions.cf
Distributions.cf — Method.

cf(d::UnivariateDistribution, t)

Evaluate the characteristic function of distribution d.

source

1.17 Distributions.components
Distributions.components — Method.

components(d::AbstractMixtureModel)

Get a list of components of the mixture model d.

source

1.18 Distributions.cquantile
Distributions.cquantile — Method.

cquantile(d::UnivariateDistribution, q::Real)

The complementary quantile value, i.e. quantile(d, 1-q).

source
1.19. DISTRIBUTIONS.FAILPROB 5

1.19 Distributions.failprob
Distributions.failprob — Method.

failprob(d::UnivariateDistribution)

Get the probability of failure.

source

1.20 Distributions.fit mle

Distributions.fit mle — Method.

fit_mle(D, x, w)

Fit a distribution of type D to a weighted data set x, with weights given by

w.
Here, w should be an array with length n, where n is the number of samples
contained in x.
source

1.21 Distributions.fit mle

Distributions.fit mle — Method.

fit_mle(D, x)

Fit a distribution of type D to a given data set x.

• For univariate distribution, x can be an array of arbitrary size.

• For multivariate distribution, x should be a matrix, where each column is

a sample.

source

1.22 Distributions.insupport
Distributions.insupport — Method.

insupport(d::MultivariateMixture, x)

Evaluate whether x is within the support of mixture distribution d.

source
6 CHAPTER 1. DISTRIBUTIONS

1.23 Distributions.insupport
Distributions.insupport — Method.

insupport(d::MultivariateDistribution, x::AbstractArray)

If x is a vector, it returns whether x is within the support of d. If x is a

matrix, it returns whether every column in x is within the support of d.
source

1.24 Distributions.insupport
Distributions.insupport — Method.

insupport(d::UnivariateDistribution, x::Any)

When x is a scalar, it returns whether x is within the support of d (e.g.,

insupport(d, x) = minimum(d) <= x <= maximum(d)). When x is an array,
it returns whether every element in x is within the support of d.
Generic fallback methods are provided, but it is often the case that insupport
can be done more efficiently, and a specialized insupport is thus desirable. You
should also override this function if the support is composed of multiple disjoint
intervals.
source

1.25 Distributions.invcov
Distributions.invcov — Method.

invcov(d::AbstractMvNormal)

Return the inversed covariance matrix of d.

source

1.26 Distributions.invlogccdf
Distributions.invlogccdf — Method.

invlogcdf(d::UnivariateDistribution, lp::Real)

The inverse function of logcdf.

source
1.27. DISTRIBUTIONS.INVLOGCDF 7

1.27 Distributions.invlogcdf
Distributions.invlogcdf — Method.

invlogcdf(d::UnivariateDistribution, lp::Real)

The inverse function of logcdf.

source

1.28 Distributions.isleptokurtic
Distributions.isleptokurtic — Method.

isleptokurtic(d)

Return whether d is leptokurtic (i.e kurtosis(d) < 0).

source

1.29 Distributions.ismesokurtic
Distributions.ismesokurtic — Method.

ismesokurtic(d)

Return whether d is mesokurtic (i.e kurtosis(d) == 0).

source

1.30 Distributions.isplatykurtic
Distributions.isplatykurtic — Method.

isplatykurtic(d)

Return whether d is platykurtic (i.e kurtosis(d) > 0).

source

1.31 Distributions.location!
Distributions.location! — Method.

location!{D<:AbstractMvLogNormal}(::Type{D},s::Symbol,m::AbstractVector,S::AbstractMatrix,::Abstr

Calculate the location vector (as above) and store the result in
source
8 CHAPTER 1. DISTRIBUTIONS

1.32 Distributions.location
Distributions.location — Method.

location(d::UnivariateDistribution)

Get the location parameter.

source

1.33 Distributions.location
Distributions.location — Method.

location(d::MvLogNormal)

Return the location vector of the distribution (the mean of the underlying
normal distribution).
source

1.34 Distributions.location
Distributions.location — Method.

location{D<:AbstractMvLogNormal}(::Type{D},s::Symbol,m::AbstractVector,S::AbstractMatri

Calculate the location vector (the mean of the underlying normal distribu-
tion).

• If s == :meancov, then m is taken as the mean, and S the covariance

matrix of a lognormal distribution.
• If s == :mean | :median | :mode, then m is taken as the mean, median
or mode of the lognormal respectively, and S is interpreted as the scale
matrix (the covariance of the underlying normal distribution).

It is not possible to analytically calculate the location vector from e.g., me-
dian + covariance, or from mode + covariance.
source

1.35 Distributions.logccdf
Distributions.logccdf — Method.

logccdf(d::UnivariateDistribution, x::Real)

The logarithm of the complementary cumulative function values evaluated

at x, i.e. log(ccdf(x)).
source
1.36. DISTRIBUTIONS.LOGCDF 9

1.36 Distributions.logcdf
Distributions.logcdf — Method.

logcdf(d::UnivariateDistribution, x::Real)

The logarithm of the cumulative function value(s) evaluated at x, i.e. log(cdf(x)).

source

1.37 Distributions.logdetcov
Distributions.logdetcov — Method.

logdetcov(d::AbstractMvNormal)

Return the log-determinant value of the covariance matrix.

source

1.38 Distributions.logpdf
Distributions.logpdf — Method.

logpdf(d::Union{UnivariateMixture, MultivariateMixture}, x)

Evaluate the logarithm of the (mixed) probability density function over x.

Here, x can be a single sample or an array of multiple samples.
source

1.39 Distributions.logpdf
Distributions.logpdf — Method.

logpdf(d::MultivariateDistribution, x::AbstractArray)

Return the logarithm of probability density evaluated at x.

• If x is a vector, it returns the result as a scalar.

• If x is a matrix with n columns, it returns a vector r of length n, where

r[i] corresponds to x[:,i].

logpdf!(r, d, x) will write the results to a pre-allocated array r.

source
10 CHAPTER 1. DISTRIBUTIONS

1.40 Distributions.logpdf
Distributions.logpdf — Method.

logpdf(d::UnivariateDistribution, x::Real)

Evaluate the logarithm of probability density (mass) at x. Whereas there

is a fallback implemented logpdf(d, x) = log(pdf(d, x)). Relying on this
fallback is not recommended in general, as it is prone to overflow or underflow.
source

1.41 Distributions.logpdf
Distributions.logpdf — Method.

logpdf(d::MatrixDistribution, AbstractMatrix)

Compute the logarithm of the probability density at the input matrix x.

source

1.42 Distributions.mgf
Distributions.mgf — Method.

mgf(d::UnivariateDistribution, t)

Evaluate the moment generating function of distribution d.

source

1.43 Distributions.ncategories
Distributions.ncategories — Method.

ncategories(d::UnivariateDistribution)

Get the number of categories.

source

1.44 Distributions.nsamples
Distributions.nsamples — Method.

nsamples(s::Sampleable)

The number of samples contained in A. Multiple samples are often organized

into an array, depending on the variate form.
source
1.45. DISTRIBUTIONS.NTRIALS 11

1.45 Distributions.ntrials
Distributions.ntrials — Method.

ntrials(d::UnivariateDistribution)

Get the number of trials.

source

1.46 Distributions.pdf
Distributions.pdf — Method.

pdf(d::Union{UnivariateMixture, MultivariateMixture}, x)

Evaluate the (mixed) probability density function over x. Here, x can be a

single sample or an array of multiple samples.
source

1.47 Distributions.pdf
Distributions.pdf — Method.

pdf(d::MultivariateDistribution, x::AbstractArray)

Return the probability density of distribution d evaluated at x.

• If x is a vector, it returns the result as a scalar.

• If x is a matrix with n columns, it returns a vector r of length n, where

r[i] corresponds

to x[:,i] (i.e. treating each column as a sample).

pdf!(r, d, x) will write the results to a pre-allocated array r.
source

1.48 Distributions.pdf
Distributions.pdf — Method.

pdf(d::UnivariateDistribution, x::Real)

Evaluate the probability density (mass) at x.

See also: logpdf.
source
12 CHAPTER 1. DISTRIBUTIONS

1.49 Distributions.pdf
Distributions.pdf — Method.

pdf(d::MatrixDistribution, x::AbstractArray)

Compute the probability density at the input matrix x.

source

1.50 Distributions.probs
Distributions.probs — Method.

probs(d::AbstractMixtureModel)

Get the vector of prior probabilities of all components of d.

source

1.51 Distributions.rate
Distributions.rate — Method.

rate(d::UnivariateDistribution)

Get the rate parameter.

source

1.52 Distributions.sampler
Distributions.sampler — Method.

sampler(d::Distribution) -> Sampleable

Samplers can often rely on pre-computed quantities (that are not parameters
themselves) to improve efficiency. If such a sampler exists, it can be provide
with this sampler method, which would be used for batch sampling. The general
fallback is sampler(d::Distribution) = d.
source

1.53 Distributions.scale
Distributions.scale — Method.

scale(d::UnivariateDistribution)

Get the scale parameter.

source
1.54. DISTRIBUTIONS.SCALE 13

1.54 Distributions.scale
Distributions.scale — Method.
scale(d::MvLogNormal)
Return the scale matrix of the distribution (the covariance matrix of the
underlying normal distribution).
source

1.55 Distributions.scale
Distributions.scale — Method.
scale{D<:AbstractMvLogNormal}(::Type{D},s::Symbol,m::AbstractVector,S::AbstractMatrix)
Calculate the scale parameter, as defined for the location parameter above.
source

1.56 Distributions.shape
Distributions.shape — Method.
shape(d::UnivariateDistribution)
Get the shape parameter.
source

1.57 Distributions.sqmahal
Distributions.sqmahal — Method.
sqmahal(d, x)
Return the squared Mahalanobis distance from x to the center of d, w.r.t.
the covariance. When x is a vector, it returns a scalar value. When x is a
matrix, it returns a vector of length size(x,2).
sqmahal!(r, d, x) with write the results to a pre-allocated array r.
source

1.58 Distributions.succprob
Distributions.succprob — Method.
succprob(d::UnivariateDistribution)
Get the probability of success.
source
14 CHAPTER 1. DISTRIBUTIONS

1.59 StatsBase.dof
StatsBase.dof — Method.

dof(d::UnivariateDistribution)

Get the degrees of freedom.

source

1.60 StatsBase.entropy
StatsBase.entropy — Method.

entropy(d::MultivariateDistribution, b::Real)

Compute the entropy value of distribution d, w.r.t. a given base.

source

1.61 StatsBase.entropy
StatsBase.entropy — Method.

entropy(d::MultivariateDistribution)

Compute the entropy value of distribution d.

source

1.62 StatsBase.entropy
StatsBase.entropy — Method.

entropy(d::UnivariateDistribution, b::Real)

Compute the entropy value of distribution d, w.r.t. a given base.

source

1.63 StatsBase.entropy
StatsBase.entropy — Method.

entropy(d::UnivariateDistribution)

Compute the entropy value of distribution d.

source
1.64. STATSBASE.KURTOSIS 15

1.64 StatsBase.kurtosis
StatsBase.kurtosis — Method.

kurtosis(d::Distribution, correction::Bool)

Computes excess kurtosis by default. Proper kurtosis can be returned with

correction=false
source

1.65 StatsBase.kurtosis
StatsBase.kurtosis — Method.

kurtosis(d::UnivariateDistribution)

Compute the excessive kurtosis.

source

1.66 StatsBase.loglikelihood
StatsBase.loglikelihood — Method.

loglikelihood(d::MultivariateDistribution, x::AbstractMatrix)

The log-likelihood of distribution d w.r.t. all columns contained in matrix

x.
source

1.67 StatsBase.loglikelihood
StatsBase.loglikelihood — Method.

loglikelihood(d::UnivariateDistribution, X::AbstractArray)

The log-likelihood of distribution d w.r.t. all samples contained in array x.

source

1.68 StatsBase.mode
StatsBase.mode — Method.

mode(d::UnivariateDistribution)

Returns the first mode.

source
16 CHAPTER 1. DISTRIBUTIONS

1.69 StatsBase.mode
StatsBase.mode — Method.

mode(d::MvLogNormal)

Return the mode vector of the lognormal distribution, which is strictly

smaller than the mean and median.
source

1.70 StatsBase.modes
StatsBase.modes — Method.

modes(d::UnivariateDistribution)

Get all modes (if this makes sense).

source

1.71 StatsBase.params!
StatsBase.params! — Method.

params!{D<:AbstractMvLogNormal}(::Type{D},m::AbstractVector,S::AbstractMatrix,::Abstrac

Calculate (scale,location) for a given mean and covariance, and store the
results in and
source

1.72 StatsBase.params
StatsBase.params — Method.

params(d::UnivariateDistribution)

Return a tuple of parameters. Let d be a distribution of type D, then

D(params(d)...) will construct exactly the same distribution as d.
source

1.73 StatsBase.params
StatsBase.params — Method.

params{D<:AbstractMvLogNormal}(::Type{D},m::AbstractVector,S::AbstractMatrix)

Return (scale,location) for a given mean and covariance

source
1.74. STATSBASE.SKEWNESS 17

1.74 StatsBase.skewness
StatsBase.skewness — Method.

skewness(d::UnivariateDistribution)

Compute the skewness.

source

1.75 Distributions.AbstractMvNormal
Distributions.AbstractMvNormal — Type.
The Multivariate normal distribution is a multidimensional generalization
of the normal distribution. The probability density function of a d-dimensional
multivariate normal distribution with mean vector µ and covariance matrix Σ
is:
 
1 1
f (x; µ, Σ) = exp − (x − µ)T Σ−1 (x − µ)
(2π)d/2 |Σ|1/2 2

We realize that the mean vector and the covariance often have special forms
in practice, which can be exploited to simplify the computation. For example,
the mean vector is sometimes just a zero vector, while the covariance matrix
can be a diagonal matrix or even in the form of σI. To take advantage of such
special cases, we introduce a parametric type MvNormal, defined as below, which
allows users to specify the special structure of the mean and covariance.
[] immutable MvNormal{Cov¡:AbstractPDMat,Mean¡:Union{Vector,ZeroVector}}
¡: AbstractMvNormal ::Mean ::Cov end
Here, the mean vector can be an instance of either Vector or ZeroVector,
where the latter is simply an empty type indicating a vector filled with zeros.
The covariance can be of any subtype of AbstractPDMat. Particularly, one can
use PDMat for full covariance, PDiagMat for diagonal covariance, and ScalMat
for the isotropic covariance – those in the form of σI. (See the Julia package
PDMats for details).
We also define a set of alias for the types using different combinations of
mean vectors and covariance:
[] const IsoNormal = MvNormal{ScalMat, Vector{Float64}} const DiagNor-
mal = MvNormal{PDiagMat, Vector{Float64}} const FullNormal = MvNor-
mal{PDMat, Vector{Float64}}
const ZeroMeanIsoNormal = MvNormal{ScalMat, ZeroVector{Float64}} const
ZeroMeanDiagNormal = MvNormal{PDiagMat, ZeroVector{Float64}} const
ZeroMeanFullNormal = MvNormal{PDMat, ZeroVector{Float64}}
source
18 CHAPTER 1. DISTRIBUTIONS

1.76 Distributions.Arcsine
Distributions.Arcsine — Type.

Arcsine(a,b)

The Arcsine distribution has probability density function

1
f (x) = p , x ∈ [a, b]
π (x − a)(b − x)

[] Arcsine() Arcsine distribution with support [0, 1] Arcsine(b) Arcsine

distribution with support [0, b] Arcsine(a, b) Arcsine distribution with support
[a, b]
params(d) Get the parameters, i.e. (a, b) minimum(d) Get the lower
bound, i.e. a maximum(d) Get the upper bound, i.e. b location(d) Get the
left bound, i.e. a scale(d) Get the span of the support, i.e. b - a
External links

• Arcsine distribution on Wikipedia

source

1.77 Distributions.Bernoulli
Distributions.Bernoulli — Type.

Bernoulli(p)

A Bernoulli distribution is parameterized by a success rate p, which takes

value 1 with probability p and 0 with probability 1-p.
(
1−p for k = 0,
P (X = k) =
p for k = 1.

[] Bernoulli() Bernoulli distribution with p = 0.5 Bernoulli(p) Bernoulli

distribution with success rate p
params(d) Get the parameters, i.e. (p,) succprob(d) Get the success rate,
i.e. p failprob(d) Get the failure rate, i.e. 1 - p
External links:

• Bernoulli distribution on Wikipedia

source
1.78. DISTRIBUTIONS.BETA 19

1.78 Distributions.Beta
Distributions.Beta — Type.

Beta(,)

The Beta distribution has probability density function

1
f (x; α, β) = xα−1 (1 − x)β−1 , x ∈ [0, 1]
B(α, β)
The Beta distribution is related to the Gamma distribution via the property
that if X ∼ Gamma(α) and Y ∼ Gamma(β) independently, then X/(X + Y ) ∼
Beta(α, β).
[] Beta() equivalent to Beta(1, 1) Beta(a) equivalent to Beta(a, a) Beta(a,
b) Beta distribution with shape parameters a and b
params(d) Get the parameters, i.e. (a, b)
External links

• Beta distribution on Wikipedia

source

1.79 Distributions.BetaBinomial
Distributions.BetaBinomial — Type.

BetaBinomial(n,,)

A Beta-binomial distribution is the compound distribution of the Binomial

distribution where the probability of success p is distributed according to the
Beta. It has three parameters: n, the number of trials and two shape parameters
,

 
n
P (X = k) = B(k + α, n − k + β)/B(α, β), for k = 0, 1, 2, . . . , n.
k

[] BetaBinomial(n, a, b) BetaBinomial distribution with n trials and shape

parameters a, b
params(d) Get the parameters, i.e. (n, a, b) ntrials(d) Get the number of
trials, i.e. n
External links:

• Beta-binomial distribution on Wikipedia

source
20 CHAPTER 1. DISTRIBUTIONS

1.80 Distributions.BetaPrime
Distributions.BetaPrime — Type.

BetaPrime(,)

The Beta prime distribution has probability density function

1
f (x; α, β) = xα−1 (1 + x)−(α+β) , x>0
B(α, β)
The Beta prime distribution is related to the Beta distribution via the rela-
X
tion ship that if X ∼ Beta(α, β) then 1−X ∼ BetaPrime(α, β)
[] BetaPrime() equivalent to BetaPrime(1, 1) BetaPrime(a) equivalent to
BetaPrime(a, a) BetaPrime(a, b) Beta prime distribution with shape parame-
ters a and b
params(d) Get the parameters, i.e. (a, b)
External links

• Beta prime distribution on Wikipedia

source

1.81 Distributions.Binomial
Distributions.Binomial — Type.

Binomial(n,p)

A Binomial distribution characterizes the number of successes in a sequence

of independent trials. It has two parameters: n, the number of trials, and p, the
probability of success in an individual trial, with the distribution:
 
n k
P (X = k) = p (1 − p)n−k , for k = 0, 1, 2, . . . , n.
k
[] Binomial() Binomial distribution with n = 1 and p = 0.5 Binomial(n)
Binomial distribution for n trials with success rate p = 0.5 Binomial(n, p)
Binomial distribution for n trials with success rate p
params(d) Get the parameters, i.e. (n, p) ntrials(d) Get the number of
trials, i.e. n succprob(d) Get the success rate, i.e. p failprob(d) Get the failure
rate, i.e. 1 - p
External links:

• Binomial distribution on Wikipedia

source
1.82. DISTRIBUTIONS.BIWEIGHT 21

1.82 Distributions.Biweight
Distributions.Biweight — Type.
Biweight(, )
source

1.83 Distributions.Categorical
Distributions.Categorical — Type.
Categorical(p)
A Categorical distribution is parameterized by a probability vector p (of
length K).

P (X = k) = p[k] for k = 1, 2, . . . , K.
[] Categorical(p) Categorical distribution with probability vector p params(d)
Get the parameters, i.e. (p,) probs(d) Get the probability vector, i.e. p ncate-
gories(d) Get the number of categories, i.e. K
Here, p must be a real vector, of which all components are nonnegative
and sum to one. Note: The input vector p is directly used as a field of the
constructed distribution, without being copied. External links:
• Categorical distribution on Wikipedia
source

1.84 Distributions.Cauchy
Distributions.Cauchy — Type.
Cauchy(, )
The Cauchy distribution with location and scale has probability density
function
1
f (x; µ, σ) =  
x−µ 2


πσ 1 + σ

[] Cauchy() Standard Cauchy distribution, i.e. Cauchy(0, 1) Cauchy(u)

Cauchy distribution with location u and unit scale, i.e. Cauchy(u, 1) Cauchy(u,
b) Cauchy distribution with location u and scale b
params(d) Get the parameters, i.e. (u, b) location(d) Get the location
parameter, i.e. u scale(d) Get the scale parameter, i.e. b
External links
• Cauchy distribution on Wikipedia
source
22 CHAPTER 1. DISTRIBUTIONS

1.85 Distributions.Chi
Distributions.Chi — Type.

Chi()

The Chi distribution degrees of freedom has probability density function

1 2
f (x; k) = 21−k/2 xk−1 e−x /2 , x>0
Γ(k/2)

It is the distribution of the square-root of a Chisq variate.

[] Chi(k) Chi distribution with k degrees of freedom
params(d) Get the parameters, i.e. (k,) dof(d) Get the degrees of freedom,
i.e. k
External links

• Chi distribution on Wikipedia

source

1.86 Distributions.Chisq
Distributions.Chisq — Type.

Chisq()

The Chi squared distribution (typically written ) with degrees of freedom

has the probability density function

xk/2−1 e−x/2
f (x; k) = , x > 0.
2k/2 Γ(k/2)

If is an integer, then it is the distribution of the sum of squares of inde-

pendent standard Normal variates.
[] Chisq(k) Chi-squared distribution with k degrees of freedom
params(d) Get the parameters, i.e. (k,) dof(d) Get the degrees of freedom,
i.e. k
External links

• Chi-squared distribution on Wikipedia

source
1.87. DISTRIBUTIONS.COSINE 23

1.87 Distributions.Cosine
Distributions.Cosine — Type.
Cosine(, )
A raised Cosine distribution.
External link:
• Cosine distribution on wikipedia
source

1.88 Distributions.Dirichlet
Distributions.Dirichlet — Type.
Dirichlet
The Dirichlet distribution is often used the conjugate prior for Categorical
or Multinomial distributions. The probability density function of a Dirichlet
distribution with parameter α = (α1 , . . . , αk ) is:

k Qk
1 Y αi −1 Γ(αi )
f (x; α) = x , with B(α) = i=1 , x1 + · · · + xk = 1
B(α) i=1 i Γ
Pk
α
i=1 i

[] Let alpha be a vector Dirichlet(alpha) Dirichlet distribution with param-

eter vector alpha
Let a be a positive scalar Dirichlet(k, a) Dirichlet distribution with param-
eter a * ones(k)
source

1.89 Distributions.DiscreteUniform
Distributions.DiscreteUniform — Type.
DiscreteUniform(a,b)
A Discrete uniform distribution is a uniform distribution over a consecutive
sequence of integers between a and b, inclusive.

P (X = k) = 1/(b − a + 1) for k = a, a + 1, . . . , b.
[] DiscreteUniform(a, b) a uniform distribution over {a, a+1, ..., b}
params(d) Get the parameters, i.e. (a, b) span(d) Get the span of the
support, i.e. (b - a + 1) probval(d) Get the probability value, i.e. 1 / (b - a +
1) minimum(d) Return a maximum(d) Return b
External links
24 CHAPTER 1. DISTRIBUTIONS

• Discrete uniform distribution on Wikipedia

source

1.90 Distributions.Epanechnikov
Distributions.Epanechnikov — Type.
Epanechnikov(, )
source

1.91 Distributions.Erlang
Distributions.Erlang — Type.
Erlang(,)
The Erlang distribution is a special case of a Gamma distribution with integer
shape parameter.
[] Erlang() Erlang distribution with unit shape and unit scale, i.e. Erlang(1,
1) Erlang(a) Erlang distribution with shape parameter a and unit scale, i.e.
Erlang(a, 1) Erlang(a, s) Erlang distribution with shape parameter a and scale
b
External links
• Erlang distribution on Wikipedia
source

1.92 Distributions.Exponential
Distributions.Exponential — Type.
Exponential()
The Exponential distribution with scale parameter has probability density
function
1 −x
f (x; θ) = e θ, x > 0
θ
[] Exponential() Exponential distribution with unit scale, i.e. Exponen-
tial(1) Exponential(b) Exponential distribution with scale b
params(d) Get the parameters, i.e. (b,) scale(d) Get the scale parameter,
i.e. b rate(d) Get the rate parameter, i.e. 1 / b
External links
• Exponential distribution on Wikipedia
source
1.93. DISTRIBUTIONS.FDIST 25

1.93 Distributions.FDist
Distributions.FDist — Type.

FDist(1, 2)

The F distribution has probability density function

s
1 (ν1 x)ν1 · ν2ν2
f (x; ν1 , ν2 ) = , x>0
xB(ν1 /2, ν2 /2) (ν1 x + ν2 )ν1 +ν2

It is related to the Chisq distribution via the property that if X1 ∼ Chisq(ν1 )

and X2 ∼ Chisq(ν2 ), then (X1 /ν1 )/(X2 /ν2 ) ∼ FDist(ν1 , ν2 ).
[] FDist(1, 2) F-Distribution with parameters 1 and 2
params(d) Get the parameters, i.e. (1, 2)
External links

• F distribution on Wikipedia

source

1.94 Distributions.Frechet
Distributions.Frechet — Type.

Frechet(,)

The Frchet distribution with shape and scale has probability density func-
tion

α  x −α−1 −(x/θ)−α
f (x; α, θ) = e , x>0
θ θ
[] Frechet() Frchet distribution with unit shape and unit scale, i.e. Frechet(1,
1) Frechet(a) Frchet distribution with shape a and unit scale, i.e. Frechet(a, 1)
Frechet(a, b) Frchet distribution with shape a and scale b
params(d) Get the parameters, i.e. (a, b) shape(d) Get the shape param-
eter, i.e. a scale(d) Get the scale parameter, i.e. b
External links

• Frchet distribution on Wikipedia

source
26 CHAPTER 1. DISTRIBUTIONS

1.95 Distributions.Gamma
Distributions.Gamma — Type.

Gamma(,)

The Gamma distribution with shape parameter and scale has probability
density function

xα−1 e−x/θ
f (x; α, θ) = , x>0
Γ(α)θα
[] Gamma() Gamma distribution with unit shape and unit scale, i.e. Gamma(1,
1) Gamma() Gamma distribution with shape and unit scale, i.e. Gamma(, 1)
Gamma(, ) Gamma distribution with shape and scale
params(d) Get the parameters, i.e. (, ) shape(d) Get the shape parameter,
i.e. scale(d) Get the scale parameter, i.e.
External links

• Gamma distribution on Wikipedia

source

1.96 Distributions.GeneralizedExtremeValue
Distributions.GeneralizedExtremeValue — Type.

GeneralizedExtremeValue(, , )

The Generalized extreme value distribution with shape parameter , scale and
location has probability density function

−1/ξ−1
−1/ξ o
(  n 
x−µ
1
exp − 1 + x−µ


σ 1+ σ ξ σ ξ for ξ 6= 0
f (x; ξ, σ, µ) = x−µ
 x−µ 
1
 
σ exp − σ exp − exp − σ for ξ = 0

for
h 
σ


 µ − ξ , +∞ for ξ > 0
x ∈ (−∞, +∞)
 i for ξ = 0
 −∞, µ − σ

for ξ < 0
ξ

[] GeneralizedExtremeValue(m, s, k) Generalized Pareto distribution with

shape k, scale s and location m.
params(d) Get the parameters, i.e. (m, s, k) location(d) Get the location
parameter, i.e. m scale(d) Get the scale parameter, i.e. s shape(d) Get the
shape parameter, i.e. k (sometimes called c)
External links
1.97. DISTRIBUTIONS.GENERALIZEDPARETO 27

• Generalized extreme value distribution on Wikipedia

source

1.97 Distributions.GeneralizedPareto
Distributions.GeneralizedPareto — Type.
GeneralizedPareto(, , )
The Generalized Pareto distribution with shape parameter , scale and loca-
tion has probability density function

x−µ − ξ1 −1
( (
1
σ (1 + ξ σ ) for ξ 6= 0 [µ, ∞] for ξ ≥ 0
f (x; µ, σ, ξ) = 1 − (x−µ)
, x∈
σe
σ for ξ = 0 [µ, µ − σ/ξ] for ξ < 0
[] GeneralizedPareto() Generalized Pareto distribution with unit shape and
unit scale, i.e. GeneralizedPareto(0, 1, 1) GeneralizedPareto(k, s) Generalized
Pareto distribution with shape k and scale s, i.e. GeneralizedPareto(0, k, s)
GeneralizedPareto(m, k, s) Generalized Pareto distribution with shape k, scale
s and location m.
params(d) Get the parameters, i.e. (m, s, k) location(d) Get the location
parameter, i.e. m scale(d) Get the scale parameter, i.e. s shape(d) Get the
shape parameter, i.e. k
External links
• Generalized Pareto distribution on Wikipedia
source

1.98 Distributions.Geometric
Distributions.Geometric — Type.
Geometric(p)
A Geometric distribution characterizes the number of failures before the first
success in a sequence of independent Bernoulli trials with success rate p.

P (X = k) = p(1 − p)k , for k = 0, 1, 2, . . . .

[] Geometric() Geometric distribution with success rate 0.5 Geometric(p)
Geometric distribution with success rate p
params(d) Get the parameters, i.e. (p,) succprob(d) Get the success rate,
i.e. p failprob(d) Get the failure rate, i.e. 1 - p
External links
• Geometric distribution on Wikipedia
source
28 CHAPTER 1. DISTRIBUTIONS

1.99 Distributions.Gumbel
Distributions.Gumbel — Type.

Gumbel(, )

The Gumbel distribution with location and scale has probability density
function

1 −(z+ez ) x−µ
f (x; µ, θ) = e , with z =
θ θ
[] Gumbel() Gumbel distribution with zero location and unit scale, i.e.
Gumbel(0, 1) Gumbel(u) Gumbel distribution with location u and unit scale,
i.e. Gumbel(u, 1) Gumbel(u, b) Gumbel distribution with location u and scale
b
params(d) Get the parameters, i.e. (u, b) location(d) Get the location
parameter, i.e. u scale(d) Get the scale parameter, i.e. b
External links

• Gumbel distribution on Wikipedia

source

1.100 Distributions.Hypergeometric
Distributions.Hypergeometric — Type.

Hypergeometric(s, f, n)

A Hypergeometric distribution describes the number of successes in n draws

without replacement from a finite population containing s successes and f fail-
ures.

s f



k n−k
P (X = k) = s+f

, for k = max(0, n − f ), . . . , min(n, s).
n

[] Hypergeometric(s, f, n) Hypergeometric distribution for a population with

s successes and f failures, and a sequence of n trials.
params(d) Get the parameters, i.e. (s, f, n)
External links

• Hypergeometric distribution on Wikipedia

source
1.101. DISTRIBUTIONS.INVERSEGAMMA 29

1.101 Distributions.InverseGamma
Distributions.InverseGamma — Type.
InverseGamma(, )
The inverse gamma distribution with shape parameter and scale has prob-
ability density function

θα x−(α+1) − θ
f (x; α, θ) = e x, x>0
Γ(α)
It is related to the Gamma distribution: if X ∼ Gamma(α, β), then ‘1 / X
∼ InverseGamma(α, βˆ{−1})“.
[] InverseGamma() Inverse Gamma distribution with unit shape and unit
scale, i.e. InverseGamma(1, 1) InverseGamma(a) Inverse Gamma distribution
with shape a and unit scale, i.e. InverseGamma(a, 1) InverseGamma(a, b)
Inverse Gamma distribution with shape a and scale b
params(d) Get the parameters, i.e. (a, b) shape(d) Get the shape param-
eter, i.e. a scale(d) Get the scale parameter, i.e. b
External links
• Inverse gamma distribution on Wikipedia
source

1.102 Distributions.InverseGaussian
Distributions.InverseGaussian — Type.
InverseGaussian(,)
The inverse Gaussian distribution with mean and shape has probability
density function
r
−λ(x − µ)2
 
λ
f (x; µ, λ) = exp , x>0
2πx3 2µ2 x
[] InverseGaussian() Inverse Gaussian distribution with unit mean and unit
shape, i.e. InverseGaussian(1, 1) InverseGaussian(mu), Inverse Gaussian distri-
bution with mean mu and unit shape, i.e. InverseGaussian(u, 1) InverseGaus-
sian(mu, lambda) Inverse Gaussian distribution with mean mu and shape
lambda
params(d) Get the parameters, i.e. (mu, lambda) mean(d) Get the mean
parameter, i.e. mu shape(d) Get the shape parameter, i.e. lambda
External links
• Inverse Gaussian distribution on Wikipedia
source
30 CHAPTER 1. DISTRIBUTIONS

1.103 Distributions.InverseWishart
Distributions.InverseWishart — Type.
InverseWishart(nu, P)
The [Inverse Wishart distribution](http://en.wikipedia.org/wiki/Inverse-Wishart distribution
is usually used a the conjugate prior for the covariance matrix of a multivariate
normal distribution, which is characterized by a degree of freedom , and a base
matrix .
source

1.104 Distributions.KSDist
Distributions.KSDist — Type.
KSDist(n)
Distribution of the (two-sided) Kolmogorov-Smirnoff statistic
p
Dn = sup |F̂n (x) − F (x)| (n)
x
Dn converges a.s. to the Kolmogorov distribution.
source

1.105 Distributions.KSOneSided
Distributions.KSOneSided — Type.
KSOneSided(n)
Distribution of the one-sided Kolmogorov-Smirnov test statistic:

Dn+ = sup(F̂n (x) − F (x))

x
source

1.106 Distributions.Kolmogorov
Distributions.Kolmogorov — Type.
Kolmogorov()
Kolmogorov distribution defined as

sup |B(t)|
t∈[0,1]

where B(t) is a Brownian bridge used in the Kolmogorov–Smirnov test for

large n.
source
1.107. DISTRIBUTIONS.LAPLACE 31

1.107 Distributions.Laplace
Distributions.Laplace — Type.

Laplace(,)

The Laplace distribution with location and scale has probability density
function
 
1 |x − µ|
f (x; µ, β) = exp −
2β β
[] Laplace() Laplace distribution with zero location and unit scale, i.e.
Laplace(0, 1) Laplace(u) Laplace distribution with location u and unit scale,
i.e. Laplace(u, 1) Laplace(u, b) Laplace distribution with location u ans scale
b
params(d) Get the parameters, i.e. (u, b) location(d) Get the location
parameter, i.e. u scale(d) Get the scale parameter, i.e. b
External links

• Laplace distribution on Wikipedia

source

1.108 Distributions.Levy
Distributions.Levy — Type.

Levy(, )

The Lvy distribution with location and scale has probability density func-
tion
r  
σ σ
f (x; µ, σ) = exp − , x>µ
2π(x − µ)3 2(x − µ)
[] Levy() Levy distribution with zero location and unit scale, i.e. Levy(0,
1) Levy(u) Levy distribution with location u and unit scale, i.e. Levy(u, 1)
Levy(u, c) Levy distribution with location u ans scale c
params(d) Get the parameters, i.e. (u, c) location(d) Get the location
parameter, i.e. u
External links

• Lvy distribution on Wikipedia

source
32 CHAPTER 1. DISTRIBUTIONS

1.109 Distributions.LocationScale
Distributions.LocationScale — Type.

LocationScale(,,)

A location-scale transformed distribution with location parameter , scale

parameter , and given distribution .
 
1 x−
f (x) =

[] LocationScale(,,) location-scale transformed distribution params(d) Get

the parameters, i.e. (, , and the base distribution) location(d) Get the location
parameter scale(d) Get the scale parameter
External links Location-Scale family on Wikipedia
source

1.110 Distributions.LogNormal
Distributions.LogNormal — Type.

LogNormal(,)

The log normal distribution is the distribution of the exponential of a Normal

variate: if X ∼ Normal(µ, σ) then exp(X) ∼ LogNormal(µ, σ). The probability
density function is

(log(x) − µ)2
 
1
f (x; µ, σ) = √ exp − , x>0
x 2πσ 2 2σ 2

[] LogNormal() Log-normal distribution with zero log-mean and unit scale

LogNormal(mu) Log-normal distribution with log-mean mu and unit scale Log-
Normal(mu, sig) Log-normal distribution with log-mean mu and scale sig
params(d) Get the parameters, i.e. (mu, sig) meanlogx(d) Get the mean of
log(X), i.e. mu varlogx(d) Get the variance of log(X), i.e. sig2 stdlogx(d)Getthestandarddeviationof log(X

External links

• Log normal distribution on Wikipedia

source
1.111. DISTRIBUTIONS.LOGISTIC 33

1.111 Distributions.Logistic
Distributions.Logistic — Type.

Logistic(,)

The Logistic distribution with location and scale has probability density
function
 
1 2 x−µ
f (x; µ, θ) = sech
4θ 2θ
[] Logistic() Logistic distribution with zero location and unit scale, i.e.
Logistic(0, 1) Logistic(u) Logistic distribution with location u and unit scale,
i.e. Logistic(u, 1) Logistic(u, b) Logistic distribution with location u ans scale
b
params(d) Get the parameters, i.e. (u, b) location(d) Get the location
parameter, i.e. u scale(d) Get the scale parameter, i.e. b
External links

• Logistic distribution on Wikipedia

source

1.112 Distributions.MixtureModel
Distributions.MixtureModel — Method.

MixtureModel(components, [prior])

Construct a mixture model with a vector of components and a prior prob-

ability vector. If no prior is provided then all components will have the same
prior probabilities.
source

1.113 Distributions.MixtureModel
Distributions.MixtureModel — Method.

MixtureModel(C, params, [prior])

Construct a mixture model with component type C, a vector of parame-

ters for constructing the components given by params, and a prior probability
vector. If no prior is provided then all components will have the same prior
probabilities.
source
34 CHAPTER 1. DISTRIBUTIONS

1.114 Distributions.Multinomial
Distributions.Multinomial — Type.
The Multinomial distribution generalizes the binomial distribution. Consider
n independent draws from a Categorical distribution over a finite set of size k,
and let X = (X1 , ..., Xk ) where Xi represents the number of times the element
i occurs, then the distribution of X is a multinomial distribution. Each sample
of a multinomial distribution is a k-dimensional integer vector that sums to n.
The probability mass function is given by
k
n! Y
f (x; n, p) = pxi , x1 + · · · + xk = n
x1 ! · · · xk ! i=1 i
[] Multinomial(n, p) Multinomial distribution for n trials with probability
vector p Multinomial(n, k) Multinomial distribution for n trials with equal
probabilities over 1:k
source

1.115 Distributions.MvLogNormal
Distributions.MvLogNormal — Type.

MvLogNormal(d::MvNormal)

The Multivariate lognormal distribution is a multidimensional generalization

of the lognormal distribution.
If X ∼ N (µ, Σ) has a multivariate normal distribution then Y = exp(X)
has a multivariate lognormal distribution.
Mean vector µ and covariance matrix Σ of the underlying normal distri-
bution are known as the location and scale parameters of the corresponding
lognormal distribution.
source

1.116 Distributions.MvNormal
Distributions.MvNormal — Type.

MvNormal

Generally, users don’t have to worry about these internal details. We pro-
vide a common constructor MvNormal, which will construct a distribution of
appropriate type depending on the input arguments.

MvNormal(sig)

Construct a multivariate normal distribution with zero mean and covariance

represented by sig.
1.117. DISTRIBUTIONS.MVNORMALCANON 35

MvNormal(mu, sig)

Construct a multivariate normal distribution with mean mu and covariance

represented by sig.

MvNormal(d, sig)

Construct a multivariate normal distribution of dimension d, with zero mean,

and an isotropic covariance as abs2(sig) * eye(d).
Arguments

• mu::Vector{T<:Real}: The mean vector.

• d::Real: dimension of distribution.
• sig: The covariance, which can in of either of the following forms (with
T<:Real):

1. subtype of AbstractPDMat
2. symmetric matrix of type Matrix{T}
3. vector of type Vector{T}: indicating a diagonal covariance as diagm(abs2(sig)).
4. real-valued number: indicating an isotropic covariance as abs2(sig)
* eye(d).

Note: The constructor will choose an appropriate covariance form inter-

nally, so that special structure of the covariance can be exploited.
source

1.117 Distributions.MvNormalCanon
Distributions.MvNormalCanon — Type.

MvNormalCanon

Multivariate normal distribution is an exponential family distribution, with

two canonical parameters: the potential vector h and the precision matrix J.
The relation between these parameters and the conventional representation (i.e.
the one using mean µ and covariance Σ) is:

h = Σ−1 µ, and J = Σ−1

The canonical parameterization is widely used in Bayesian analysis. We pro-
vide a type MvNormalCanon, which is also a subtype of AbstractMvNormal to
represent a multivariate normal distribution using canonical parameters. Par-
ticularly, MvNormalCanon is defined as:
[] immutable MvNormalCanon{P¡:AbstractPDMat,V¡:Union{Vector,ZeroVector}}
¡: AbstractMvNormal ::V the mean vector h::V potential vector, i.e. inv() *
J::P precision matrix, i.e. inv() end
36 CHAPTER 1. DISTRIBUTIONS

We also define aliases for common specializations of this parametric type:

[] const FullNormalCanon = MvNormalCanon{PDMat, Vector{Float64}}
const DiagNormalCanon = MvNormalCanon{PDiagMat, Vector{Float64}} const
IsoNormalCanon = MvNormalCanon{ScalMat, Vector{Float64}}
const ZeroMeanFullNormalCanon = MvNormalCanon{PDMat, ZeroVector{Float64}}
const ZeroMeanDiagNormalCanon = MvNormalCanon{PDiagMat, ZeroVector{Float64}}
const ZeroMeanIsoNormalCanon = MvNormalCanon{ScalMat, ZeroVector{Float64}}

A multivariate distribution with canonical parameterization can be con-

structed using a common constructor MvNormalCanon as:

MvNormalCanon(h, J)

Construct a multivariate normal distribution with potential vector h and

precision matrix represented by J.

MvNormalCanon(J)

Construct a multivariate normal distribution with zero mean (thus zero po-
tential vector) and precision matrix represented by J.

MvNormalCanon(d, J)

Construct a multivariate normal distribution of dimension d, with zero mean

and a precision matrix as J * eye(d).
Arguments

• d::Int: dimension of distribution

• h::Vector{T<:Real}: the potential vector, of type Vector{T} with T<:Real.

• J: the representation of the precision matrix, which can be in either of the

following forms (T<:Real):

1. an instance of a subtype of AbstractPDMat

2. a square matrix of type Matrix{T}
3. a vector of type Vector{T}: indicating a diagonal precision matrix
as diagm(J).
4. a real number: indicating an isotropic precision matrix as J * eye(d).

Note: MvNormalCanon share the same set of methods as MvNormal.

source
1.118. DISTRIBUTIONS.NEGATIVEBINOMIAL 37

1.118 Distributions.NegativeBinomial
Distributions.NegativeBinomial — Type.

NegativeBinomial(r,p)

A Negative binomial distribution describes the number of failures before the

rth success in a sequence of independent Bernoulli trials. It is parameterized
by r, the number of successes, and p, the probability of success in an individual
trial.
 
k+r−1 r
P (X = k) = p (1 − p)k , for k = 0, 1, 2, . . . .
k
The distribution remains well-defined for any positive r, in which case

Γ(k + r) r
P (X = k) = p (1 − p)k , for k = 0, 1, 2, . . . .
k!Γ(r)
[] NegativeBinomial() Negative binomial distribution with r = 1 and p =
0.5 NegativeBinomial(r, p) Negative binomial distribution with r successes and
success rate p
params(d) Get the parameters, i.e. (r, p) succprob(d) Get the success rate,
i.e. p failprob(d) Get the failure rate, i.e. 1 - p
External links:

• Negative binomial distribution on Wikipedia

source

1.119 Distributions.NoncentralBeta
Distributions.NoncentralBeta — Type.

NoncentralBeta(, , )

source

1.120 Distributions.NoncentralChisq
Distributions.NoncentralChisq — Type.

NoncentralChisq(, )

The noncentral chi-squared distribution with degrees of freedom and non-

centrality parameter has the probability density function

1 −(x+λ)/2  x ν/4−1/2 √
f (x; ν, λ) = e Iν/2−1 ( λx), x>0
2 λ
38 CHAPTER 1. DISTRIBUTIONS

It is the distribution of the sum of squares of independent Normal variates

with individual means µi and
ν
X
λ= µ2i
i=1
[] NoncentralChisq(, ) Noncentral chi-squared distribution with degrees of
freedom and noncentrality parameter
params(d) Get the parameters, i.e. (, )
External links
• Noncentral chi-squared distribution on Wikipedia
source

1.121 Distributions.NoncentralF
Distributions.NoncentralF — Type.
NoncentralF(1, 2, )
source

1.122 Distributions.NoncentralT
Distributions.NoncentralT — Type.
NoncentralT(, )
source

1.123 Distributions.Normal
Distributions.Normal — Type.
Normal(,)
The Normal distribution with mean and standard deviation has probability
density function

(x − µ)2
 
1
f (x; µ, σ) = √ exp −
2πσ 2 2σ 2
[] Normal() standard Normal distribution with zero mean and unit variance
Normal(mu) Normal distribution with mean mu and unit variance Normal(mu,
sig) Normal distribution with mean mu and variance sig2
params(d) Get the parameters, i.e. (mu, sig) mean(d) Get the mean, i.e.
mu std(d) Get the standard deviation, i.e. sig
External links
1.124. DISTRIBUTIONS.NORMALCANON 39

• Normal distribution on Wikipedia

source

1.124 Distributions.NormalCanon
Distributions.NormalCanon — Type.
NormalCanon(, )
Canonical Form of Normal distribution
source

1.125 Distributions.NormalInverseGaussian
Distributions.NormalInverseGaussian — Type.
NormalInverseGaussian(,,,)
The Normal-inverse Gaussian distribution with location , tail heaviness ,
asymmetry parameter and scale has probability density function
 p 
αδK1 α δ 2 + (x − µ)2
f (x; µ, α, β, δ) = p eδγ+β(x−µ)
π δ 2 + (x − µ)2
where Kj denotes a modified Bessel function of the third kind.
External links
• Normal-inverse Gaussian distribution on Wikipedia
source

1.126 Distributions.Pareto
Distributions.Pareto — Type.
Pareto(,)
The Pareto distribution with shape and scale has probability density func-
tion
αθα
f (x; α, θ) = , x≥θ
xα+1
[] Pareto() Pareto distribution with unit shape and unit scale, i.e. Pareto(1,
1) Pareto(a) Pareto distribution with shape a and unit scale, i.e. Pareto(a, 1)
Pareto(a, b) Pareto distribution with shape a and scale b
params(d) Get the parameters, i.e. (a, b) shape(d) Get the shape param-
eter, i.e. a scale(d) Get the scale parameter, i.e. b
External links
40 CHAPTER 1. DISTRIBUTIONS

• Pareto distribution on Wikipedia

source

1.127 Distributions.Poisson
Distributions.Poisson — Type.
Poisson()
A Poisson distribution descibes the number of independent events occurring
within a unit time interval, given the average rate of occurrence .

λk −λ
P (X = k) = e , for k = 0, 1, 2, . . . .
k!
[] Poisson() Poisson distribution with rate parameter 1 Poisson(lambda)
Poisson distribution with rate parameter lambda
params(d) Get the parameters, i.e. (,) mean(d) Get the mean arrival rate,
i.e.
External links:
• Poisson distribution on Wikipedia
source

1.128 Distributions.PoissonBinomial
Distributions.PoissonBinomial — Type.
PoissonBinomial(p)
A Poisson-binomial distribution describes the number of successes in a se-
quence of independent trials, wherein each trial has a different success rate. It
is parameterized by a vector p (of length K), where K is the total number of
trials and p[i] corresponds to the probability of success of the ith trial.

X Y Y
P (X = k) = p[i] (1 − p[j]), for k = 0, 1, 2, . . . , K,
A∈Fk i∈A j∈Ac

where Fk is the set of all subsets of k integers that can be selected from
{1, 2, 3, ..., K}.
[] PoissonBinomial(p) Poisson Binomial distribution with success rate vector
p
params(d) Get the parameters, i.e. (p,) succprob(d) Get the vector of
success rates, i.e. p failprob(d) Get the vector of failure rates, i.e. 1-p
External links:
• Poisson-binomial distribution on Wikipedia
source
1.129. DISTRIBUTIONS.RAYLEIGH 41

1.129 Distributions.Rayleigh
Distributions.Rayleigh — Type.

Rayleigh()

The Rayleigh distribution with scale has probability density function

x − x22
f (x; σ) = e 2σ , x>0
σ2
It is related to the Normal distribution
√ via the property that if X, Y ∼
Normal(0, σ), independently, then X 2 + Y 2 ∼ Rayleigh(σ).
[] Rayleigh() Rayleigh distribution with unit scale, i.e. Rayleigh(1) Rayleigh(s)
Rayleigh distribution with scale s
params(d) Get the parameters, i.e. (s,) scale(d) Get the scale parameter,
i.e. s
External links

• Rayleigh distribution on Wikipedia

source

1.130 Distributions.Semicircle
Distributions.Semicircle — Type.

Semicircle(r)

The Wigner semicircle distribution with radius parameter r has probability

density function

2 p 2
f (x; r) = r − x2 , x ∈ [−r, r].
πr2
[] Semicircle(r) Wigner semicircle distribution with radius r
params(d) Get the radius parameter, i.e. (r,)
External links

• Wigner semicircle distribution on Wikipedia

source
42 CHAPTER 1. DISTRIBUTIONS

1.131 Distributions.Skellam
Distributions.Skellam — Type.

Skellam(1, 2)

A Skellam distribution describes the difference between two independent

Poisson variables, respectively with rate 1 and 2.
 k/2
µ1 √
P (X = k) = e−(µ1 +µ2 ) Ik (2 µ1 µ2 ) for integer k
µ2
where Ik is the modified Bessel function of the first kind.
[] Skellam(mu1, mu2) Skellam distribution for the difference between two
Poisson variables, respectively with expected values mu1 and mu2.
params(d) Get the parameters, i.e. (mu1, mu2)
External links:

• Skellam distribution on Wikipedia

source

1.132 Distributions.SymTriangularDist
Distributions.SymTriangularDist — Type.

SymTriangularDist(,)

The Symmetric triangular distribution with location and scale has proba-
bility density function
  
1 x − µ
f (x; µ, σ) = 1−
  , µ−σ ≤x≤µ+σ
σ σ 
[] SymTriangularDist() Symmetric triangular distribution with zero loca-
tion and unit scale SymTriangularDist(u) Symmetric triangular distribution
with location u and unit scale SymTriangularDist(u, s) Symmetric triangular
distribution with location u and scale s
params(d) Get the parameters, i.e. (u, s) location(d) Get the location
parameter, i.e. u scale(d) Get the scale parameter, i.e. s
source

1.133 Distributions.TDist
Distributions.TDist — Type.

TDist()
1.134. DISTRIBUTIONS.TRIANGULARDIST 43

The Students T distribution with degrees of freedom has probability density

function
− d+1
x2

1 2

f (x; d) = √ 1+
dB(1/2, d/2) d
[] TDist(d) t-distribution with d degrees of freedom
params(d) Get the parameters, i.e. (d,) dof(d) Get the degrees of freedom,
i.e. d
External links
Student’s T distribution on Wikipedia
source

1.134 Distributions.TriangularDist
Distributions.TriangularDist — Type.

TriangularDist(a,b,c)

The triangular distribution with lower limit a, upper limit b and mode c has
probability density function



0 for x < a,
2(x−a)
for a ≤ x ≤ c,


 (b−a)(c−a)
f (x; a, b, c) = 2(b−x)


 (b−a)(b−c) for c < x ≤ b,


0 for b < x,
[] TriangularDist(a, b) Triangular distribution with lower limit a, upper
limit b, and mode (a+b)/2 TriangularDist(a, b, c) Triangular distribution with
lower limit a, upper limit b, and mode c
params(d) Get the parameters, i.e. (a, b, c) minimum(d) Get the lower
bound, i.e. a maximum(d) Get the upper bound, i.e. b mode(d) Get the mode,
i.e. c
External links

• Triangular distribution on Wikipedia

source

1.135 Distributions.Triweight
Distributions.Triweight — Type.

Triweight(, )

source
44 CHAPTER 1. DISTRIBUTIONS

1.136 Distributions.Truncated
Distributions.Truncated — Type.

Truncated(d, l, u):

Construct a truncated distribution.

Arguments

• d::UnivariateDistribution: The original distribution.

• l::Real: The lower bound of the truncation, which can be a finite value
or -Inf.

• u::Real: The upper bound of the truncation, which can be a finite value
of Inf.

source

1.137 Distributions.Uniform
Distributions.Uniform — Type.

Uniform(a,b)

The continuous uniform distribution over an interval [a, b] has probability

density function

1
f (x; a, b) = , a≤x≤b
b−a
[] Uniform() Uniform distribution over [0, 1] Uniform(a, b) Uniform distri-
bution over [a, b]
params(d) Get the parameters, i.e. (a, b) minimum(d) Get the lower
bound, i.e. a maximum(d) Get the upper bound, i.e. b location(d) Get the
location parameter, i.e. a scale(d) Get the scale parameter, i.e. b - a
External links

• Uniform distribution (continuous) on Wikipedia

source

1.138 Distributions.VonMises
Distributions.VonMises — Type.

VonMises(, )
1.139. DISTRIBUTIONS.WEIBULL 45

The von Mises distribution with mean and concentration has probability
density function
1
f (x; µ, κ) = exp (κ cos(x − µ))
2πI0 (κ)
[] VonMises() von Mises distribution with zero mean and unit concentration
VonMises() von Mises distribution with zero mean and concentration Von-
Mises(, ) von Mises distribution with mean and concentration
External links

• von Mises distribution on Wikipedia

source

1.139 Distributions.Weibull
Distributions.Weibull — Type.

Weibull(,)

The Weibull distribution with shape and scale has probability density
function
α  x α−1 −(x/θ)α
f (x; α, θ) = e , x≥0
θ θ
[] Weibull() Weibull distribution with unit shape and unit scale, i.e. Weibull(1,
1) Weibull(a) Weibull distribution with shape a and unit scale, i.e. Weibull(a,
1) Weibull(a, b) Weibull distribution with shape a and scale b
params(d) Get the parameters, i.e. (a, b) shape(d) Get the shape param-
eter, i.e. a scale(d) Get the scale parameter, i.e. b
External links

• Weibull distribution on Wikipedia

source

1.140 Distributions.Wishart
Distributions.Wishart — Type.

Wishart(nu, S)

The Wishart distribution is a multidimensional generalization of the Chi-

square distribution, which is characterized by a degree of freedom , and a base
matrix S.
source
Chapter 2

StatsBase

2.1 Base.stdm
Base.stdm — Method.

stdm(v, w::AbstractWeights, m, [dim]; corrected=false)

Compute the standard deviation of a real-valued array x with a known mean

m, optionally over a dimension dim. Observations in x are weighted using weight
vector w. The uncorrected (when corrected=false) sample standard deviation
is defined as:
v
u n
u 1 X 2
tP wi (xi − m)
w i=1

where n is the length of the input. The unbiased estimate (when corrected=true)
of the population standard deviation is computed by replacing P1w with a factor
dependent on the type of weights used:

• AnalyticWeights: P P1 2 P
w− w / w

• FrequencyWeights: P1
w−1

nP
• ProbabilityWeights: (n−1) w where n equals count(!iszero, w)

• Weights: ArgumentError (bias correction not supported)

source

46
2.2. BASE.VARM 47

2.2 Base.varm
Base.varm — Method.

varm(x, w::AbstractWeights, m, [dim]; corrected=false)

Compute the variance of a real-valued array x with a known mean m, option-

ally over a dimension dim. Observations in x are weighted using weight vector
w. The uncorrected (when corrected=false) sample variance is defined as:
n
1 X 2
P wi (xi − m)
w i=1

where n is the length of the input. The unbiased estimate (when corrected=true)
of the population variance is computed by replacing P1w with a factor dependent
on the type of weights used:

• AnalyticWeights: P P1 2 P
w− w / w

• FrequencyWeights: P1
w−1

nP
• ProbabilityWeights: (n−1) w where n equals count(!iszero, w)

• Weights: ArgumentError (bias correction not supported)

source

2.3 StatsBase.L1dist
StatsBase.L1dist — Method.

L1dist(a, b)
Pn
Compute the L1 distance between two arrays: i=1 |ai −bi |. Efficient equiv-
alent of sum(abs, a - b).
source

2.4 StatsBase.L2dist
StatsBase.L2dist — Method.

L2dist(a, b)
pPn
Compute the L2 distance between two arrays: i=1 |ai − bi |2 . Efficient
equivalent of sqrt(sumabs2(a - b)).
source
48 CHAPTER 2. STATSBASE

2.5 StatsBase.Linfdist
StatsBase.Linfdist — Method.

Linfdist(a, b)

Compute the L distance, also called the Chebyshev distance, between two
arrays: maxi∈1:n |ai − bi |. Efficient equivalent of maxabs(a - b).
source

2.6 StatsBase.addcounts!
StatsBase.addcounts! — Method.

addcounts!(r, x, levels::UnitRange{<:Int}, [wv::AbstractWeights])

Add the number of occurrences in x of each value in levels to an existing

array r. If a weighting vector wv is specified, the sum of weights is used rather
than the raw counts.
source

2.7 StatsBase.addcounts!
StatsBase.addcounts! — Method.

addcounts!(dict, x[, wv]; alg = :auto)

Add counts based on x to a count map. New entries will be added if new
values come up. If a weighting vector wv is specified, the sum of the weights is
used rather than the raw counts.
alg can be one of:

• :auto (default): if StatsBase.radixsort safe(eltype(x)) == true then

use :radixsort, otherwise use :dict.

• :radixsort: if radixsort safe(eltype(x)) == true then use the radix

sort algorithm to sort the input vector which will generally lead to shorter
running time. However the radix sort algorithm creates a copy of the
input vector and hence uses more RAM. Choose :dict if the amount of
available RAM is a limitation.

• :dict: use Dict-based method which is generally slower but uses less
RAM and is safe for any data type.

source
2.8. STATSBASE.ADJR2 49

2.8 StatsBase.adjr2
StatsBase.adjr2 — Method.

adjr2(obj::StatisticalModel, variant::Symbol)
adjr(obj::StatisticalModel, variant::Symbol)

Adjusted pseudo-coefficient of determination (adjusted pseudo R-squared).

For nonlinear models, one of the several pseudo R definitions must be chosen
via variant. The only currently supported variant is :MacFadden, defined as
1 − (log L − k)/ log L0. In this formula, L is the likelihood of the model, L0
that of the null model (the model including only the intercept). These two
quantities are taken to be minus half deviance of the corresponding models.
k is the number of consumed degrees of freedom of the model (as returned by
dof).
source

2.9 StatsBase.adjr2
StatsBase.adjr2 — Method.

adjr2(obj::StatisticalModel)
adjr(obj::StatisticalModel)

Adjusted coefficient of determination (adjusted R-squared).

For linear models, the adjusted R is defined as 1−(1−(1−R2 )(n−1)/(n−p)),
with R2 the coefficient of determination, n the number of observations, and p
the number of coefficients (including the intercept). This definition is generally
known as the Wherry Formula I.
source

2.10 StatsBase.aic
StatsBase.aic — Method.

aic(obj::StatisticalModel)

Akaike’s Information Criterion, defined as −2 log L+2k, with L the likelihood

of the model, and k its number of consumed degrees of freedom (as returned by
dof).
source
50 CHAPTER 2. STATSBASE

2.11 StatsBase.aicc
StatsBase.aicc — Method.

aicc(obj::StatisticalModel)

Corrected Akaike’s Information Criterion for small sample sizes (Hurvich and
Tsai 1989), defined as −2 log L+2k +2k(k −1)/(n−k −1), with L the likelihood
of the model, k its number of consumed degrees of freedom (as returned by dof),
and n the number of observations (as returned by nobs).
source

2.12 StatsBase.autocor!
StatsBase.autocor! — Method.

autocor!(r, x, lags; demean=true)

Compute the autocorrelation function (ACF) of a vector or matrix x at

lags and store the result in r. demean denotes whether the mean of x should
be subtracted from x before computing the ACF.
If x is a vector, r must be a vector of the same length as x. If x is a matrix, r
must be a matrix of size (length(lags), size(x,2)), and where each column
in the result will correspond to a column in x.
The output is normalized by the variance of x, i.e. so that the lag 0 auto-
correlation is 1. See autocov! for the unnormalized form.
source

2.13 StatsBase.autocor
StatsBase.autocor — Method.

autocor(x, [lags]; demean=true)

Compute the autocorrelation function (ACF) of a vector or matrix x, op-

tionally specifying the lags. demean denotes whether the mean of x should be
subtracted from x before computing the ACF.
If x is a vector, return a vector of the same length as x. If x is a matrix,
return a matrix of size (length(lags), size(x,2)), where each column in the
result corresponds to a column in x.
When left unspecified, the lags used are the integers from 0 to min(size(x,1)-1,
10*log10(size(x,1))).
The output is normalized by the variance of x, i.e. so that the lag 0 auto-
correlation is 1. See autocov for the unnormalized form.
source
2.14. STATSBASE.AUTOCOV! 51

2.14 StatsBase.autocov!
StatsBase.autocov! — Method.

autocov!(r, x, lags; demean=true)

Compute the autocovariance of a vector or matrix x at lags and store the

result in r. demean denotes whether the mean of x should be subtracted from
x before computing the autocovariance.
If x is a vector, r must be a vector of the same length as x. If x is a matrix, r
must be a matrix of size (length(lags), size(x,2)), and where each column
in the result will correspond to a column in x.
The output is not normalized. See autocor! for a method with normaliza-
tion.
source

2.15 StatsBase.autocov
StatsBase.autocov — Method.

autocov(x, [lags]; demean=true)

Compute the autocovariance of a vector or matrix x, optionally specifying

the lags at which to compute the autocovariance. demean denotes whether the
mean of x should be subtracted from x before computing the autocovariance.
If x is a vector, return a vector of the same length as x. If x is a matrix,
return a matrix of size (length(lags), size(x,2)), where each column in the
result corresponds to a column in x.
When left unspecified, the lags used are the integers from 0 to min(size(x,1)-1,
10*log10(size(x,1))).
The output is not normalized. See autocor for a function with normaliza-
tion.
source

2.16 StatsBase.aweights
StatsBase.aweights — Method.

aweights(vs)

Construct an AnalyticWeights vector from array vs. See the documenta-

tion for AnalyticWeights for more details.
source
52 CHAPTER 2. STATSBASE

2.17 StatsBase.bic
StatsBase.bic — Method.

bic(obj::StatisticalModel)

Bayesian Information Criterion, defined as −2 log L+k log n, with L the like-
lihood of the model, k its number of consumed degrees of freedom (as returned
by dof), and n the number of observations (as returned by nobs).
source

2.18 StatsBase.coef
StatsBase.coef — Method.

coef(obj::StatisticalModel)

Return the coefficients of the model.

source

2.19 StatsBase.coefnames
StatsBase.coefnames — Method.

coefnames(obj::StatisticalModel)

Return the names of the coefficients.

source

2.20 StatsBase.coeftable
StatsBase.coeftable — Method.

coeftable(obj::StatisticalModel)

Return a table of class CoefTable with coefficients and related statistics.

source

2.21 StatsBase.competerank
StatsBase.competerank — Method.

competerank(x; lt = isless, rev::Bool = false)

2.22. STATSBASE.CONFINT 53

Return the standard competition ranking (“1224” ranking) of an array. The

lt keyword allows providing a custom “less than” function; use rev=true to
reverse the sorting order. Items that compare equal are given the same rank,
then a gap is left in the rankings the size of the number of tied items - 1. Missing
values are assigned rank missing.
source

2.22 StatsBase.confint
StatsBase.confint — Method.

confint(obj::StatisticalModel)

Compute confidence intervals for coefficients.

source

2.23 StatsBase.cor2cov
StatsBase.cor2cov — Method.

cor2cov(C, s)

Compute the covariance matrix from the correlation matrix C and a vector
of standard deviations s. Use StatsBase.cor2cov! for an in-place version.
source

2.24 StatsBase.corkendall
StatsBase.corkendall — Method.

corkendall(x, y=x)

Compute Kendall’s rank correlation coefficient, . x and y must both be either

matrices or vectors.
source

2.25 StatsBase.corspearman
StatsBase.corspearman — Method.

corspearman(x, y=x)

Compute Spearman’s rank correlation coefficient. If x and y are vectors,

the output is a float, otherwise it’s a matrix corresponding to the pairwise
correlations of the columns of x and y.
source
54 CHAPTER 2. STATSBASE

2.26 StatsBase.counteq
StatsBase.counteq — Method.

counteq(a, b)

Count the number of indices at which the elements of the arrays a and b are
equal.
source

2.27 StatsBase.countmap
StatsBase.countmap — Method.

countmap(x; alg = :auto)

Return a dictionary mapping each unique value in x to its number of occur-

rences.

• :auto (default): if StatsBase.radixsort safe(eltype(x)) == true then

use :radixsort, otherwise use :dict.

• :radixsort: if radixsort safe(eltype(x)) == true then use the radix

sort algorithm to sort the input vector which will generally lead to shorter
running time. However the radix sort algorithm creates a copy of the
input vector and hence uses more RAM. Choose :dict if the amount of
available RAM is a limitation.

• :dict: use Dict-based method which is generally slower but uses less
RAM and is safe for any data type.

source

2.28 StatsBase.countne
StatsBase.countne — Method.

countne(a, b)

Count the number of indices at which the elements of the arrays a and b are
not equal.
source
2.29. STATSBASE.COUNTS 55

2.29 StatsBase.counts
StatsBase.counts — Function.

counts(x, [wv::AbstractWeights])
counts(x, levels::UnitRange{<:Integer}, [wv::AbstractWeights])
counts(x, k::Integer, [wv::AbstractWeights])

Count the number of times each value in x occurs. If levels is provided,

only values falling in that range will be considered (the others will be ignored
without raising an error or a warning). If an integer k is provided, only values
in the range 1:k will be considered.
If a weighting vector wv is specified, the sum of the weights is used rather
than the raw counts.
The output is a vector of length length(levels).
source

2.30 StatsBase.cov2cor
StatsBase.cov2cor — Method.

cov2cor(C, s)

Compute the correlation matrix from the covariance matrix C and a vector
of standard deviations s. Use StatsBase.cov2cor! for an in-place version.
source

2.31 StatsBase.crosscor!
StatsBase.crosscor! — Method.

crosscor!(r, x, y, lags; demean=true)

Compute the cross correlation between real-valued vectors or matrices x

and y at lags and store the result in r. demean specifies whether the respective
means of x and y should be subtracted from them before computing their cross
correlation.
If both x and y are vectors, r must be a vector of the same length as lags. If
either x is a matrix and y is a vector, r must be a matrix of size (length(lags),
size(x, 2)); if x is a vector and y is a matrix, r must be a matrix of size
(length(lags), size(y, 2)). If both x and y are matrices, r must be a
three-dimensional array of size (length(lags), size(x, 2), size(y, 2)).
The output is normalized by sqrt(var(x)*var(y)). See crosscov! for the
unnormalized form.
source
56 CHAPTER 2. STATSBASE

2.32 StatsBase.crosscor
StatsBase.crosscor — Method.

crosscor(x, y, [lags]; demean=true)

Compute the cross correlation between real-valued vectors or matrices x

and y, optionally specifying the lags. demean specifies whether the respective
means of x and y should be subtracted from them before computing their cross
correlation.
If both x and y are vectors, return a vector of the same length as lags.
Otherwise, compute cross covariances between each pairs of columns in x and
y.
When left unspecified, the lags used are the integers from -min(size(x,1)-1,
10*log10(size(x,1))) to min(size(x,1), 10*log10(size(x,1))).
The output is normalized by sqrt(var(x)*var(y)). See crosscov for the
unnormalized form.
source

2.33 StatsBase.crosscov!
StatsBase.crosscov! — Method.

crosscov!(r, x, y, lags; demean=true)

Compute the cross covariance function (CCF) between real-valued vectors or

matrices x and y at lags and store the result in r. demean specifies whether the
respective means of x and y should be subtracted from them before computing
their CCF.
If both x and y are vectors, r must be a vector of the same length as lags. If
either x is a matrix and y is a vector, r must be a matrix of size (length(lags),
size(x, 2)); if x is a vector and y is a matrix, r must be a matrix of size
(length(lags), size(y, 2)). If both x and y are matrices, r must be a
three-dimensional array of size (length(lags), size(x, 2), size(y, 2)).
The output is not normalized. See crosscor! for a function with normal-
ization.
source

2.34 StatsBase.crosscov
StatsBase.crosscov — Method.

crosscov(x, y, [lags]; demean=true)

2.35. STATSBASE.CROSSENTROPY 57

Compute the cross covariance function (CCF) between real-valued vectors or

matrices x and y, optionally specifying the lags. demean specifies whether the
respective means of x and y should be subtracted from them before computing
their CCF.
If both x and y are vectors, return a vector of the same length as lags.
Otherwise, compute cross covariances between each pairs of columns in x and
y.
When left unspecified, the lags used are the integers from -min(size(x,1)-1,
10*log10(size(x,1))) to min(size(x,1), 10*log10(size(x,1))).
The output is not normalized. See crosscor for a function with normaliza-
tion.
source

2.35 StatsBase.crossentropy
StatsBase.crossentropy — Method.

crossentropy(p, q, [b])

Compute the cross entropy between p and q, optionally specifying a real

number b such that the result is scaled by 1/log(b).
source

2.36 StatsBase.denserank
StatsBase.denserank — Method.

denserank(x)

Return the dense ranking (“1223” ranking) of an array. The lt keyword

allows providing a custom “less than” function; use rev=true to reverse the
sorting order. Items that compare equal receive the same ranking, and the
next subsequent rank is assigned with no gap. Missing values are assigned rank
missing.
source

2.37 StatsBase.describe
StatsBase.describe — Method.

describe(a)

Pretty-print the summary statistics provided by summarystats: the mean,

minimum, 25th percentile, median, 75th percentile, and maximum.
source
58 CHAPTER 2. STATSBASE

2.38 StatsBase.deviance
StatsBase.deviance — Method.

deviance(obj::StatisticalModel)

Return the deviance of the model relative to a reference, which is usually

when applicable the saturated model. It is equal, up to a constant, to −2 log L,
with L the likelihood of the model.
source

2.39 StatsBase.dof
StatsBase.dof — Method.

dof(obj::StatisticalModel)

Return the number of degrees of freedom consumed in the model, including

when applicable the intercept and the distribution’s dispersion parameter.
source

2.40 StatsBase.dof residual

StatsBase.dof residual — Method.

dof_residual(obj::RegressionModel)

Return the residual degrees of freedom of the model.

source

2.41 StatsBase.ecdf
StatsBase.ecdf — Method.

ecdf(X)

Return an empirical cumulative distribution function (ECDF) based on a

vector of samples given in X.
Note: this is a higher-level function that returns a function, which can then
be applied to evaluate CDF values on other samples.
source
2.42. STATSBASE.ENTROPY 59

2.42 StatsBase.entropy
StatsBase.entropy — Method.

entropy(p, [b])

Compute the entropy of an array p, optionally specifying a real number b

such that the entropy is scaled by 1/log(b).
source

2.43 StatsBase.findat
StatsBase.findat — Method.

findat(a, b)

For each element in b, find its first index in a. If the value does not occur
in a, the corresponding index is 0.
source

2.44 StatsBase.fit!
StatsBase.fit! — Method.
Fit a statistical model in-place.
source

2.45 StatsBase.fit
StatsBase.fit — Method.
Fit a statistical model.
source

2.46 StatsBase.fit
StatsBase.fit — Method.

fit(Histogram, data[, weight][, edges]; closed=:right, nbins)

Fit a histogram to data.

Arguments

• data: either a vector (for a 1-dimensional histogram), or a tuple of vectors

of equal length (for an n-dimensional histogram).
60 CHAPTER 2. STATSBASE

• weight: an optional AbstractWeights (of the same length as the data

vectors), denoting the weight each observation contributes to the bin. If
no weight vector is supplied, each observation has weight 1.

• edges: a vector (typically an AbstractRange object), or tuple of vectors,

that gives the edges of the bins along each dimension. If no edges are
provided, these are determined from the data.

Keyword arguments

• closed=:right: if :left, the bin intervals are left-closed [a,b); if :right

(the default), intervals are right-closed (a,b].

• nbins: if no edges argument is supplied, the approximate number of bins

to use along each dimension (can be either a single integer, or a tuple of
integers).

Examples
[] Univariate h = fit(Histogram, rand(100)) h = fit(Histogram, rand(100),
0:0.1:1.0) h = fit(Histogram, rand(100), nbins=10) h = fit(Histogram, rand(100),
weights(rand(100)), 0:0.1:1.0) h = fit(Histogram, [20], 0:20:100) h = fit(Histogram,
[20], 0:20:100, closed=:left)
Multivariate h = fit(Histogram, (rand(100),rand(100))) h = fit(Histogram,
(rand(100),rand(100)),nbins=10)
source

2.47 StatsBase.fitted
StatsBase.fitted — Method.

fitted(obj::RegressionModel)

Return the fitted values of the model.

source

2.48 StatsBase.fweights
StatsBase.fweights — Method.

fweights(vs)

Construct a FrequencyWeights vector from a given array. See the docu-

mentation for FrequencyWeights for more details.
source
2.49. STATSBASE.GENMEAN 61

2.49 StatsBase.genmean
StatsBase.genmean — Method.
genmean(a, p)
Return the generalized/power mean with exponent p of a real-valued array,
Pn
1
i.e. n1 i=1 api p , where n = length(a). It is taken to be the geometric mean
when p == 0.
source

2.50 StatsBase.geomean
StatsBase.geomean — Method.
geomean(a)
Return the geometric mean of a real-valued array.
source

2.51 StatsBase.gkldiv
StatsBase.gkldiv — Method.
gkldiv(a, b)

PnCompute the generalized Kullback-Leibler divergence between two arrays:

i=1 (ai log(ai /bi ) − ai + bi ). Efficient equivalent of sum(a*log(a/b)-a+b).
source

2.52 StatsBase.harmmean
StatsBase.harmmean — Method.
harmmean(a)
Return the harmonic mean of a real-valued array.
source

2.53 StatsBase.indexmap
StatsBase.indexmap — Method.
indexmap(a)
Construct a dictionary that maps each unique value in a to the index of its
first occurrence in a.
source
62 CHAPTER 2. STATSBASE

2.54 StatsBase.indicatormat
StatsBase.indicatormat — Method.

indicatormat(x, c=sort(unique(x)); sparse=false)

Construct a boolean matrix I of size (length(c), length(x)). Let ci be

the index of x[i] in c. Then I[ci, i] = true and all other elements are
false.
source

2.55 StatsBase.indicatormat
StatsBase.indicatormat — Method.

indicatormat(x, k::Integer; sparse=false)

Construct a boolean matrix I of size (k, length(x)) such that I[x[i], i]

= true and all other elements are set to false. If sparse is true, the output
will be a sparse matrix, otherwise it will be dense (default).
Examples

julia> using StatsBase

julia> indicatormat([1 2 2], 2)

23 Array{Bool,2}:
true false false
false true true

source

2.56 StatsBase.informationmatrix
StatsBase.informationmatrix — Method.

informationmatrix(model::StatisticalModel; expected::Bool = true)

Return the information matrix. By default the Fisher information ma-

trix is returned, while the observed information matrix can be requested with
expected = false.
source
2.57. STATSBASE.INVERSE RLE 63

2.57 StatsBase.inverse rle

StatsBase.inverse rle — Method.

inverse_rle(vals, lens)

Reconstruct a vector from its run-length encoding (see rle). vals is a vector
of the values and lens is a vector of the corresponding run lengths.
source

2.58 StatsBase.iqr
StatsBase.iqr — Method.

iqr(v)

Compute the interquartile range (IQR) of an array, i.e. the 75th percentile
minus the 25th percentile.
source

2.59 StatsBase.isfitted
StatsBase.isfitted — Method.

isfitted(obj::StatisticalModel)

Indicate whether the model has been fitted.

source

2.60 StatsBase.islinear
StatsBase.islinear — Method.

islinear(obj::StatisticalModel)

Indicate whether the model is linear.

source

2.61 StatsBase.kldivergence
StatsBase.kldivergence — Method.

kldivergence(p, q, [b])

Compute the Kullback-Leibler divergence of q from p, optionally specifying

a real number b such that the divergence is scaled by 1/log(b).
source
64 CHAPTER 2. STATSBASE

2.62 StatsBase.kurtosis
StatsBase.kurtosis — Method.

kurtosis(v, [wv::AbstractWeights], m=mean(v))

Compute the excess kurtosis of a real-valued array v, optionally specifying

a weighting vector wv and a center m.
source

2.63 StatsBase.levelsmap
StatsBase.levelsmap — Method.

levelsmap(a)

Construct a dictionary that maps each of the n unique values in a to a

number between 1 and n.
source

2.64 StatsBase.leverage
StatsBase.leverage — Method.

leverage(obj::RegressionModel)

Return the diagonal of the projection matrix.

source

2.65 StatsBase.loglikelihood
StatsBase.loglikelihood — Method.

loglikelihood(obj::StatisticalModel)

Return the log-likelihood of the model.

source

2.66 StatsBase.mad!
StatsBase.mad! — Method.

StatsBase.mad!(v; center=median!(v), normalize=true)

2.67. STATSBASE.MAD 65

Compute the median absolute deviation (MAD) of v around center (by

default, around the median), overwriting v in the process.
If normalize is set to true, the MAD is multiplied by 1 / quantile(Normal(),
3/4) 1.4826, in order to obtain a consistent estimator of the standard devia-
tion under the assumption that the data is normally distributed.
source

2.67 StatsBase.mad
StatsBase.mad — Method.

mad(v; center=median(v), normalize=true)

Compute the median absolute deviation (MAD) of v around center (by

default, around the median).
If normalize is set to true, the MAD is multiplied by 1 / quantile(Normal(),
3/4) 1.4826, in order to obtain a consistent estimator of the standard devia-
tion under the assumption that the data is normally distributed.
source

2.68 StatsBase.maxad
StatsBase.maxad — Method.

maxad(a, b)

Return the maximum absolute deviation between two arrays: maxabs(a -

b).
source

2.69 StatsBase.mean and cov

StatsBase.mean and cov — Function.

mean_and_cov(x, [wv::AbstractWeights]; vardim=1, corrected=false) -> (mean, cov)

Return the mean and covariance matrix as a tuple. A weighting vector wv

can be specified. vardim that designates whether the variables are columns in
the matrix (1) or rows (2). Finally, bias correction is applied to the covariance
calculation if corrected=true. See cov documentation for more details.
source
66 CHAPTER 2. STATSBASE

2.70 StatsBase.mean and std

StatsBase.mean and std — Method.

mean_and_std(x, [w::AbstractWeights], [dim]; corrected=false) -> (mean, std)

Return the mean and standard deviation of a real-valued array x, optionally

over a dimension dim, as a tuple. A weighting vector w can be specified to weight
the estimates. Finally, bias correction is applied to the standard deviation
calculation if corrected=true. See std documentation for more details.
source

2.71 StatsBase.mean and var

StatsBase.mean and var — Method.

mean_and_var(x, [w::AbstractWeights], [dim]; corrected=false) -> (mean, var)

Return the mean and variance of a real-valued array x, optionally over a

dimension dim, as a tuple. Observations in x can be weighted using weight
vector w. Finally, bias correction is be applied to the variance calculation if
corrected=true. See var documentation for more details.
source

2.72 StatsBase.meanad
StatsBase.meanad — Method.

meanad(a, b)

Return the mean absolute deviation between two arrays: mean(abs(a -

b)).
source

2.73 StatsBase.meanresponse
StatsBase.meanresponse — Method.

meanresponse(obj::RegressionModel)

Return the mean of the response.

source
2.74. STATSBASE.MODE 67

2.74 StatsBase.mode
StatsBase.mode — Method.
mode(a, [r])
Return the mode (most common number) of an array, optionally over a
specified range r. If several modes exist, the first one (in order of appearance)
is returned.
source

2.75 StatsBase.modelmatrix
StatsBase.modelmatrix — Method.
modelmatrix(obj::RegressionModel)
Return the model matrix (a.k.a. the design matrix).
source

2.76 StatsBase.modes
StatsBase.modes — Method.
modes(a, [r])::Vector
Return all modes (most common numbers) of an array, optionally over a
specified range r.
source

2.77 StatsBase.moment
StatsBase.moment — Method.
moment(v, k, [wv::AbstractWeights], m=mean(v))
Return the kth order central moment of a real-valued array v, optionally
specifying a weighting vector wv and a center m.
source

2.78 StatsBase.msd
StatsBase.msd — Method.
msd(a, b)
Return the mean squared deviation between two arrays: mean(abs2(a -
b)).
source
68 CHAPTER 2. STATSBASE

2.79 StatsBase.mss
StatsBase.mss — Method.

mss(obj::StatisticalModel)

Return the model sum of squares.

source

2.80 StatsBase.nobs
StatsBase.nobs — Method.

nobs(obj::StatisticalModel)

Return the number of independent observations on which the model was fit-
ted. Be careful when using this information, as the definition of an independent
observation may vary depending on the model, on the format used to pass the
data, on the sampling plan (if specified), etc.
source

2.81 StatsBase.nquantile
StatsBase.nquantile — Method.

nquantile(v, n)

Return the n-quantiles of a real-valued array, i.e. the values which partition
v into n subsets of nearly equal size.
Equivalent to quantile(v, [0:n]/n). For example, nquantiles(x, 5) re-
turns a vector of quantiles, respectively at [0.0, 0.2, 0.4, 0.6, 0.8, 1.0].
source

2.82 StatsBase.nulldeviance
StatsBase.nulldeviance — Method.

nulldeviance(obj::StatisticalModel)

Return the deviance of the null model, that is the one including only the
intercept.
source
2.83. STATSBASE.NULLLOGLIKELIHOOD 69

2.83 StatsBase.nullloglikelihood
StatsBase.nullloglikelihood — Method.

loglikelihood(obj::StatisticalModel)

Return the log-likelihood of the null model corresponding to model obj. This
is usually the model containing only the intercept.
source

2.84 StatsBase.ordinalrank
StatsBase.ordinalrank — Method.

ordinalrank(x; lt = isless, rev::Bool = false)

Return the ordinal ranking (“1234” ranking) of an array. The lt keyword

allows providing a custom “less than” function; use rev=true to reverse the
sorting order. All items in x are given distinct, successive ranks based on their
position in sort(x; lt = lt, rev = rev). Missing values are assigned rank
missing.
source

2.85 StatsBase.pacf !
StatsBase.pacf! — Method.

pacf!(r, X, lags; method=:regression)

Compute the partial autocorrelation function (PACF) of a matrix X at lags

and store the result in r. method designates the estimation method. Rec-
ognized values are :regression, which computes the partial autocorrelations
via successive regression models, and :yulewalker, which computes the partial
autocorrelations using the Yule-Walker equations.
r must be a matrix of size (length(lags), size(x, 2)).
source

2.86 StatsBase.pacf
StatsBase.pacf — Method.

pacf(X, lags; method=:regression)

70 CHAPTER 2. STATSBASE

Compute the partial autocorrelation function (PACF) of a real-valued vector

or matrix X at lags. method designates the estimation method. Recognized val-
ues are :regression, which computes the partial autocorrelations via successive
regression models, and :yulewalker, which computes the partial autocorrela-
tions using the Yule-Walker equations.
If x is a vector, return a vector of the same length as lags. If x is a matrix,
return a matrix of size (length(lags), size(x, 2)), where each column in
the result corresponds to a column in x.
source

2.87 StatsBase.percentile
StatsBase.percentile — Method.

percentile(v, p)

Return the pth percentile of a real-valued array v, i.e. quantile(x, p /

100).
source

2.88 StatsBase.predict
StatsBase.predict — Function.

predict(obj::RegressionModel, [newX])

Form the predicted response of model obj. An object with new covariate
values newX can be supplied, which should have the same type and structure as
that used to fit obj; e.g. for a GLM it would generally be a DataFrame with the
same variable names as the original predictors.
source

2.89 StatsBase.predict!
StatsBase.predict! — Function.

predict!

In-place version of predict.

source
2.90. STATSBASE.PROPORTIONMAP 71

2.90 StatsBase.proportionmap
StatsBase.proportionmap — Method.
proportionmap(x)
Return a dictionary mapping each unique value in x to its proportion in x.
source

2.91 StatsBase.proportions
StatsBase.proportions — Method.
proportions(x, k::Integer, [wv::AbstractWeights])
Return the proportion of integers in 1 to k that occur in x.
source

2.92 StatsBase.proportions
StatsBase.proportions — Method.
proportions(x, levels=span(x), [wv::AbstractWeights])
Return the proportion of values in the range levels that occur in x. Equiva-
lent to counts(x, levels) / length(x). If a weighting vector wv is specified,
the sum of the weights is used rather than the raw counts.
source

2.93 StatsBase.psnr
StatsBase.psnr — Method.
psnr(a, b, maxv)
Compute the peak signal-to-noise ratio between two arrays a and b. maxv is
the maximum possible value either array can take. The PSNR is computed as
10 * log10(maxv^2 / msd(a, b)).
source

2.94 StatsBase.pweights
StatsBase.pweights — Method.
pweights(vs)
Construct a ProbabilityWeights vector from a given array. See the docu-
mentation for ProbabilityWeights for more details.
source
72 CHAPTER 2. STATSBASE

2.95 StatsBase.r2
StatsBase.r2 — Method.

r2(obj::StatisticalModel, variant::Symbol)
r(obj::StatisticalModel, variant::Symbol)

Pseudo-coefficient of determination (pseudo R-squared).

For nonlinear models, one of several pseudo R definitions must be chosen
via variant. Supported variants are:

• :MacFadden (a.k.a. likelihood ratio index), defined as 1 − log L/ log L0.

• :CoxSnell, defined as 1 − (L0/L)2/n

• :Nagelkerke, defined as (1 − (L0/L)2/n )/(1 − L02/n ), with n the number

of observations (as returned by nobs).

In the above formulas, L is the likelihood of the model, L0 that of the null
model (the model including only the intercept). These two quantities are taken
to be minus half deviance of the corresponding models.
source

2.96 StatsBase.r2
StatsBase.r2 — Method.

r2(obj::StatisticalModel)
r(obj::StatisticalModel)

Coefficient of determination (R-squared).

For a linear model, the R is defined as ESS/T SS, with ESS the explained
sum of squares and T SS the total sum of squares.
source

2.97 StatsBase.renyientropy
StatsBase.renyientropy — Method.

renyientropy(p, )

Compute the Rnyi (generalized) entropy of order of an array p.

source
2.98. STATSBASE.RESIDUALS 73

2.98 StatsBase.residuals
StatsBase.residuals — Method.

residuals(obj::RegressionModel)

Return the residuals of the model.

source

2.99 StatsBase.response
StatsBase.response — Method.

response(obj::RegressionModel)

Return the model response (a.k.a. the dependent variable).

source

2.100 StatsBase.rle
StatsBase.rle — Method.

rle(v) -> (vals, lens)

Return the run-length encoding of a vector as a tuple. The first element

of the tuple is a vector of values of the input and the second is the number of
consecutive occurrences of each element.
Examples

julia> using StatsBase

julia> rle([1,1,1,2,2,3,3,3,3,2,2,2])
([1, 2, 3, 2], [3, 2, 4, 3])

source

2.101 StatsBase.rmsd
StatsBase.rmsd — Method.

rmsd(a, b; normalize=false)

Return the root mean squared deviation between two optionally normalized
arrays. The root mean squared deviation is computed as sqrt(msd(a, b)).
source
74 CHAPTER 2. STATSBASE

2.102 StatsBase.rss
StatsBase.rss — Method.

rss(obj::StatisticalModel)

Return the residual sum of squares.

source

2.103 StatsBase.sample!
StatsBase.sample! — Method.

sample!([rng], a, [wv::AbstractWeights], x; replace=true, ordered=false)

Draw a random sample of length(x) elements from an array a and store the
result in x. A polyalgorithm is used for sampling. Sampling probabilities are
proportional to the weights given in wv, if provided. replace dictates whether
sampling is performed with replacement and order dictates whether an ordered
sample, also called a sequential sample, should be taken.
Optionally specify a random number generator rng as the first argument
(defaults to Random.GLOBAL RNG).
source

2.104 StatsBase.sample
StatsBase.sample — Method.

sample([rng], a, [wv::AbstractWeights])

Select a single random element of a. Sampling probabilities are proportional

to the weights given in wv, if provided.
Optionally specify a random number generator rng as the first argument
(defaults to Random.GLOBAL RNG).
source

2.105 StatsBase.sample
StatsBase.sample — Method.

sample([rng], wv::AbstractWeights)

Select a single random integer in 1:length(wv) with probabilities propor-

tional to the weights given in wv.
Optionally specify a random number generator rng as the first argument
(defaults to Random.GLOBAL RNG).
source
2.106. STATSBASE.SAMPLE 75

2.106 StatsBase.sample
StatsBase.sample — Method.

sample([rng], a, [wv::AbstractWeights], n::Integer; replace=true, ordered=false)

Select a random, optionally weighted sample of size n from an array a using

a polyalgorithm. Sampling probabilities are proportional to the weights given in
wv, if provided. replace dictates whether sampling is performed with replace-
ment and order dictates whether an ordered sample, also called a sequential
sample, should be taken.
Optionally specify a random number generator rng as the first argument
(defaults to Random.GLOBAL RNG).
source

2.107 StatsBase.sample
StatsBase.sample — Method.

sample([rng], a, [wv::AbstractWeights], dims::Dims; replace=true, ordered=false)

Select a random, optionally weighted sample from an array a specifying the

dimensions dims of the output array. Sampling probabilities are proportional
to the weights given in wv, if provided. replace dictates whether sampling is
performed with replacement and order dictates whether an ordered sample,
also called a sequential sample, should be taken.
Optionally specify a random number generator rng as the first argument
(defaults to Random.GLOBAL RNG).
source

2.108 StatsBase.samplepair
StatsBase.samplepair — Method.

samplepair([rng], a)

Draw a pair of distinct elements from the array a without replacement.

Optionally specify a random number generator rng as the first argument
(defaults to Random.GLOBAL RNG).
source

2.109 StatsBase.samplepair
StatsBase.samplepair — Method.

samplepair([rng], n)
76 CHAPTER 2. STATSBASE

Draw a pair of distinct integers between 1 and n without replacement.

Optionally specify a random number generator rng as the first argument
(defaults to Random.GLOBAL RNG).
source

2.110 StatsBase.scattermat
StatsBase.scattermat — Function.

scattermat(X, [wv::AbstractWeights]; mean=nothing, vardim=1)

Compute the scatter matrix, which is an unnormalized covariance matrix.

A weighting vector wv can be specified to weight the estimate.
Arguments

• mean=nothing: a known mean value. nothing indicates that the mean

is unknown, and the function will compute the mean. Specifying mean=0
indicates that the data are centered and hence there’s no need to subtract
the mean.

• vardim=1: the dimension along which the variables are organized. When
vardim = 1, the variables are considered columns with observations in
rows; when vardim = 2, variables are in rows with observations in columns.

source

2.111 StatsBase.score
StatsBase.score — Method.

score(obj::StatisticalModel)

Return the score of the statistical model. The score is the gradient of the
log-likelihood with respect to the coefficients.
source

2.112 StatsBase.sem
StatsBase.sem — Method.

sem(a)

Return the standard error of the mean of a, i.e. sqrt(var(a) / length(a)).

source
2.113. STATSBASE.SKEWNESS 77

2.113 StatsBase.skewness
StatsBase.skewness — Method.
skewness(v, [wv::AbstractWeights], m=mean(v))
Compute the standardized skewness of a real-valued array v, optionally spec-
ifying a weighting vector wv and a center m.
source

2.114 StatsBase.span
StatsBase.span — Method.
span(x)
Return the span of an integer array, i.e. the range minimum(x):maximum(x).
The minimum and maximum of x are computed in one-pass using extrema.
source

2.115 StatsBase.sqL2dist
StatsBase.sqL2dist — Method.
sqL2dist(a, b)
Pn
Compute the squared L2 distance between two arrays: i=1 |ai − bi |2 . Ef-
ficient equivalent of sumabs2(a - b).
source

2.116 StatsBase.stderror
StatsBase.stderror — Method.
stderror(obj::StatisticalModel)
Return the standard errors for the coefficients of the model.
source

2.117 StatsBase.summarystats
StatsBase.summarystats — Method.
summarystats(a)
Compute summary statistics for a real-valued array a. Returns a SummaryStats
object containing the mean, minimum, 25th percentile, median, 75th percentile,
and maxmimum.
source
78 CHAPTER 2. STATSBASE

2.118 StatsBase.tiedrank
StatsBase.tiedrank — Method.

tiedrank(x)

Return the tied ranking, also called fractional or “1 2.5 2.5 4” ranking, of
an array. The lt keyword allows providing a custom “less than” function; use
rev=true to reverse the sorting order. Items that compare equal receive the
mean of the rankings they would have been assigned under ordinal ranking.
Missing values are assigned rank missing.
source

2.119 StatsBase.trim!
StatsBase.trim! — Method.

trim!(x; prop=0.0, count=0)

A variant of trim that modifies x in place.

source

2.120 StatsBase.trim
StatsBase.trim — Method.

trim(x; prop=0.0, count=0)

Return a copy of x with either count or proportion prop of the high-

est and lowest elements removed. To compute the trimmed mean of x use
mean(trim(x)); to compute the variance use trimvar(x) (see trimvar).
Example
[] julia¿ trim([1,2,3,4,5], prop=0.2) 3-element Array{Int64,1}: 2 3 4
source

2.121 StatsBase.trimvar
StatsBase.trimvar — Method.

trimvar(x; prop=0.0, count=0)

Compute the variance of the trimmed mean of x. This function uses the
Winsorized variance, as described in Wilcox (2010).
source
2.122. STATSBASE.VARIATION 79

2.122 StatsBase.variation
StatsBase.variation — Method.

variation(x, m=mean(x))

Return the coefficient of variation of an array x, optionally specifying a

precomputed mean m. The coefficient of variation is the ratio of the standard
deviation to the mean.
source

2.123 StatsBase.vcov
StatsBase.vcov — Method.

vcov(obj::StatisticalModel)

Return the variance-covariance matrix for the coefficients of the model.

source

2.124 StatsBase.weights
StatsBase.weights — Method.

weights(vs)

Construct a Weights vector from array vs. See the documentation for
Weights for more details.
source

2.125 StatsBase.weights
StatsBase.weights — Method.

weights(obj::StatisticalModel)

Return the weights used in the model.

source

2.126 StatsBase.winsor!
StatsBase.winsor! — Method.

winsor!(x; prop=0.0, count=0)

A variant of winsor that modifies vector x in place.

source
80 CHAPTER 2. STATSBASE

2.127 StatsBase.winsor
StatsBase.winsor — Method.

winsor(x; prop=0.0, count=0)

Return a copy of x with either count or proportion prop of the lowest

elements of x replaced with the next-lowest, and an equal number of the highest
elements replaced with the previous-highest. To compute the Winsorized mean
of x use mean(winsor(x)).
Example
[] julia¿ winsor([1,2,3,4,5], prop=0.2) 5-element Array{Int64,1}: 2 2 3 4 4
source

2.128 StatsBase.wmean
StatsBase.wmean — Method.

wmean(v, w::AbstractVector)

Compute the weighted mean of an array v with weights w.

source

2.129 StatsBase.wmedian
StatsBase.wmedian — Method.

wmedian(v, w)

Compute the weighted median of an array v with weights w, given as either

a vector or an AbstractWeights vector.
source

2.130 StatsBase.wquantile
StatsBase.wquantile — Method.

wquantile(v, w, p)

Compute the pth quantile(s) of v with weights w, given as either a vector or

an AbstractWeights vector.
source
2.131. STATSBASE.WSAMPLE! 81

2.131 StatsBase.wsample!
StatsBase.wsample! — Method.

wsample!([rng], a, w, x; replace=true, ordered=false)

Select a weighted sample from an array a and store the result in x. Sam-
pling probabilities are proportional to the weights given in w. replace dictates
whether sampling is performed with replacement and order dictates whether
an ordered sample, also called a sequential sample, should be taken.
Optionally specify a random number generator rng as the first argument
(defaults to Random.GLOBAL RNG).
source

2.132 StatsBase.wsample
StatsBase.wsample — Method.

wsample([rng], [a], w)

Select a weighted random sample of size 1 from a with probabilities propor-

tional to the weights given in w. If a is not present, select a random weight from
w.
Optionally specify a random number generator rng as the first argument
(defaults to Random.GLOBAL RNG).
source

2.133 StatsBase.wsample
StatsBase.wsample — Method.

wsample([rng], [a], w, n::Integer; replace=true, ordered=false)

Select a weighted random sample of size n from a with probabilities pro-

portional to the weights given in w if a is present, otherwise select a random
sample of size n of the weights given in w. replace dictates whether sampling
is performed with replacement and order dictates whether an ordered sample,
also called a sequential sample, should be taken.
Optionally specify a random number generator rng as the first argument
(defaults to Random.GLOBAL RNG).
source
82 CHAPTER 2. STATSBASE

2.134 StatsBase.wsample
StatsBase.wsample — Method.

wsample([rng], [a], w, dims::Dims; replace=true, ordered=false)

Select a weighted random sample from a with probabilities proportional to

the weights given in w if a is present, otherwise select a random sample of size
n of the weights given in w. The dimensions of the output are given by dims.
Optionally specify a random number generator rng as the first argument
(defaults to Random.GLOBAL RNG).
source

2.135 StatsBase.wsum!
StatsBase.wsum! — Method.

wsum!(R, A, w, dim; init=true)

Compute the weighted sum of A with weights w over the dimension dim and
store the result in R. If init=false, the sum is added to R rather than starting
from zero.
source

2.136 StatsBase.wsum
StatsBase.wsum — Method.

wsum(v, w::AbstractVector, [dim])

Compute the weighted sum of an array v with weights w, optionally over the
dimension dim.
source

2.137 StatsBase.zscore!
StatsBase.zscore! — Method.

zscore!([Z], X, , )

Compute the z-scores of an array X with mean and standard deviation .

z-scores are the signed number of standard deviations above the mean that an
observation lies, i.e. (x − )/.
If a destination array Z is provided, the scores are stored in Z and it must
have the same shape as X. Otherwise X is overwritten.
source
2.138. STATSBASE.ZSCORE 83

2.138 StatsBase.zscore
StatsBase.zscore — Method.

zscore(X, [, ])

Compute the z-scores of X, optionally specifying a precomputed mean and

standard deviation . z-scores are the signed number of standard deviations above
the mean that an observation lies, i.e. (x − )/.
and should be both scalars or both arrays. The computation is broadcast-
ing. In particular, when and are arrays, they should have the same size, and
size(, i) == 1 || size(, i) == size(X, i) for each dimension.
source
Chapter 3

PyPlot

3.1 Base.step
Base.step — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“step”,))
source

3.2 PyPlot.Axes3D
PyPlot.Axes3D — Method.
PyPlot.LazyHelp(PyPlot.LazyPyModule(“mpl toolkits.mplot3d.axes3d”, Py-
Call.PyObject(Ptr{PyCall.PyObject struct} @0x0000000000000000)), (“Axes3D”,))
source

3.3 PyPlot.acorr
PyPlot.acorr — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“acorr”,))
source

3.4 PyPlot.annotate
PyPlot.annotate — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“annotate”,))
source

84
3.5. PYPLOT.ARROW 85

3.5 PyPlot.arrow
PyPlot.arrow — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“arrow”,))
source

3.6 PyPlot.autoscale
PyPlot.autoscale — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“autoscale”,))
source

3.7 PyPlot.autumn
PyPlot.autumn — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“autumn”,))
source

3.8 PyPlot.axes
PyPlot.axes — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“axes”,))
source

3.9 PyPlot.axhline
PyPlot.axhline — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“axhline”,))
source

3.10 PyPlot.axhspan
PyPlot.axhspan — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“axhspan”,))
source
86 CHAPTER 3. PYPLOT

3.11 PyPlot.axis
PyPlot.axis — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“axis”,))
source

3.12 PyPlot.axvline
PyPlot.axvline — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“axvline”,))
source

3.13 PyPlot.axvspan
PyPlot.axvspan — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“axvspan”,))
source

3.14 PyPlot.bar
PyPlot.bar — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“bar”,))
source

3.15 PyPlot.bar3D
PyPlot.bar3D — Method.
PyPlot.LazyHelp(PyPlot.LazyPyModule(“mpl toolkits.mplot3d.axes3d”, Py-
Call.PyObject(Ptr{PyCall.PyObject struct} @0x0000000000000000)), (“Axes3D”,
“bar3d”))
source

3.16 PyPlot.barbs
PyPlot.barbs — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“barbs”,))
source
3.17. PYPLOT.BARH 87

3.17 PyPlot.barh
PyPlot.barh — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“barh”,))
source

3.18 PyPlot.bone
PyPlot.bone — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“bone”,))
source

3.19 PyPlot.box
PyPlot.box — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“box”,))
source

3.20 PyPlot.boxplot
PyPlot.boxplot — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“boxplot”,))
source

3.21 PyPlot.broken barh

PyPlot.broken barh — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“broken barh”,))
source

3.22 PyPlot.cla
PyPlot.cla — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“cla”,))
source
88 CHAPTER 3. PYPLOT

3.23 PyPlot.clabel
PyPlot.clabel — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“clabel”,))
source

3.24 PyPlot.clf
PyPlot.clf — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“clf”,))
source

3.25 PyPlot.clim
PyPlot.clim — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“clim”,))
source

3.26 PyPlot.cohere
PyPlot.cohere — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“cohere”,))
source

3.27 PyPlot.colorbar
PyPlot.colorbar — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“colorbar”,))
source

3.28 PyPlot.colors
PyPlot.colors — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“colors”,))
source
3.29. PYPLOT.CONTOUR 89

3.29 PyPlot.contour
PyPlot.contour — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“contour”,))
source

3.30 PyPlot.contour3D
PyPlot.contour3D — Method.
PyPlot.LazyHelp(PyPlot.LazyPyModule(“mpl toolkits.mplot3d.axes3d”, Py-
Call.PyObject(Ptr{PyCall.PyObject struct} @0x0000000000000000)), (“Axes3D”,
“contour3D”))
source

3.31 PyPlot.contourf
PyPlot.contourf — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“contourf”,))
source

3.32 PyPlot.contourf3D
PyPlot.contourf3D — Method.
PyPlot.LazyHelp(PyPlot.LazyPyModule(“mpl toolkits.mplot3d.axes3d”, Py-
Call.PyObject(Ptr{PyCall.PyObject struct} @0x0000000000000000)), (“Axes3D”,
“contourf3D”))
source

3.33 PyPlot.cool
PyPlot.cool — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“cool”,))
source

3.34 PyPlot.copper
PyPlot.copper — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“copper”,))
source
90 CHAPTER 3. PYPLOT

3.35 PyPlot.csd
PyPlot.csd — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“csd”,))
source

3.36 PyPlot.delaxes
PyPlot.delaxes — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“delaxes”,))
source

3.37 PyPlot.disconnect
PyPlot.disconnect — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“disconnect”,))
source

3.38 PyPlot.draw
PyPlot.draw — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“draw”,))
source

3.39 PyPlot.errorbar
PyPlot.errorbar — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“errorbar”,))
source

3.40 PyPlot.eventplot
PyPlot.eventplot — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“eventplot”,))
source
3.41. PYPLOT.FIGASPECT 91

3.41 PyPlot.figaspect
PyPlot.figaspect — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“figaspect”,))
source

3.42 PyPlot.figimage
PyPlot.figimage — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“figimage”,))
source

3.43 PyPlot.figlegend
PyPlot.figlegend — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“figlegend”,))
source

3.44 PyPlot.figtext
PyPlot.figtext — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“figtext”,))
source

3.45 PyPlot.figure
PyPlot.figure — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x00000000383be0d0),
())
source

3.46 PyPlot.fill between

PyPlot.fill between — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“fill between”,))
source
92 CHAPTER 3. PYPLOT

3.47 PyPlot.fill betweenx

PyPlot.fill betweenx — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“fill betweenx”,))
source

3.48 PyPlot.findobj
PyPlot.findobj — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“findobj”,))
source

3.49 PyPlot.flag
PyPlot.flag — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“flag”,))
source

3.50 PyPlot.gca
PyPlot.gca — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“gca”,))
source

3.51 PyPlot.gcf
PyPlot.gcf — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x00000000383be1e0),
())
source

3.52 PyPlot.gci
PyPlot.gci — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“gci”,))
source
3.53. PYPLOT.GET CMAP 93

3.53 PyPlot.get cmap

PyPlot.get cmap — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x0000000037bc50d0),
())
source

3.54 PyPlot.get current fig manager

PyPlot.get current fig manager — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“get current fig manager”,))
source

3.55 PyPlot.get figlabels

PyPlot.get figlabels — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“get figlabels”,))
source

3.56 PyPlot.get fignums

PyPlot.get fignums — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“get fignums”,))
source

3.57 PyPlot.get plot commands

PyPlot.get plot commands — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“get plot commands”,))
source

3.58 PyPlot.ginput
PyPlot.ginput — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“ginput”,))
source
94 CHAPTER 3. PYPLOT

3.59 PyPlot.gray
PyPlot.gray — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“gray”,))
source

3.60 PyPlot.grid
PyPlot.grid — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“grid”,))
source

3.61 PyPlot.hexbin
PyPlot.hexbin — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“hexbin”,))
source

3.62 PyPlot.hist2D
PyPlot.hist2D — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“hist2d”,))
source

3.63 PyPlot.hlines
PyPlot.hlines — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“hlines”,))
source

3.64 PyPlot.hold
PyPlot.hold — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“hold”,))
source
3.65. PYPLOT.HOT 95

3.65 PyPlot.hot
PyPlot.hot — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“hot”,))
source

3.66 PyPlot.hsv
PyPlot.hsv — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“hsv”,))
source

3.67 PyPlot.imread
PyPlot.imread — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“imread”,))
source

3.68 PyPlot.imsave
PyPlot.imsave — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“imsave”,))
source

3.69 PyPlot.imshow
PyPlot.imshow — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“imshow”,))
source

3.70 PyPlot.ioff
PyPlot.ioff — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“ioff”,))
source
96 CHAPTER 3. PYPLOT

3.71 PyPlot.ion
PyPlot.ion — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“ion”,))
source

3.72 PyPlot.ishold
PyPlot.ishold — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“ishold”,))
source

3.73 PyPlot.jet
PyPlot.jet — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“jet”,))
source

3.74 PyPlot.legend
PyPlot.legend — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“legend”,))
source

3.75 PyPlot.locator params

PyPlot.locator params — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“locator params”,))
source

3.76 PyPlot.loglog
PyPlot.loglog — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“loglog”,))
source
3.77. PYPLOT.MARGINS 97

3.77 PyPlot.margins
PyPlot.margins — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“margins”,))
source

3.78 PyPlot.matshow
PyPlot.matshow — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“matshow”,))
source

3.79 PyPlot.mesh
PyPlot.mesh — Method.
PyPlot.LazyHelp(PyPlot.LazyPyModule(“mpl toolkits.mplot3d.axes3d”, Py-
Call.PyObject(Ptr{PyCall.PyObject struct} @0x0000000000000000)), (“Axes3D”,
“plot wireframe”))
source

3.80 PyPlot.minorticks off

PyPlot.minorticks off — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“minorticks off”,))
source

3.81 PyPlot.minorticks on
PyPlot.minorticks on — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“minorticks on”,))
source

3.82 PyPlot.over
PyPlot.over — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“over”,))
source
98 CHAPTER 3. PYPLOT

3.83 PyPlot.pause
PyPlot.pause — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“pause”,))
source

3.84 PyPlot.pcolor
PyPlot.pcolor — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“pcolor”,))
source

3.85 PyPlot.pcolormesh
PyPlot.pcolormesh — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“pcolormesh”,))
source

3.86 PyPlot.pie
PyPlot.pie — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“pie”,))
source

3.87 PyPlot.pink
PyPlot.pink — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“pink”,))
source

3.88 PyPlot.plot
PyPlot.plot — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“plot”,))
source
3.89. PYPLOT.PLOT3D 99

3.89 PyPlot.plot3D
PyPlot.plot3D — Method.
PyPlot.LazyHelp(PyPlot.LazyPyModule(“mpl toolkits.mplot3d.axes3d”, Py-
Call.PyObject(Ptr{PyCall.PyObject struct} @0x0000000000000000)), (“Axes3D”,
“plot3D”))
source

3.90 PyPlot.plot date

PyPlot.plot date — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“plot date”,))
source

3.91 PyPlot.plot surface

PyPlot.plot surface — Method.
PyPlot.LazyHelp(PyPlot.LazyPyModule(“mpl toolkits.mplot3d.axes3d”, Py-
Call.PyObject(Ptr{PyCall.PyObject struct} @0x0000000000000000)), (“Axes3D”,
“plot surface”))
source

3.92 PyPlot.plot trisurf

PyPlot.plot trisurf — Method.
PyPlot.LazyHelp(PyPlot.LazyPyModule(“mpl toolkits.mplot3d.axes3d”, Py-
Call.PyObject(Ptr{PyCall.PyObject struct} @0x0000000000000000)), (“Axes3D”,
“plot trisurf”))
source

3.93 PyPlot.plot wireframe

PyPlot.plot wireframe — Method.
PyPlot.LazyHelp(PyPlot.LazyPyModule(“mpl toolkits.mplot3d.axes3d”, Py-
Call.PyObject(Ptr{PyCall.PyObject struct} @0x0000000000000000)), (“Axes3D”,
“plot wireframe”))
source

3.94 PyPlot.plotfile
PyPlot.plotfile — Method.
100 CHAPTER 3. PYPLOT

PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),

(“plotfile”,))
source

3.95 PyPlot.polar
PyPlot.polar — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“polar”,))
source

3.96 PyPlot.prism
PyPlot.prism — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“prism”,))
source

3.97 PyPlot.psd
PyPlot.psd — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“psd”,))
source

3.98 PyPlot.quiver
PyPlot.quiver — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“quiver”,))
source

3.99 PyPlot.quiverkey
PyPlot.quiverkey — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“quiverkey”,))
source
3.100. PYPLOT.RC 101

3.100 PyPlot.rc
PyPlot.rc — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“rc”,))
source

3.101 PyPlot.rc context

PyPlot.rc context — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“rc context”,))
source

3.102 PyPlot.rcdefaults
PyPlot.rcdefaults — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“rcdefaults”,))
source

3.103 PyPlot.register cmap

PyPlot.register cmap — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000001305aa60),
())
source

3.104 PyPlot.rgrids
PyPlot.rgrids — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“rgrids”,))
source

3.105 PyPlot.savefig
PyPlot.savefig — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“savefig”,))
source
102 CHAPTER 3. PYPLOT

3.106 PyPlot.sca
PyPlot.sca — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“sca”,))
source

3.107 PyPlot.scatter
PyPlot.scatter — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“scatter”,))
source

3.108 PyPlot.scatter3D
PyPlot.scatter3D — Method.
PyPlot.LazyHelp(PyPlot.LazyPyModule(“mpl toolkits.mplot3d.axes3d”, Py-
Call.PyObject(Ptr{PyCall.PyObject struct} @0x0000000000000000)), (“Axes3D”,
“scatter3D”))
source

3.109 PyPlot.sci
PyPlot.sci — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“sci”,))
source

3.110 PyPlot.semilogx
PyPlot.semilogx — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“semilogx”,))
source

3.111 PyPlot.semilogy
PyPlot.semilogy — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“semilogy”,))
source
3.112. PYPLOT.SET CMAP 103

3.112 PyPlot.set cmap

PyPlot.set cmap — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“set cmap”,))
source

3.113 PyPlot.setp
PyPlot.setp — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“setp”,))
source

3.114 PyPlot.specgram
PyPlot.specgram — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“specgram”,))
source

3.115 PyPlot.spectral
PyPlot.spectral — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“spectral”,))
source

3.116 PyPlot.spring
PyPlot.spring — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“spring”,))
source

3.117 PyPlot.spy
PyPlot.spy — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“spy”,))
source
104 CHAPTER 3. PYPLOT

3.118 PyPlot.stackplot
PyPlot.stackplot — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“stackplot”,))
source

3.119 PyPlot.stem
PyPlot.stem — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“stem”,))
source

3.120 PyPlot.streamplot
PyPlot.streamplot — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“streamplot”,))
source

3.121 PyPlot.subplot
PyPlot.subplot — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“subplot”,))
source

3.122 PyPlot.subplot2grid
PyPlot.subplot2grid — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“subplot2grid”,))
source

3.123 PyPlot.subplot tool

PyPlot.subplot tool — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“subplot tool”,))
source
3.124. PYPLOT.SUBPLOTS 105

3.124 PyPlot.subplots
PyPlot.subplots — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“subplots”,))
source

3.125 PyPlot.subplots adjust

PyPlot.subplots adjust — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“subplots adjust”,))
source

3.126 PyPlot.summer
PyPlot.summer — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“summer”,))
source

3.127 PyPlot.suptitle
PyPlot.suptitle — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“suptitle”,))
source

3.128 PyPlot.surf
PyPlot.surf — Method.
PyPlot.LazyHelp(PyPlot.LazyPyModule(“mpl toolkits.mplot3d.axes3d”, Py-
Call.PyObject(Ptr{PyCall.PyObject struct} @0x0000000000000000)), (“Axes3D”,
“plot surface”))
source

3.129 PyPlot.table
PyPlot.table — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“table”,))
source
106 CHAPTER 3. PYPLOT

3.130 PyPlot.text
PyPlot.text — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“text”,))
source

3.131 PyPlot.text2D
PyPlot.text2D — Method.
PyPlot.LazyHelp(PyPlot.LazyPyModule(“mpl toolkits.mplot3d.axes3d”, Py-
Call.PyObject(Ptr{PyCall.PyObject struct} @0x0000000000000000)), (“Axes3D”,
“text2D”))
source

3.132 PyPlot.text3D
PyPlot.text3D — Method.
PyPlot.LazyHelp(PyPlot.LazyPyModule(“mpl toolkits.mplot3d.axes3d”, Py-
Call.PyObject(Ptr{PyCall.PyObject struct} @0x0000000000000000)), (“Axes3D”,
“text3D”))
source

3.133 PyPlot.thetagrids
PyPlot.thetagrids — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“thetagrids”,))
source

3.134 PyPlot.tick params

PyPlot.tick params — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“tick params”,))
source

3.135 PyPlot.ticklabel format

PyPlot.ticklabel format — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“ticklabel format”,))
source
3.136. PYPLOT.TIGHT LAYOUT 107

3.136 PyPlot.tight layout

PyPlot.tight layout — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“tight layout”,))
source

3.137 PyPlot.title
PyPlot.title — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“title”,))
source

3.138 PyPlot.tricontour
PyPlot.tricontour — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“tricontour”,))
source

3.139 PyPlot.tricontourf
PyPlot.tricontourf — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“tricontourf”,))
source

3.140 PyPlot.tripcolor
PyPlot.tripcolor — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“tripcolor”,))
source

3.141 PyPlot.triplot
PyPlot.triplot — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“triplot”,))
source
108 CHAPTER 3. PYPLOT

3.142 PyPlot.twinx
PyPlot.twinx — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“twinx”,))
source

3.143 PyPlot.twiny
PyPlot.twiny — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“twiny”,))
source

3.144 PyPlot.using3D
PyPlot.using3D — Method.

using3D()

This function ensures that the mplot3d module is loaded for 3d plotting.
This occurs automatically if you call any of the 3d plotting functions like plot3D
or surf, but it may be necessary to call this function manually if you are passing
projection="3d" explicitly to axes or subplot objects.
source

3.145 PyPlot.vlines
PyPlot.vlines — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“vlines”,))
source

3.146 PyPlot.waitforbuttonpress
PyPlot.waitforbuttonpress — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“waitforbuttonpress”,))
source
3.147. PYPLOT.WINTER 109

3.147 PyPlot.winter
PyPlot.winter — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“winter”,))
source

3.148 PyPlot.xkcd
PyPlot.xkcd — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“xkcd”,))
source

3.149 PyPlot.xlabel
PyPlot.xlabel — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“xlabel”,))
source

3.150 PyPlot.xlim
PyPlot.xlim — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“xlim”,))
source

3.151 PyPlot.xscale
PyPlot.xscale — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“xscale”,))
source

3.152 PyPlot.xticks
PyPlot.xticks — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“xticks”,))
source
110 CHAPTER 3. PYPLOT

3.153 PyPlot.ylabel
PyPlot.ylabel — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“ylabel”,))
source

3.154 PyPlot.ylim
PyPlot.ylim — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“ylim”,))
source

3.155 PyPlot.yscale
PyPlot.yscale — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“yscale”,))
source

3.156 PyPlot.yticks
PyPlot.yticks — Method.
PyPlot.LazyHelp(PyCall.PyObject(Ptr{PyCall.PyObject struct} @0x000000002ba7ed68),
(“yticks”,))
source

3.157 PyPlot.zlabel
PyPlot.zlabel — Method.
PyPlot.LazyHelp(PyPlot.LazyPyModule(“mpl toolkits.mplot3d.axes3d”, Py-
Call.PyObject(Ptr{PyCall.PyObject struct} @0x0000000000000000)), (“Axes3D”,
“set zlabel”))
source

3.158 PyPlot.zlim
PyPlot.zlim — Method.
PyPlot.LazyHelp(PyPlot.LazyPyModule(“mpl toolkits.mplot3d.axes3d”, Py-
Call.PyObject(Ptr{PyCall.PyObject struct} @0x0000000000000000)), (“Axes3D”,
“set zlim”))
source
3.159. PYPLOT.ZSCALE 111

3.159 PyPlot.zscale
PyPlot.zscale — Method.
PyPlot.LazyHelp(PyPlot.LazyPyModule(“mpl toolkits.mplot3d.axes3d”, Py-
Call.PyObject(Ptr{PyCall.PyObject struct} @0x0000000000000000)), (“Axes3D”,
“set zscale”))
source

3.160 PyPlot.zticks
PyPlot.zticks — Method.
PyPlot.LazyHelp(PyPlot.LazyPyModule(“mpl toolkits.mplot3d.axes3d”, Py-
Call.PyObject(Ptr{PyCall.PyObject struct} @0x0000000000000000)), (“Axes3D”,
“set zticks”))
source
Chapter 4

IndexedTables

4.1 Base.Sort.select
Base.Sort.select — Method.
select(t::Table, which::Selection)
Select all or a subset of columns, or a single column from the table.
Selection is a type union of many types that can select from a table. It
can be:

1. Integer – returns the column at this position.

2. Symbol – returns the column with this name.
3. Pair{Selection => Function} – selects and maps a function over the
selection, returns the result.
4. AbstractArray – returns the array itself. This must be the same length
as the table.
5. Tuple of Selection – returns a table containing a column for every se-
lector in the tuple. The tuple may also contain the type Pair{Symbol,
Selection}, which the selection a name. The most useful form of this
when introducing a new column.

Examples:
Selection with Integer – returns the column at this position.

julia> tbl = table([0.01, 0.05], [2,1], [3,4], names=[:t, :x, :y], pkey=:t)
Table with 2 rows, 3 columns:
t x y

0.01 2 3
0.05 1 4

112
4.1. BASE.SORT.SELECT 113

julia> select(tbl, 2)
2-element Array{Int64,1}:
2
1

Selection with Symbol – returns the column with this name.

julia> select(tbl, :t)

2-element Array{Float64,1}:
0.01
0.05

Selection with Pair{Selection => Function} – selects some columns and

maps a function over it, then returns the mapped column.

julia> select(tbl, :t=>t->1/t)

2-element Array{Float64,1}:
100.0
20.0

Selection with AbstractArray – returns the array itself.

julia> select(tbl, [3,4])

2-element Array{Int64,1}:
3
4

Selection with Tuple– returns a table containing a column for every selector
in the tuple.

julia> select(tbl, (2,1))

Table with 2 rows, 2 columns:
x t

2 0.01
1 0.05

julia> vx = select(tbl, (:x, :t)=>p->p.x/p.t)

2-element Array{Float64,1}:
200.0
20.0

julia> select(tbl, (:x,:t=>-))

Table with 2 rows, 2 columns:
x t

1 -0.05
2 -0.01
114 CHAPTER 4. INDEXEDTABLES

Note that since tbl was initialized with t as the primary key column, se-
lections that retain the key column will retain its status as a key. The same
applies when multiple key columns are selected.
Selection with a custom array in the tuple will cause the name of the columns
to be removed and replaced with integers.

julia> select(tbl, (:x, :t, [3,4]))

Table with 2 rows, 3 columns:
1 2 3

2 0.01 3
1 0.05 4

This is because the third column’s name is unknown. In general if a column’s

name cannot be determined, then selection returns an iterable of tuples rather
than named tuples. In other words, it strips column names.
To specify a new name to a custom column, you can use Symbol => Selection
selector.

julia> select(tbl, (:x,:t,:z=>[3,4]))

Table with 2 rows, 3 columns:
x t z

2 0.01 3
1 0.05 4

julia> select(tbl, (:x, :t, :minust=>:t=>-))

Table with 2 rows, 3 columns:
x t minust

2 0.01 -0.01
1 0.05 -0.05

julia> select(tbl, (:x, :t, :vx=>(:x,:t)=>p->p.x/p.t))

Table with 2 rows, 3 columns:
x t vx

2 0.01 200.0
1 0.05 20.0

source

4.2 Base.reduce
Base.reduce — Method.
reduce(f, t::Table; select::Selection)
4.2. BASE.REDUCE 115

Reduce t by applying f pair-wise on values or structs selected by select.

f can be:
1. A function
2. An OnlineStat
3. A tuple of functions and/or OnlineStats
4. A named tuple of functions and/or OnlineStats
5. A named tuple of (selector => function or OnlineStat) pairs
julia> t = table([0.1, 0.5, 0.75], [0,1,2], names=[:t, :x])
Table with 3 rows, 2 columns:
t x

0.1 0
0.5 1
0.75 2
When f is a function, it reduces the selection as usual:
julia> reduce(+, t, select=:t)
1.35
If select is omitted, the rows themselves are passed to reduce as tuples.
julia> reduce((a, b) -> @NT(t=a.t+b.t, x=a.x+b.x), t)
(t = 1.35, x = 3)
If f is an OnlineStat object from the OnlineStats package, the statistic is
computed on the selection.
julia> using OnlineStats

julia> reduce(Mean(), t, select=:t)

Mean: n=3 | value=0.45
Reducing with multiple functions
Often one needs many aggregate values from a table. This is when f can be
passed as a tuple of functions:
julia> y = reduce((min, max), t, select=:x)
(min = 0, max = 2)

julia> y.max
2

julia> y.min
0
116 CHAPTER 4. INDEXEDTABLES

Note that the return value of invoking reduce with a tuple of functions will
be a named tuple which has the function names as the keys. In the example,
we reduced using min and max functions to obtain the minimum and maximum
values in column x.
If you want to give a different name to the fields in the output, use a named
tuple as f instead:

julia> y = reduce(@NT(sum=+, prod=*), t, select=:x)

(sum = 3, prod = 0)

You can also compute many OnlineStats by passing tuple or named tuple of
OnlineStat objects as the reducer.

julia> y = reduce((Mean(), Variance()), t, select=:t)

(Mean = Mean: n=3 | value=0.45, Variance = Variance: n=3 | value=0.1075)

julia> y.Mean
Mean: n=3 | value=0.45

julia> y.Variance
Variance: n=3 | value=0.1075

Combining reduction and selection

In the above section where we computed many reduced values at once, we
have been using the same selection for all reducers, that specified by select.
It’s possible to select different inputs for different reducers by using a named
tuple of slector => function pairs:

julia> reduce(@NT(xsum=:x=>+, negtsum=(:t=>-)=>+), t)

(xsum = 3, negtsum = -1.35)

See Selection for more on what selectors can be specified. Here since
each output can select its own input, select keyword is unsually unnecessary.
If specified, the slections in the reducer tuple will be done over the result of
selecting with the select argument.
source

4.3 DataValues.dropna
DataValues.dropna — Function.
dropna(t[, select])
Drop rows which contain NA values.

julia> t = table([0.1, 0.5, NA,0.7], [2,NA,4,5], [NA,6,NA,7],

names=[:t,:x,:y])
Table with 4 rows, 3 columns:
t x y
4.4. INDEXEDTABLES.AGGREGATE! 117

0.1 2 #NA
0.5 #NA 6
#NA 4 #NA
0.7 5 7

julia> dropna(t)
Table with 1 rows, 3 columns:
t x y

0.7 5 7

Optionally select can be speicified to limit columns to look for NAs in.

julia> dropna(t, :y)

Table with 2 rows, 3 columns:
t x y

0.5 #NA 6
0.7 5 7

julia> t1 = dropna(t, (:t, :x))

Table with 2 rows, 3 columns:
t x y

0.1 2 #NA
0.7 5 7

Any columns whose NA rows have been dropped will be converted to non-
na array type. In our last example, columns t and x got converted from
Array{DataValue{Int}} to Array{Int}. Similarly if the vectors are of type
DataValueArray{T} (default for loadtable) they will be converted to Array{T}.
[] julia¿ typeof(column(dropna(t,:x), :x)) Array{Int64,1}
source

4.4 IndexedTables.aggregate!
IndexedTables.aggregate! — Method.
aggregate!(f::Function, arr::NDSparse)
Combine adjacent rows with equal indices using the given 2-argument re-
duction function, in place.
source
118 CHAPTER 4. INDEXEDTABLES

4.5 IndexedTables.asofjoin
IndexedTables.asofjoin — Method.
asofjoin(left::NDSparse, right::NDSparse)
asofjoin is most useful on two time-series. It joins rows from left with the
“most recent” value from right.

julia> x = ndsparse((["ko","ko", "xrx","xrx"],

Date.(["2017-11-11", "2017-11-12",
"2017-11-11", "2017-11-12"])), [1,2,3,4]);

julia> y = ndsparse((["ko","ko", "xrx","xrx"],

Date.(["2017-11-12", "2017-11-13",
"2017-11-10", "2017-11-13"])), [5,6,7,8])

julia> asofjoin(x,y)
2-d NDSparse with 4 values (Int64):
1 2

"ko" 2017-11-11 1
"ko" 2017-11-12 5
"xrx" 2017-11-11 7
"xrx" 2017-11-12 7

source

4.6 IndexedTables.collect columns

IndexedTables.collect columns — Method.
collect columns(itr)
Collect an iterable as a Columns object if it iterates Tuples or NamedTuples,
as a normal Array otherwise.
Examples

julia> s = [(1,2), (3,4)];

julia> collect_columns(s)
2-element Columns{Tuple{Int64,Int64}}:
(1, 2)
(3, 4)

julia> s = Iterators.filter(isodd, 1:8);

julia> collect_columns(s)
4-element Array{Int64,1}:
1
4.7. INDEXEDTABLES.COLNAMES 119

3
5
7

source

4.7 IndexedTables.colnames
IndexedTables.colnames — Function.

colnames(itr)

Returns the names of the “columns” in itr.

Examples:

julia> colnames([1,2,3])
1-element Array{Int64,1}:
1

julia> colnames(Columns([1,2,3], [3,4,5]))

2-element Array{Int64,1}:
1
2

julia> colnames(table([1,2,3], [3,4,5]))

2-element Array{Int64,1}:
1
2

julia> colnames(Columns(x=[1,2,3], y=[3,4,5]))

2-element Array{Symbol,1}:
:x
:y

julia> colnames(table([1,2,3], [3,4,5], names=[:x,:y]))

2-element Array{Symbol,1}:
:x
:y

julia> colnames(ndsparse(Columns(x=[1,2,3]), Columns(y=[3,4,5])))

2-element Array{Symbol,1}:
:x
:y

julia> colnames(ndsparse(Columns(x=[1,2,3]), [3,4,5]))

2-element Array{Any,1}:
120 CHAPTER 4. INDEXEDTABLES

:x
1
2

julia> colnames(ndsparse(Columns(x=[1,2,3]), [3,4,5]))

2-element Array{Any,1}:
:x
2

julia> colnames(ndsparse(Columns([1,2,3], [4,5,6]), Columns(x=[6,7,8])))

3-element Array{Any,1}:
1
2
:x

julia> colnames(ndsparse(Columns(x=[1,2,3]), Columns([3,4,5],[6,7,8])))

3-element Array{Any,1}:
:x
2
3

source

4.8 IndexedTables.columns
IndexedTables.columns — Function.
columns(itr[, select::Selection])
Select one or more columns from an iterable of rows as a tuple of vectors.
select specifies which columns to select. See Selection convention for
possible values. If unspecified, returns all columns.
itr can be NDSparse, Columns and AbstractVector, and their distributed
counterparts.
Examples

julia> t = table([1,2],[3,4], names=[:x,:y])

Table with 2 rows, 2 columns:
x y

1 3
2 4

julia> columns(t)
(x = [1, 2], y = [3, 4])

julia> columns(t, :x)

2-element Array{Int64,1}:
4.9. INDEXEDTABLES.COLUMNS 121

1
2

julia> columns(t, (:x,))

(x = [1, 2])

julia> columns(t, (:y,:x=>-))

(y = [3, 4], x = [-1, -2])

source

4.9 IndexedTables.columns
IndexedTables.columns — Method.
columns(itr, which)
Returns a vector or a tuple of vectors from the iterator.
source

4.10 IndexedTables.convertdim
IndexedTables.convertdim — Method.
convertdim(x::NDSparse, d::DimName, xlate; agg::Function, vecagg::Function,
name)
Apply function or dictionary xlate to each index in the specified dimension.
If the mapping is many-to-one, agg or vecagg is used to aggregate the results.
If agg is passed, it is used as a 2-argument reduction function over the data. If
vecagg is passed, it is used as a vector-to-scalar function to aggregate the data.
name optionally specifies a new name for the translated dimension.
source

4.11 IndexedTables.dimlabels
IndexedTables.dimlabels — Method.
dimlabels(t::NDSparse)
Returns an array of integers or symbols giving the labels for the dimensions
of t. ndims(t) == length(dimlabels(t)).
source

4.12 IndexedTables.flatten
IndexedTables.flatten — Function.
flatten(t::Table, col=length(columns(t)))
Flatten col column which may contain a vector of vectors while repeating
the other fields. If column argument is not provided, default to last column.
122 CHAPTER 4. INDEXEDTABLES

Examples:

julia> x = table([1,2], [[3,4], [5,6]], names=[:x, :y])

Table with 2 rows, 2 columns:
x y

1 [3, 4]
2 [5, 6]

julia> flatten(x, 2)
Table with 4 rows, 2 columns:
x y

1 3
1 4
2 5
2 6

julia> x = table([1,2], [table([3,4],[5,6], names=[:a,:b]),

table([7,8], [9,10], names=[:a,:b])], names=[:x, :y]);

julia> flatten(x, :y)

Table with 4 rows, 3 columns:
x a b

1 3 5
1 4 6
2 7 9
2 8 10

source

4.13 IndexedTables.flush!
IndexedTables.flush! — Method.
flush!(arr::NDSparse)
Commit queued assignment operations, by sorting and merging the internal
temporary buffer.
source

4.14 IndexedTables.groupby
IndexedTables.groupby — Function.
groupby(f, t[, by::Selection]; select::Selection, flatten)
4.14. INDEXEDTABLES.GROUPBY 123

Group rows by by, and apply f to each group. f can be a function or a tuple
of functions. The result of f on each group is put in a table keyed by unique by
values. flatten will flatten the result and can be used when f returns a vector
instead of a single scalar value.
Examples

julia> t=table([1,1,1,2,2,2], [1,1,2,2,1,1], [1,2,3,4,5,6],

names=[:x,:y,:z]);

julia> groupby(mean, t, :x, select=:z)

Table with 2 rows, 2 columns:
x mean

1 2.0
2 5.0

julia> groupby(identity, t, (:x, :y), select=:z)

Table with 4 rows, 3 columns:
x y identity

1 1 [1, 2]
1 2 [3]
2 1 [5, 6]
2 2 [4]

julia> groupby(mean, t, (:x, :y), select=:z)

Table with 4 rows, 3 columns:
x y mean

1 1 1.5
1 2 3.0
2 1 5.5
2 2 4.0
multiple aggregates can be computed by passing a tuple of functions:

julia> groupby((mean, std, var), t, :y, select=:z)

Table with 2 rows, 4 columns:
y mean std var

1 3.5 2.38048 5.66667

2 3.5 0.707107 0.5

julia> groupby(@NT(q25=z->quantile(z, 0.25), q50=median,

q75=z->quantile(z, 0.75)), t, :y, select=:z)
Table with 2 rows, 4 columns:
y q25 q50 q75
124 CHAPTER 4. INDEXEDTABLES

1 1.75 3.5 5.25

2 3.25 3.5 3.75

Finally, it’s possible to select different inputs for different functions by using
a named tuple of slector => function pairs:

julia> groupby(@NT(xmean=:z=>mean, ystd=(:y=>-)=>std), t, :x)

Table with 2 rows, 3 columns:
x xmean ystd

1 2.0 0.57735
2 5.0 0.57735

By default, the result of groupby when f returns a vector or iterator of values

will not be expanded. Pass the flatten option as true to flatten the grouped
column:

julia> t = table([1,1,2,2], [3,4,5,6], names=[:x,:y])

julia> groupby((:normy => x->Iterators.repeated(mean(x), length(x)),),

t, :x, select=:y, flatten=true)
Table with 4 rows, 2 columns:
x normy

1 3.5
1 3.5
2 5.5
2 5.5

The keyword option usekey = true allows to use information from the in-
dexing column. f will need to accept two arguments, the first being the key (as
a Tuple or NamedTuple) the second the data (as Columns).

julia> t = table([1,1,2,2], [3,4,5,6], names=[:x,:y])

julia> groupby((:x_plus_mean_y => (key, d) -> key.x + mean(d),),

t, :x, select=:y, usekey = true)
Table with 2 rows, 2 columns:
x x_plus_mean_y

1 4.5
2 7.5

source
4.15. INDEXEDTABLES.GROUPJOIN 125

4.15 IndexedTables.groupjoin
IndexedTables.groupjoin — Method.
groupjoin([f, ] left, right; how, <options>)
Join left and right creating groups of values with matching keys.
Inner join
Inner join is the default join (when how is unspecified). It looks up keys from
left in right and only joins them when there is a match. This generates the
“intersection” of keys from left and right.
One-to-many and many-to-many matches
If the same key appears multiple times in either table (say, m and n times
respectively), each row with a key from left is matched with each row from
right with that key. The resulting group has mn output elements.

julia> l = table([1,1,1,2], [1,2,2,1], [1,2,3,4],

names=[:a,:b,:c], pkey=(:a, :b))
Table with 4 rows, 3 columns:
a b c

1 1 1
1 2 2
1 2 3
2 1 4

julia> r = table([0,1,1,2], [1,2,2,1], [1,2,3,4],

names=[:a,:b,:d], pkey=(:a, :b))
Table with 4 rows, 3 columns:
a b d

0 1 1
1 2 2
1 2 3
2 1 4

julia> groupjoin(l,r)
Table with 2 rows, 3 columns:
a b groups

1 2 NamedTuples._NT_c_d{Int64,Int64}[(c = 2, d = 2), (c = 2, d = 3), (c = 3, d = 2), (c = 3, d

2 1 NamedTuples._NT_c_d{Int64,Int64}[(c = 4, d = 4)]

Left join
Left join looks up rows from right where keys match that in left. If there
are no such rows in right, an NA value is used for every selected field from
right.

julia> groupjoin(l,r, how=:left)

126 CHAPTER 4. INDEXEDTABLES

Table with 3 rows, 3 columns:

a b groups

1 1 NamedTuples._NT_c_d{Int64,Int64}[]
1 2 NamedTuples._NT_c_d{Int64,Int64}[(c = 2, d = 2), (c = 2, d = 3), (c = 3, d = 2),
2 1 NamedTuples._NT_c_d{Int64,Int64}[(c = 4, d = 4)]

Outer join
Outer (aka Union) join looks up rows from right where keys match that in
left, and also rows from left where keys match those in left, if there are no
matches on either side, a tuple of NA values is used. The output is guarranteed
to contain

julia> groupjoin(l,r, how=:outer)

Table with 4 rows, 3 columns:
a b groups

0 1 NamedTuples._NT_c_d{Int64,Int64}[]
1 1 NamedTuples._NT_c_d{Int64,Int64}[]
1 2 NamedTuples._NT_c_d{Int64,Int64}[(c = 2, d = 2), (c = 2, d = 3), (c = 3, d = 2),
2 1 NamedTuples._NT_c_d{Int64,Int64}[(c = 4, d = 4)]

Options

• how::Symbol – join method to use. Described above.

• lkey::Selection – fields from left to match on

• rkey::Selection – fields from right to match on

• lselect::Selection – fields from left to use as input to use as output

columns, or input to f if it is specified. By default, this is all fields not
selected in lkey.

• rselect::Selection – fields from left to use as input to use as output

columns, or input to f if it is specified. By default, this is all fields not
selected in rkey.

julia> groupjoin(l,r, lkey=:a, rkey=:a, lselect=:c, rselect=:d, how=:outer)

Table with 3 rows, 2 columns:
a groups

0 NamedTuples._NT_c_d{Int64,Int64}[]
1 NamedTuples._NT_c_d{Int64,Int64}[(c = 1, d = 2), (c = 1, d = 3), (c = 2, d = 2), (c
2 NamedTuples._NT_c_d{Int64,Int64}[(c = 4, d = 4)]

source
4.16. INDEXEDTABLES.GROUPREDUCE 127

4.16 IndexedTables.groupreduce
IndexedTables.groupreduce — Function.
groupreduce(f, t[, by::Selection]; select::Selection)
Group rows by by, and apply f to reduce each group. f can be a function,
OnlineStat or a struct of these as described in reduce. Recommended: see
documentation for reduce first. The result of reducing each group is put in a
table keyed by unique by values, the names of the output columns are the same
as the names of the fields of the reduced tuples.
Examples

julia> t=table([1,1,1,2,2,2], [1,1,2,2,1,1], [1,2,3,4,5,6],

names=[:x,:y,:z]);

julia> groupreduce(+, t, :x, select=:z)

Table with 2 rows, 2 columns:
x +

1 6
2 15

julia> groupreduce(+, t, (:x, :y), select=:z)

Table with 4 rows, 3 columns:
x y +

1 1 3
1 2 3
2 1 11
2 2 4

julia> groupreduce((+, min, max), t, (:x, :y), select=:z)

Table with 4 rows, 5 columns:
x y + min max

1 1 3 1 2
1 2 3 3 3
2 1 11 5 6
2 2 4 4 4

If f is a single function or a tuple of functions, the output columns will be

named the same as the functions themselves. To change the name, pass a named
tuple:

julia> groupreduce(@NT(zsum=+, zmin=min, zmax=max), t, (:x, :y), select=:z)

Table with 4 rows, 5 columns:
x y zsum zmin zmax
128 CHAPTER 4. INDEXEDTABLES

1 1 3 1 2
1 2 3 3 3
2 1 11 5 6
2 2 4 4 4

Finally, it’s possible to select different inputs for different reducers by using
a named tuple of slector => function pairs:

julia> groupreduce(@NT(xsum=:x=>+, negysum=(:y=>-)=>+), t, :x)

Table with 2 rows, 3 columns:
x xsum negysum

1 3 -4
2 6 -4

source

4.17 IndexedTables.insertcol
IndexedTables.insertcol — Method.
insertcol(t, position::Integer, name, x)
Insert a column x named name at position. Returns a new table.

julia> t = table([0.01, 0.05], [2,1], [3,4], names=[:t, :x, :y], pkey=:t)

Table with 2 rows, 3 columns:
t x y

0.01 2 3
0.05 1 4

julia> insertcol(t, 2, :w, [0,1])

Table with 2 rows, 4 columns:
t w x y

0.01 0 2 3
0.05 1 1 4

source

4.18 IndexedTables.insertcolafter
IndexedTables.insertcolafter — Method.
insertcolafter(t, after, name, col)
Insert a column col named name after after. Returns a new table.
4.19. INDEXEDTABLES.INSERTCOLBEFORE 129

julia> t = table([0.01, 0.05], [2,1], [3,4], names=[:t, :x, :y], pkey=:t)

Table with 2 rows, 3 columns:
t x y

0.01 2 3
0.05 1 4

julia> insertcolafter(t, :t, :w, [0,1])

Table with 2 rows, 4 columns:
t w x y

0.01 0 2 3
0.05 1 1 4
source

4.19 IndexedTables.insertcolbefore
IndexedTables.insertcolbefore — Method.
insertcolbefore(t, before, name, col)
Insert a column col named name before before. Returns a new table.
julia> t = table([0.01, 0.05], [2,1], [3,4], names=[:t, :x, :y], pkey=:t)
Table with 2 rows, 3 columns:
t x y

0.01 2 3
0.05 1 4

julia> insertcolbefore(t, :x, :w, [0,1])

Table with 2 rows, 4 columns:
t w x y

0.01 0 2 3
0.05 1 1 4
source

4.20 IndexedTables.map rows

IndexedTables.map rows — Method.
map rows(f, c...)
Transform collection c by applying f to each element. For multiple collection
arguments, apply f elementwise. Collect output as Columns if f returns Tuples
or NamedTuples with constant fields, as Array otherwise.
Examples
130 CHAPTER 4. INDEXEDTABLES

julia> map_rows(i -> @NT(exp = exp(i), log = log(i)), 1:5)

5-element IndexedTables.Columns{NamedTuples._NT_exp_log{Float64,Float64},NamedTuples._N
(exp = 2.71828, log = 0.0)
(exp = 7.38906, log = 0.693147)
(exp = 20.0855, log = 1.09861)
(exp = 54.5982, log = 1.38629)
(exp = 148.413, log = 1.60944)

source

4.21 IndexedTables.ncols
IndexedTables.ncols — Function.
ncols(itr)
Returns the number of columns in itr.

julia> ncols([1,2,3])
1

julia> d = ncols(rows(([1,2,3],[4,5,6])))
2

julia> ncols(table(([1,2,3],[4,5,6])))
2

julia> ncols(table(@NT(x=[1,2,3],y=[4,5,6])))
2

julia> ncols(ndsparse(d, [7,8,9]))

3

source

4.22 IndexedTables.ndsparse
IndexedTables.ndsparse — Function.
ndsparse(indices, data; agg, presorted, copy, chunks)
Construct an NDSparse array with the given indices and data. Each vector
in indices represents the index values for one dimension. On construction, the
indices and data are sorted in lexicographic order of the indices.
Arguments:

• agg::Function: If indices contains duplicate entries, the corresponding

data items are reduced using this 2-argument function.
4.22. INDEXEDTABLES.NDSPARSE 131

• presorted::Bool: If true, the indices are assumed to already be sorted

and no sorting is done.
• copy::Bool: If true, the storage for the new array will not be shared with
the passed indices and data. If false (the default), the passed arrays will
be copied only if necessary for sorting. The only way to guarantee sharing
of data is to pass presorted=true.
• chunks::Integer: distribute the table into chunks (Integer) chunks (a
safe bet is nworkers()). Not distributed by default. See Distributed docs.

Examples:
1-dimensional NDSparse can be constructed with a single array as index.

julia> x = ndsparse(["a","b"],[3,4])
1-d NDSparse with 2 values (Int64):
1

"a" 3
"b" 4

julia> keytype(x), eltype(x)

(Tuple{String}, Int64)

A dimension will be named if constructed with a named tuple of columns as

index.

julia> x = ndsparse(@NT(date=Date.(2014:2017)), [4:7;])

1-d NDSparse with 4 values (Int64):
date

2014-01-01 4
2015-01-01 5
2016-01-01 6
2017-01-01 7

julia> x[Date("2015-01-01")]
5

julia> keytype(x), eltype(x)

(Tuple{Date}, Int64)

Multi-dimensional NDSparse can be constructed by passing a tuple of index

columns:

julia> x = ndsparse((["a","b"],[3,4]), [5,6])

2-d NDSparse with 2 values (Int64):
1 2
132 CHAPTER 4. INDEXEDTABLES

"a" 3 5
"b" 4 6

julia> keytype(x), eltype(x)

(Tuple{String,Int64}, Int64)

julia> x["a", 3]
5

The data itself can also contain tuples (these are stored in columnar format,
just like in table.)

julia> x = ndsparse((["a","b"],[3,4]), ([5,6], [7.,8.]))

2-d NDSparse with 2 values (2-tuples):
1 2 3 4

"a" 3 5 7.0
"b" 4 6 8.0

julia> x = ndsparse(@NT(x=["a","a","b"],y=[3,4,4]),
@NT(p=[5,6,7], q=[8.,9.,10.]))
2-d NDSparse with 3 values (2 field named tuples):
x y p q

"a" 3 5 8.0
"a" 4 6 9.0
"b" 4 7 10.0

julia> keytype(x), eltype(x)

(Tuple{String,Int64}, NamedTuples._NT_p_q{Int64,Float64})

julia> x["a", :]
2-d NDSparse with 2 values (2 field named tuples):
x y p q

"a" 3 5 8.0
"a" 4 6 9.0

Passing a chunks option to ndsparse, or constructing with a distributed

array will cause the result to be distributed. Use distribute function to dis-
tribute an array.

julia> x = ndsparse(@NT(date=Date.(2014:2017)), [4:7.;], chunks=2)

1-d Distributed NDSparse with 4 values (Float64) in 2 chunks:
date
4.23. INDEXEDTABLES.PAIRS 133

2014-01-01 4.0
2015-01-01 5.0
2016-01-01 6.0
2017-01-01 7.0

julia> x = ndsparse(@NT(date=Date.(2014:2017)), distribute([4:7.0;], 2))

1-d Distributed NDSparse with 4 values (Float64) in 2 chunks:
date

2014-01-01 4.0
2015-01-01 5.0
2016-01-01 6.0
2017-01-01 7.0

Distribution is done to match the first distributed column from left to right.
Specify chunks to override this.
source

4.23 IndexedTables.pairs
IndexedTables.pairs — Method.
pairs(arr::NDSparse, indices...)
Similar to where, but returns an iterator giving index=>value pairs. index
will be a tuple.
source

4.24 IndexedTables.pkeynames
IndexedTables.pkeynames — Method.
pkeynames(t::Table)
Names of the primary key columns in t.
Example

julia> t = table([1,2], [3,4]);

julia> pkeynames(t)
()

julia> t = table([1,2], [3,4], pkey=1);

julia> pkeynames(t)
(1,)

julia> t = table([2,1],[1,3],[4,5], names=[:x,:y,:z], pkey=(1,2));

134 CHAPTER 4. INDEXEDTABLES

julia> pkeys(t)
2-element IndexedTables.Columns{NamedTuples._NT_x_y{Int64,Int64},NamedTuples._NT_x_y{Ar
(x = 1, y = 3)
(x = 2, y = 1)

source

4.25 IndexedTables.pkeynames
IndexedTables.pkeynames — Method.

pkeynames(t::NDSparse)

Names of the primary key columns in t.

Example

julia> x = ndsparse([1,2],[3,4])
1-d NDSparse with 2 values (Int64):
1

1 3
2 4

julia> pkeynames(x)
(1,)

source

4.26 IndexedTables.pkeys
IndexedTables.pkeys — Method.

pkeys(itr::Table)

Primary keys of the table. If Table doesn’t have any designated primary
key columns (constructed without pkey argument) then a default key of tuples
(1,):(n,) is generated.
Example

julia> a = table(["a","b"], [3,4]) # no pkey

Table with 2 rows, 2 columns:
1 2
4.27. INDEXEDTABLES.POPCOL 135

"a" 3
"b" 4

julia> pkeys(a)
2-element Columns{Tuple{Int64}}:
(1,)
(2,)

julia> a = table(["a","b"], [3,4], pkey=1)

Table with 2 rows, 2 columns:
1 2

"a" 3
"b" 4

julia> pkeys(a)
2-element Columns{Tuple{String}}:
("a",)
("b",)

source

4.27 IndexedTables.popcol
IndexedTables.popcol — Method.
popcol(t, col)
Remove the column col from the table. Returns a new table.
popcol(t, cols...)
Remove many columns at a time.

julia> t = table([0.01, 0.05], [2,1], [3,4], names=[:t, :x, :y], pkey=:t)

Table with 2 rows, 3 columns:
t x y

0.01 2 3
0.05 1 4

julia> popcol(t, :x)

Table with 2 rows, 2 columns:
t y

0.01 3
0.05 4

source
136 CHAPTER 4. INDEXEDTABLES

4.28 IndexedTables.pushcol
IndexedTables.pushcol — Method.
pushcol(t, name, x)
Push a column x to the end of the table. name is the name for the new
column. Returns a new table.
pushcol(t, map::Pair...)
Push many columns at a time.
Example:

julia> t = table([0.01, 0.05], [2,1], [3,4], names=[:t, :x, :y], pkey=:t)

Table with 2 rows, 3 columns:
t x y

0.01 2 3
0.05 1 4

julia> pushcol(t, :z, [1//2, 3//4])

Table with 2 rows, 4 columns:
t x y z

0.01 2 3 1//2
0.05 1 4 3//4

source

4.29 IndexedTables.reducedim vec

IndexedTables.reducedim vec — Method.
reducedim vec(f::Function, arr::NDSparse, dims)
Like reducedim, except uses a function mapping a vector of values to a scalar
instead of a 2-argument scalar function.
source

4.30 IndexedTables.reindex
IndexedTables.reindex — Function.
reindex(t::Table, by[, select])
Reindex t by columns selected in by. Keeps columns selected by select as
non-indexed columns. By default all columns not mentioned in by are kept.
Use selectkeys to reindex and NDSparse object.

julia> t = table([2,1],[1,3],[4,5], names=[:x,:y,:z], pkey=(1,2))

julia> reindex(t, (:y, :z))

4.31. INDEXEDTABLES.RENAMECOL 137

Table with 2 rows, 3 columns:

y z x

1 4 2
3 5 1

julia> pkeynames(t)
(:y, :z)

julia> reindex(t, (:w=>[4,5], :z))

Table with 2 rows, 4 columns:
w z x y

4 5 1 3
5 4 2 1

julia> pkeynames(t)
(:w, :z)

source

4.31 IndexedTables.renamecol
IndexedTables.renamecol — Method.
renamecol(t, col, newname)
Set newname as the new name for column col in t. Returns a new table.
renamecol(t, map::Pair...)
Rename many columns at a time.

julia> t = table([0.01, 0.05], [2,1], names=[:t, :x])

Table with 2 rows, 2 columns:
t x

0.01 2
0.05 1

julia> renamecol(t, :t, :time)

Table with 2 rows, 2 columns:
time x

0.01 2
0.05 1

source
138 CHAPTER 4. INDEXEDTABLES

4.32 IndexedTables.rows
IndexedTables.rows — Function.
rows(itr[, select::Selection])
Select one or more fields from an iterable of rows as a vector of their values.
select specifies which fields to select. See Selection convention for pos-
sible values. If unspecified, returns all columns.
itr can be NDSparse, Columns and AbstractVector, and their distributed
counterparts.
Examples

julia> t = table([1,2],[3,4], names=[:x,:y])

Table with 2 rows, 2 columns:
x y

1 3
2 4

julia> rows(t)
2-element IndexedTables.Columns{NamedTuples._NT_x_y{Int64,Int64},NamedTuples._NT_x_y{Ar
(x = 1, y = 3)
(x = 2, y = 4)

julia> rows(t, :x)

2-element Array{Int64,1}:
1
2

julia> rows(t, (:x,))

2-element IndexedTables.Columns{NamedTuples._NT_x{Int64},NamedTuples._NT_x{Array{Int64,
(x = 1)
(x = 2)

julia> rows(t, (:y,:x=>-))

2-element IndexedTables.Columns{NamedTuples._NT_y_x{Int64,Int64},NamedTuples._NT_y_x{Ar
(y = 3, x = -1)
(y = 4, x = -2)

Note that vectors of tuples returned are Columns object and have columnar
internal storage.
source

4.33 IndexedTables.setcol
IndexedTables.setcol — Method.
setcol(t::Table, col::Union{Symbol, Int}, x::Selection)
4.33. INDEXEDTABLES.SETCOL 139

Sets a x as the column identified by col. Returns a new table.

setcol(t::Table, map::Pair...)
Set many columns at a time.
Examples:
julia> t = table([1,2], [3,4], names=[:x, :y])
Table with 2 rows, 2 columns:
x y

1 3
2 4

julia> setcol(t, 2, [5,6])

Table with 2 rows, 2 columns:
x y

1 5
2 6
x can be any selection that transforms existing columns.
julia> setcol(t, :x, :x => x->1/x)
Table with 2 rows, 2 columns:
x y

1.0 5
0.5 6
setcol will result in a re-sorted copy if a primary key column is replaced.
julia> t = table([0.01, 0.05], [1,2], [3,4], names=[:t, :x, :y], pkey=:t)
Table with 2 rows, 3 columns:
t x y

0.01 1 3
0.05 2 4

julia> t2 = setcol(t, :t, [0.1,0.05])

Table with 2 rows, 3 columns:
t x y

0.05 2 4
0.1 1 3

julia> t == t2
false
If col is not an existing column, setcol will add it:
140 CHAPTER 4. INDEXEDTABLES

julia> t = table([1,2], [2,3], names = [:a,:b])

Table with 2 rows, 2 columns:
a b

1 2
2 3

julia> setcol(t, :c, [1,2])

Table with 2 rows, 3 columns:
a b c

1 2 1
2 3 2

source

4.34 IndexedTables.stack
IndexedTables.stack — Method.
stack(t, by = pkeynames(t); select = excludecols(t, by), variable
= :variable, value = :value)
Reshape a table from the wide to the long format. Columns in by are
kept as indexing columns. Columns in select are stacked. In addition to the
id columns, two additional columns labeled variable and value are added,
containg the column identifier and the stacked columns.
Examples

julia> t = table(1:4, [1, 4, 9, 16], [1, 8, 27, 64], names = [:x, :xsquare, :xcube], pk

julia> stack(t)
Table with 8 rows, 3 columns:
x variable value

1 :xsquare 1
1 :xcube 1
2 :xsquare 4
2 :xcube 8
3 :xsquare 9
3 :xcube 27
4 :xsquare 16
4 :xcube 64

source
4.35. INDEXEDTABLES.SUMMARIZE 141

4.35 IndexedTables.summarize
IndexedTables.summarize — Function.
summarize(f, t, by = pkeynames(t); select = excludecols(t, by),
stack = false, variable = :variable)
Apply summary functions column-wise to a table. Return a NamedTuple in
the non-grouped case and a table in the grouped case. Use stack=true to stack
results of the same summary function for different columns.
Examples

julia> t = table([1, 2, 3], [1, 1, 1], names = [:x, :y]);

julia> summarize((mean, std), t)

(x_mean = 2.0, y_mean = 1.0, x_std = 1.0, y_std = 0.0)

julia> s = table(["a","a","b","b"], [1,3,5,7], [2,2,2,2], names = [:x, :y, :z], pkey = :x);

julia> summarize(mean, s)
Table with 2 rows, 3 columns:
x y z

"a" 2.0 2.0

"b" 6.0 2.0

julia> summarize(mean, s, stack = true)

Table with 4 rows, 3 columns:
x variable mean

"a" :y 2.0
"a" :z 2.0
"b" :y 6.0
"b" :z 2.0

Use a NamedTuple to have different names for the summary functions:

julia> summarize(@NT(m = mean, s = std), t)

(x_m = 2.0, y_m = 1.0, x_s = 1.0, y_s = 0.0)

Use select to only summarize some columns:

julia> summarize(@NT(m = mean, s = std), t, select = :x)

(m = 2.0, s = 1.0)

source
142 CHAPTER 4. INDEXEDTABLES

4.36 IndexedTables.table
IndexedTables.table — Function.
table(cols::AbstractVector...; names, <options>)
Create a table with columns given by cols.

julia> a = table([1,2,3], [4,5,6])

Table with 3 rows, 2 columns:
1 2

1 4
2 5
3 6

names specify names for columns. If specified, the table will be an iterator
of named tuples.

julia> b = table([1,2,3], [4,5,6], names=[:x, :y])

Table with 3 rows, 2 columns:
x y

1 4
2 5
3 6

table(cols::Union{Tuple, NamedTuple}; <options>)

Convert a struct of columns to a table of structs.

julia> table(([1,2,3], [4,5,6])) == a

true

julia> table(@NT(x=[1,2,3], y=[4,5,6])) == b

true

table(cols::Columns; <options>)
Construct a table from a vector of tuples. See rows.

julia> table(Columns([1,2,3], [4,5,6])) == a

true

julia> table(Columns(x=[1,2,3], y=[4,5,6])) == b

true

table(t::Union{Table, NDSparse}; <options>)

Copy a Table or NDSparse to create a new table. The same primary keys
as the input are used.

julia> b == table(b)
true
4.36. INDEXEDTABLES.TABLE 143

table(iter; <options>)
Construct a table from an iterable table.
Options:

• pkey: select columns to act as the primary key. By default, no columns

are used as primary key.

• presorted: is the data pre-sorted by primary key columns? If so, skip

sorting. false by default. Irrelevant if chunks is specified.

• copy: creates a copy of the input vectors if true. true by default. Irrela-
vant if chunks is specified.

• chunks: distribute the table into chunks (Integer) chunks (a safe bet is
nworkers()). Table is not distributed by default. See Distributed docs.

Examples:
Specifying pkey will cause the table to be sorted by the columns named in
pkey:

julia> b = table([2,3,1], [4,5,6], names=[:x, :y], pkey=:x)

Table with 3 rows, 2 columns:
x y

1 6
2 4
3 5

julia> b = table([2,1,2,1],[2,3,1,3],[4,5,6,7],
names=[:x, :y, :z], pkey=(:x,:y))
Table with 4 rows, 3 columns:
x y z

1 3 5
1 3 7
2 1 6
2 2 4

Note that the keys do not have to be unique.

chunks option creates a distributed table.
chunks can be:

1. An integer – number of chunks to create

2. An vector of k integers – number of elements in each of the k chunks.

3. The distribution of another array. i.e. vec.subdomains where vec is a

distributed array.
144 CHAPTER 4. INDEXEDTABLES

julia> t = table([2,3,1,4], [4,5,6,7],

names=[:x, :y], pkey=:x, chunks=2)
Distributed Table with 4 rows in 2 chunks:
x y

1 6
2 4
3 5
4 7
A distributed table will be constructed if one of the arrays passed into table
constructor is a distributed array. A distributed Array can be constructed using
distribute:

julia> x = distribute([1,2,3,4], 2);

julia> t = table(x, [5,6,7,8], names=[:x,:y])

Distributed Table with 4 rows in 2 chunks:
x y

1 5
2 6
3 7
4 8

julia> table(columns(t)..., [9,10,11,12],

names=[:x,:y,:z])
Distributed Table with 4 rows in 2 chunks:
x y z

1 5 9
2 6 10
3 7 11
4 8 12
Distribution is done to match the first distributed column from left to right.
Specify chunks to override this.
source

4.37 IndexedTables.unstack
IndexedTables.unstack — Method.
unstack(t, by = pkeynames(t); variable = :variable, value = :value)
Reshape a table from the long to the wide format. Columns in by are kept
as indexing columns. Keyword arguments variable and value denote which
column contains the column identifier and which the corresponding values.
4.38. INDEXEDTABLES.UPDATE! 145

Examples

julia> t = table(1:4, [1, 4, 9, 16], [1, 8, 27, 64], names = [:x, :xsquare, :xcube], pkey = :x);

julia> long = stack(t)

Table with 8 rows, 3 columns:
x variable value

1 :xsquare 1
1 :xcube 1
2 :xsquare 4
2 :xcube 8
3 :xsquare 9
3 :xcube 27
4 :xsquare 16
4 :xcube 64

julia> unstack(long)
Table with 4 rows, 3 columns:
x xsquare xcube

1 1 1
2 4 8
3 9 27
4 16 64

source

4.38 IndexedTables.update!
IndexedTables.update! — Method.
update!(f::Function, arr::NDSparse, indices...)
Replace data values x with f(x) at each location that matches the given
indices.
source

4.39 IndexedTables.where
IndexedTables.where — Method.
where(arr::NDSparse, indices...)
Returns an iterator over data items where the given indices match. Accepts
the same index arguments as getindex.
source
Chapter 5

Images

5.1 ColorTypes.blue
ColorTypes.blue — Function.
b = blue(img) extracts the blue channel from an RGB image img
source

5.2 ColorTypes.green
ColorTypes.green — Function.
g = green(img) extracts the green channel from an RGB image img
source

5.3 ColorTypes.red
ColorTypes.red — Function.
r = red(img) extracts the red channel from an RGB image img
source

5.4 Images.adjust gamma

Images.adjust gamma — Method.

gamma_corrected_img = adjust_gamma(img, gamma)

Returns a gamma corrected image.

The adjust gamma function can handle a variety of input types. The re-
turned image depends on the input type. If the input is an Image then the
resulting image is of the same type and has the same properties.

146
5.5. IMAGES.BILINEAR INTERPOLATION 147

For coloured images, the input is converted to YIQ type and the Y channel
is gamma corrected. This is the combined with the I and Q channels and the
resulting image converted to the same type as the input.
source

5.5 Images.bilinear interpolation

Images.bilinear interpolation — Method.

P = bilinear_interpolation(img, r, c)

Bilinear Interpolation is used to interpolate functions of two variables on a

rectilinear 2D grid.
The interpolation is done in one direction first and then the values obtained
are used to do the interpolation in the second direction.
source

5.6 Images.blob LoG

Images.blob LoG — Method.

blob_LoG(img, scales, [edges], [shape]) -> Vector{BlobLoG}

Find “blobs” in an N-D image using the negative Lapacian of Gaussians with
the specifed vector or tuple of values. The algorithm searches for places where
the filtered image (for a particular ) is at a peak compared to all spatially- and
-adjacent voxels, where is scales[i] * shape for some i. By default, shape
is an ntuple of 1s.
The optional edges argument controls whether peaks on the edges are in-
cluded. edges can be true or false, or a N+1-tuple in which the first entry
controls whether edge- values are eligible to serve as peaks, and the remaining
N entries control each of the N dimensions of img.
Citation:
Lindeberg T (1998), “Feature Detection with Automatic Scale Selection”,
International Journal of Computer Vision, 30(2), 79–116.
See also: BlobLoG.
source

5.7 Images.boxdiff
Images.boxdiff — Method.

sum = boxdiff(integral_image, ytop:ybot, xtop:xbot)

sum = boxdiff(integral_image, CartesianIndex(tl_y, tl_x), CartesianIndex(br_y, br_x))
sum = boxdiff(integral_image, tl_y, tl_x, br_y, br_x)
148 CHAPTER 5. IMAGES

An integral image is a data structure which helps in efficient calculation of

sum of pixels in a rectangular subset of an image. It stores at each pixel the sum
of all pixels above it and to its left. The sum of a window in an image can be
directly calculated using four array references of the integral image, irrespective
of the size of the window, given the yrange and xrange of the window. Given
an integral image -

A - - - - - - B -
- * * * * * * * -
- * * * * * * * -
- * * * * * * * -
- * * * * * * * -
- * * * * * * * -
C * * * * * * D -
- - - - - - - - -

The sum of pixels in the area denoted by * is given by S = D + A - B - C.

source

5.8 Images.canny
Images.canny — Method.

canny_edges = canny(img, (upper, lower), sigma=1.4)

Performs Canny Edge Detection on the input image.

Parameters :
(upper, lower) : Bounds for hysteresis thresholding sigma : Specifies the
standard deviation of the gaussian filter
Example
[] imgedg = canny(img, (Percentile(80), Percentile(20)))
source

5.9 Images.clahe
Images.clahe — Method.

hist_equalised_img = clahe(img, nbins, xblocks = 8, yblocks = 8, clip = 3)

Performs Contrast Limited Adaptive Histogram Equalisation (CLAHE) on

the input image. It differs from ordinary histogram equalization in the respect
that the adaptive method computes several histograms, each corresponding to
a distinct section of the image, and uses them to redistribute the lightness
values of the image. It is therefore suitable for improving the local contrast and
enhancing the definitions of edges in each region of an image.
5.10. IMAGES.CLEARBORDER 149

In the straightforward form, CLAHE is done by calculation a histogram

of a window around each pixel and using the transformation function of the
equalised histogram to rescale the pixel. Since this is computationally expen-
sive, we use interpolation which gives a significant rise in efficiency without
compromising the result. The image is divided into a grid and equalised his-
tograms are calculated for each block. Then, each pixel is interpolated using
the closest histograms.
The xblocks and yblocks specify the number of blocks to divide the input
image into in each direction. nbins specifies the granularity of histogram cal-
culation of each local region. clip specifies the value at which the histogram is
clipped. The excess in the histogram bins with value exceeding clip is redis-
tributed among the other bins.
source

5.10 Images.clearborder
Images.clearborder — Function.

cleared_img = clearborder(img)
cleared_img = clearborder(img, width)
cleared_img = clearborder(img, width, background)

Returns a copy of the original image after clearing objects connected to the
border of the image.
Parameters:

• img = Binary/Grayscale input image

• width = Width of the border examined (Default value is 1)

• background = Value to be given to pixels/elements that are cleared (De-

fault value is 0)

source

5.11 Images.cliphist
Images.cliphist — Method.

clipped_hist = cliphist(hist, clip)

Clips the histogram above a certain value clip. The excess left in the bins
exceeding clip is redistributed among the remaining bins.
source
150 CHAPTER 5. IMAGES

5.12 Images.complement
Images.complement — Method.

y = complement(x)

Take the complement 1-x of x. If x is a color with an alpha channel, the

alpha channel is left untouched. Don’t forget to add a dot when x is an array:
complement.(x)
source

5.13 Images.component boxes

Images.component boxes — Method.
component boxes(labeled array) -> an array of bounding boxes for each
label, including the background label 0
source

5.14 Images.component centroids

Images.component centroids — Method.
component centroids(labeled array) -> an array of centroids for each
label, including the background label 0
source

5.15 Images.component indices

Images.component indices — Method.
component indices(labeled array) -> an array of pixels for each label,
including the background label 0
source

5.16 Images.component lengths

Images.component lengths — Method.
component lengths(labeled array) -> an array of areas (2D), volumes
(3D), etc. for each label, including the background label 0
source

5.17 Images.component subscripts

Images.component subscripts — Method.
component subscripts(labeled array) -> an array of pixels for each la-
bel, including the background label 0
5.18. IMAGES.CONVEXHULL 151

source

5.18 Images.convexhull
Images.convexhull — Method.

chull = convexhull(img)

Computes the convex hull of a binary image and returns the vertices of
convex hull as a CartesianIndex array.
source

5.19 Images.corner2subpixel
Images.corner2subpixel — Method.

corners = corner2subpixel(responses::AbstractMatrix,corner_indicator::AbstractMatrix{Bool})
-> Vector{HomogeneousPoint{Float64,3}}

Refines integer corner coordinates to sub-pixel precision.

The function takes as input a matrix representing corner responses and a
boolean indicator matrix denoting the integer coordinates of a corner in the
image. The output is a vector of type HomogeneousPoint storing the sub-pixel
coordinates of the corners.
The algorithm computes a correction factor which is added to the original
integer coordinates. In particular, a univariate quadratic polynomial is fit sep-
arately to the x-coordinates and y-coordinates of a corner and its immediate
east/west, and north/south neighbours. The fit is achieved using a local coor-
dinate system for each corner, where the origin of the coordinate system is a
given corner, and its immediate neighbours are assigned coordinates of minus
one and plus one.
The corner and its two neighbours form a system of three equations. For
example, let x1 = −1, x2 = 0 and x3 = 1 denote the local x coordinates of
the west, center and east pixels and let the vector b = [r1 , r2 , r3 ] denote the
corresponding corner response values. With
 2 
x1 x1 1
A = x22 x2 1 ,
x23 x3 1
the coefficients of the quadratic polynomial can be found by solving the
system of equations b = Ax. The result is given by x = A−1 b.
The vertex of the quadratic polynomial yields a sub-pixel estimate of the true
corner position. For example, for a univariate quadratic polynomial px2 +qx+r,
the x-coordinate of the vertex is −q
2p . Hence, the refined sub-pixel coordinate is
equal to: c + −q
2p , where c is the integer coordinate.
152 CHAPTER 5. IMAGES

!!! note Corners on the boundary of the image are not refined to sub-pixel
precision.
source

5.20 Images.entropy
Images.entropy — Method.

entropy(log, img)
entropy(img; [kind=:shannon])

Compute the entropy of a grayscale image defined as -sum(p.*log(p)). The

base of the logarithm (a.k.a. entropy unit) is one of the following:

• :shannon (log base 2, default), or use log = log2

• :nat (log base e), or use log = log

• :hartley (log base 10), or use log = log10

source

5.21 Images.fastcorners
Images.fastcorners — Method.

fastcorners(img, n, threshold) -> corners

Performs FAST Corner Detection. n is the number of contiguous pixels which

need to be greater (lesser) than intensity + threshold (intensity - threshold) for
a pixel to be marked as a corner. The default value for n is 12.
source

5.22 Images.findlocalmaxima
Images.findlocalmaxima — Function.
findlocalmaxima(img, [region, edges]) -> Vector{CartesianIndex}
Returns the coordinates of elements whose value is larger than all of their
immediate neighbors. region is a list of dimensions to consider. edges is a
boolean specifying whether to include the first and last elements of each dimen-
sion, or a tuple-of-Bool specifying edge behavior for each dimension separately.
source
5.23. IMAGES.FINDLOCALMINIMA 153

5.23 Images.findlocalminima
Images.findlocalminima — Function.
Like findlocalmaxima, but returns the coordinates of the smallest elements.
source

5.24 Images.gaussian pyramid

Images.gaussian pyramid — Method.

pyramid = gaussian_pyramid(img, n_scales, downsample, sigma)

Returns a gaussian pyramid of scales n scales, each downsampled by a

factor downsample and sigma for the gaussian kernel.
source

5.25 Images.harris
Images.harris — Method.

harris_response = harris(img; [k], [border], [weights])

Performs Harris corner detection. The covariances can be taken using either
a mean weighted filter or a gamma kernel.
source

5.26 Images.histeq
Images.histeq — Method.

hist_equalised_img = histeq(img, nbins)

hist_equalised_img = histeq(img, nbins, minval, maxval)

Returns a histogram equalised image with a granularity of approximately

nbins number of bins.
Details
Histogram equalisation was initially conceived to improve the contrast in a
single-channel grayscale image. The method transforms the distribution of the
intensities in an image so that they are as uniform as possible [1]. The natural
justification for uniformity is that the image has better contrast if the intensity
levels of an image span a wide range on the intensity scale. As it turns out, the
necessary transformation is a mapping based on the cumulative histogram.
One can consider an L-bit single-channel I × J image with gray values in the
set {0, 1, . . . , L − 1}, as a collection of independent and identically distributed
random variables. Specifically, let the sample space Ω be the set of all IJ-tuples
154 CHAPTER 5. IMAGES

ω = (ω11 , ω12 , . . . , ω1J , ω21 , ω22 , . . . , ω2J , ωI1 , ωI2 , . . . , ωIJ ), where each ωij ∈
{0, 1, . . . , L − 1}. Furthermore, impose a probability measure on Ω such that
the functions Ω 3 ω → ωij ∈ {0, 1, . . . , L − 1} are independent and identically
distributed.
One can then regard an image as a matrix of random variables G = [Gi,j (ω)],
where each function Gi,j : Ω → R is defined by
ωij
Gi,j (ω) = ,
L−1
and each Gi,j is distributed according to some unknown density fG . While
fG is unknown, one can approximate it with a normalised histogram of gray
levels,
nv
fˆG (v) = ,
IJ
where

nv = |{(i, j) | Gi,j (ω) = v}|

represents the number of times a gray level with intensity v occurs in G. To
transforming the distribution of the intensities so that they are as uniform as
possible one needs to find a mapping T (·) such that T (Gi,j ) ∼ U . The required
mapping turns out to be the cumulative distribution function (CDF) of the
empirical density fˆG ,
Z Gi,j
T (Gi,j ) = fˆG (w)dw.
0

Options
Various options for the parameters of this function are described in more
detail below.
Choices for img
The histeq function can handle a variety of input types. The returned
image depends on the input type. If the input is an Image then the resulting
image is of the same type and has the same properties.
For coloured images, the input is converted to YIQ type and the Y channel
is equalised. This is the combined with the I and Q channels and the resulting
image converted to the same type as the input.
Choices for nbins
You can specify the total number of bins in the histogram.
Choices for minval and maxval
If minval and maxval are specified then intensities are equalized to the range
(minval, maxval). The default values are 0 and 1.
Example
[]
using TestImages, FileIO, ImageView
img = testimage(”mandrilg ray”);imgeq = histeq(img,256);
5.27. IMAGES.HISTMATCH 155

imshow(img) imshow(imgeq)
References

1. R. C. Gonzalez and R. E. Woods. Digital Image Processing (3rd Edition).

Upper Saddle River, NJ, USA: Prentice-Hall, 2006.

See also: histmatch,clahe, imhist and adjust gamma.

source

5.27 Images.histmatch
Images.histmatch — Function.

hist_matched_img = histmatch(img, oimg, nbins)

Returns a grayscale histogram matched image with a granularity of nbins

number of bins. img is the image to be matched and oimg is the image having
the desired histogram to be matched to.
source

5.28 Images.imROF
Images.imROF — Method.

imgr = imROF(img, , iterations)

Perform Rudin-Osher-Fatemi (ROF) filtering, more commonly known as To-

tal Variation (TV) denoising or TV regularization. is the regularization coeffi-
cient for the derivative, and iterations is the number of relaxation iterations
taken. 2d only.
See https://en.wikipedia.org/wiki/Total variation denoising and Chambolle,
A. (2004). “An algorithm for total variation minimization and applications”.
Journal of Mathematical Imaging and Vision. 20: 89–97
source

5.29 Images.imadjustintensity
Images.imadjustintensity — Method.
imadjustintensity(img [, (minval,maxval)]) -> Image
Map intensities over the interval (minval,maxval) to the interval [0,1].
This is equivalent to map(ScaleMinMax(eltype(img), minval, maxval),
img). (minval,maxval) defaults to extrema(img).
source
156 CHAPTER 5. IMAGES

5.30 Images.imcorner
Images.imcorner — Method.
corners = imcorner(img; [method])
corners = imcorner(img, threshold, percentile; [method])
Performs corner detection using one of the following methods -
1. harris
2. shi_tomasi
3. kitchen_rosenfeld
The parameters of the individual methods are described in their documen-
tation. The maxima values of the resultant responses are taken as corners. If a
threshold is specified, the values of the responses are thresholded to give the cor-
ner pixels. The threshold is assumed to be a percentile value unless percentile
is set to false.
source

5.31 Images.imcorner subpixel

Images.imcorner subpixel — Method.
corners = imcorner_subpixel(img; [method])
-> Vector{HomogeneousPoint{Float64,3}}
corners = imcorner_subpixel(img, threshold, percentile; [method])
-> Vector{HomogeneousPoint{Float64,3}}
Same as imcorner, but estimates corners to sub-pixel precision.
Sub-pixel precision is achieved by interpolating the corner response val-
ues using the 4-connected neighbourhood of a maximum response value. See
corner2subpixel for more details of the interpolation scheme.
source

5.32 Images.imedge
Images.imedge — Function.
grad_y, grad_x, mag, orient = imedge(img, kernelfun=KernelFactors.ando3, border="replic
Edge-detection filtering. kernelfun is a valid kernel function for imgradients,
defaulting to KernelFactors.ando3. border is any of the boundary conditions
specified in padarray.
Returns a tuple (grad y, grad x, mag, orient), which are the horizontal
gradient, vertical gradient, and the magnitude and orientation of the strongest
edge, respectively.
source
5.33. IMAGES.IMHIST 157

5.33 Images.imhist
Images.imhist — Method.

edges, count = imhist(img, nbins)

edges, count = imhist(img, nbins, minval, maxval)
edges, count = imhist(img, edges)

Generates a histogram for the image over nbins spread between (minval,
maxval]. Color images are automatically converted to grayscale.
Output
Returns edges which is a range type that specifies how the interval (minval,
maxval] is divided into bins, and an array count which records the concomitant
bin frequencies. In particular, count has the following properties:

• count[i+1] is the number of values x that satisfy edges[i] <= x <

edges[i+1].

• count[1] is the number satisfying x < edges[1], and

• count[end] is the number satisfying x >= edges[end].

• length(count) == length(edges)+1.

Details
One can consider a histogram as a piecewise-constant model of a probability
density function f [1]. Suppose that f has support on some interval I = [a, b].
Let m be an integer and a = a1 < a2 < . . . < am < am+1 = b a sequence of real
numbers. Construct a sequence of intervals

I1 = [a1 , a2 ], I2 = (a2 , a3 ], . . . , Im = (am , am+1 ]

which partition I into subsets Ij (j = 1, . . . , m) on which f is constant. These

subsets satisfy Ii ∩Ij = ∅, ∀i 6= j, and are commonly referredP to as bins. Together
they encompass the entire range of data values such that j |Ij | = |I|. Each bin
has width wj = |Ij | = aj+1 − aj and height hj which is the constant probability
density over the region of the bin. Integrating the constant probability density
over the width of the bin wj yields a probability mass of πj = hj wj for the bin.
For a sample x1 , x2 , . . . , xN , let

N
(
X 1 ifx ∈ Ij ,
nj = 1(Ij ) (xn ), where 1(Ij ) (x) = ,
n=1
0 otherwise,

represents the number of samples falling into the interval Ij . An estimate

n
for the probability mass of the jth bin is given by the relative frequency π̂ = Nj ,
and the histogram estimator of the probability density function is defined as
158 CHAPTER 5. IMAGES

m
X nj
fˆn (x) = 1(I ) (x)
j=1
N wj j
m
X π̂j
= 1(Ij ) (x)
w
j=1 j
m
X
= ĥj 1(Ij ) (x).
j=1

The function fˆn (x) is a genuine density estimator because fˆn (x) ≥ 0 and
Z ∞ m
X nj
fˆn (x) d x = wj
−∞ j=1
N wj
= 1.

Options
Various options for the parameters of this function are described in more
detail below.
Choices for nbins
You can specify the number of discrete bins for the histogram.
Choices for minval
You have the option to specify the lower bound of the interval over which
the histogram will be computed. If minval is not specified then the minimum
value present in the image is taken as the lower bound.
Choices for maxval
You have the option to specify the upper bound of the interval over which
the histogram will be computed. If maxval is not specified then the maximum
value present in the image is taken as the upper bound.
Choices for edges
If you do not designate the number of bins, nor the lower or upper bound
of the interval, then you have the option to directly stipulate how the intervals
will be divided by specifying a range type.
Example
Compute the histogram of a grayscale image.
[]
using TestImages, FileIO, ImageView
img = testimage(”mandrilg ray”);edges, counts = imhist(img,256);
Given a color image, compute the hisogram of the red channel.
[] img = testimage(”mandrill”) r = red(img) edges, counts = imhist(r,256);
References
[1] E. Herrholz, “Parsimonious Histograms,” Ph.D. dissertation, Inst. of
Math. and Comp. Sci., University of Greifswald, Greifswald, Germany, 2011.
source
5.34. IMAGES.IMSTRETCH 159

5.34 Images.imstretch
Images.imstretch — Method.
imgs = imstretch(img, m, slope) enhances or reduces (for slope > 1 or
< 1, respectively) the contrast near saturation (0 and 1). This is essentially a
symmetric gamma-correction. For a pixel of brightness p, the new intensity is
1/(1+(m/(p+eps))^slope).
This assumes the input img has intensities between 0 and 1.
source

5.35 Images.integral image

Images.integral image — Method.

integral_img = integral_image(img)

Returns the integral image of an image. The integral image is calculated

by assigning to each pixel the sum of all pixels above it and to its left, i.e. the
rectangle from (1, 1) to the pixel. An integral image is a data structure which
helps in efficient calculation of sum of pixels in a rectangular subset of an image.
See boxdiff for more information.
source

5.36 Images.kitchen rosenfeld

Images.kitchen rosenfeld — Method.

kitchen_rosenfeld_response = kitchen_rosenfeld(img; [border])

Performs Kitchen Rosenfeld corner detection. The covariances can be taken

using either a mean weighted filter or a gamma kernel.
source

5.37 Images.label components

Images.label components — Function.

label = label_components(tf, [connectivity])

label = label_components(tf, [region])

Find the connected components in a binary array tf. There are two forms
that connectivity can take:

• It can be a boolean array of the same dimensionality as tf, of size 1 or 3

160 CHAPTER 5. IMAGES

along each dimension. Each entry in the array determines whether a given
neighbor is used for connectivity analyses. For example, connectivity = trues(3,3)
would use 8-connectivity and test all pixels that touch the current one, even the
corners.

• You can provide a list indicating which dimensions are used to

determine connectivity. For example, region = [1,3] would not test neigh-
bors along dimension 2 for connectivity. This corresponds to just the nearest
neighbors, i.e., 4-connectivity in 2d and 6-connectivity in 3d.
The default is region = 1:ndims(A).
The output label is an integer array, where 0 is used for background pixels,
and each connected region gets a different integer index.
source

5.38 Images.magnitude
Images.magnitude — Method.

m = magnitude(grad_x, grad_y)

Calculates the magnitude of the gradient images given by grad x and grad y.
Equivalent to sqrt(gradx .2 + grady .2 ).
Returns a magnitude image the same size as grad x and grad y.
source

5.39 Images.magnitude phase

Images.magnitude phase — Method.

magnitude_phase(grad_x, grad_y) -> m, p

Convenience function for calculating the magnitude and phase of the gradient
images given in grad x and grad y. Returns a tuple containing the magnitude
and phase images. See magnitude and phase for details.
source

5.40 Images.maxabsfinite
Images.maxabsfinite — Method.
m = maxabsfinite(A) calculates the maximum absolute value in A, ignoring
any values that are not finite (Inf or NaN).
source
5.41. IMAGES.MAXFINITE 161

5.41 Images.maxfinite
Images.maxfinite — Method.
m = maxfinite(A) calculates the maximum value in A, ignoring any values
that are not finite (Inf or NaN).
source

5.42 Images.meanfinite
Images.meanfinite — Method.
M = meanfinite(img, region) calculates the mean value along the dimen-
sions listed in region, ignoring any non-finite values.
source

5.43 Images.minfinite
Images.minfinite — Method.
m = minfinite(A) calculates the minimum value in A, ignoring any values
that are not finite (Inf or NaN).
source

5.44 Images.ncc
Images.ncc — Method.
C = ncc(A, B) computes the normalized cross-correlation of A and B.
source

5.45 Images.orientation
Images.orientation — Method.

orientation(grad_x, grad_y) -> orient

Calculate the orientation angle of the strongest edge from gradient images
given by grad x and grad y. Equivalent to atan2(grad x, grad y). When
both grad x and grad y are effectively zero, the corresponding angle is set to
zero.
source

5.46 Images.otsu threshold

Images.otsu threshold — Method.
162 CHAPTER 5. IMAGES

thres = otsu_threshold(img)
thres = otsu_threshold(img, bins)

Computes threshold for grayscale image using Otsu’s method.

Parameters:

• img = Grayscale input image

• bins = Number of bins used to compute the histogram. Needed for

floating-point images.

source

5.47 Images.phase
Images.phase — Method.

phase(grad_x, grad_y) -> p

Calculate the rotation angle of the gradient given by grad x and grad y.
Equivalent to atan2(-grad y, grad x), except that when both grad x and
grad y are effectively zero, the corresponding angle is set to zero.
source

5.48 Images.sad
Images.sad — Method.
s = sad(A, B) computes the sum-of-absolute differences over arrays/images
A and B
source

5.49 Images.sadn
Images.sadn — Method.
s = sadn(A, B) computes the sum-of-absolute differences over arrays/images
A and B, normalized by array size
source

5.50 Images.shepp logan

Images.shepp logan — Method.

phantom = shepp_logan(N,[M]; highContrast=true)

5.51. IMAGES.SHI TOMASI 163

output the NxM Shepp-Logan phantom, which is a standard test image usu-
ally used for comparing image reconstruction algorithms in the field of computed
tomography (CT) and magnetic resonance imaging (MRI). If the argument M
is omitted, the phantom is of size NxN. When setting the keyword argument
highConstrast to false, the CT version of the phantom is created. Otherwise,
the high contrast MRI version is calculated.
source

5.51 Images.shi tomasi

Images.shi tomasi — Method.

shi_tomasi_response = shi_tomasi(img; [border], [weights])

Performs Shi Tomasi corner detection. The covariances can be taken using
either a mean weighted filter or a gamma kernel.
source

5.52 Images.ssd
Images.ssd — Method.
s = ssd(A, B) computes the sum-of-squared differences over arrays/images
A and B
source

5.53 Images.ssdn
Images.ssdn — Method.
s = ssdn(A, B) computes the sum-of-squared differences over arrays/images
A and B, normalized by array size
source

5.54 Images.thin edges

Images.thin edges — Method.

thinned = thin_edges(img, gradientangle, [border])

thinned, subpix = thin_edges_subpix(img, gradientangle, [border])
thinned, subpix = thin_edges_nonmaxsup(img, gradientangle, [border]; [radius::Float64=1.35], [the
thinned, subpix = thin_edges_nonmaxsup_subpix(img, gradientangle, [border]; [radius::Float64=1.35

Edge thinning for 2D edge images. Currently the only algorithm available
is non-maximal suppression, which takes an edge image and its gradient angle,
164 CHAPTER 5. IMAGES

and checks each edge point for local maximality in the direction of the gradient.
The returned image is non-zero only at maximal edge locations.
border is any of the boundary conditions specified in padarray.
In addition to the maximal edge image, the subpix versions of these func-
tions also return an estimate of the subpixel location of each local maxima, as a
2D array or image of Graphics.Point objects. Additionally, each local maxima
is adjusted to the estimated value at the subpixel location.
Currently, the nonmaxsup functions are identical to the first two function
calls, except that they also accept additional keyword arguments. radius indi-
cates the step size to use when searching in the direction of the gradient; values
between 1.2 and 1.5 are suggested (default 1.35). theta indicates the step size
to use when discretizing angles in the gradientangle image, in radians (default:
1 degree in radians = pi/180).
Example:

g = rgb2gray(rgb_image)
gx, gy = imgradients(g)
mag, grad_angle = magnitude_phase(gx,gy)
mag[mag .< 0.5] = 0.0 # Threshold magnitude image
thinned, subpix = thin_edges_subpix(mag, grad_angle)

source

5.55 Images.yen threshold

Images.yen threshold — Method.

thres = yen_threshold(img)
thres = yen_threshold(img, bins)

Computes threshold for grayscale image using Yen’s maximum correlation

criterion for bilevel thresholding
Parameters:

• img = Grayscale input image

• bins = Number of bins used to compute the histogram. Needed for
floating-point images.

5.56 Citation Yen J.C., Chang F.J., and Chang

S. (1995) “A New Criterion for Automatic
Multilevel Thresholding” IEEE Trans. on
Image Processing, 4(3): 370-378. DOI:10.1109/83.366472
source
Chapter 6

DataFrames

6.1 Base.insert!
Base.insert! — Method.
Insert a column into a data frame in place.
[] insert!(df::DataFrame, coli nd :: Int, item ::AbstractV ector, name ::Symbol;makeunique ::Bool= f alse)

Arguments

• df : the DataFrame to which we want to add a column

• col ind : a position at which we want to insert a column

• item : a column to be inserted into df

• name : column name

• makeunique : Defines what to do if name already exists in df; if it is false

an error will be thrown; if it is true a new unique name will be generated
by adding a suffix

Result

• ::DataFrame : a DataFrame with added column.

Examples

julia> d = DataFrame(a=1:3)
31 DataFrames.DataFrame
Row a

1 1
2 2
3 3

165
166 CHAPTER 6. DATAFRAMES

julia> insert!(d, 1, ’a’:’c’, :b)

32 DataFrames.DataFrame
Row b a

1 ’a’ 1
2 ’b’ 2
3 ’c’ 3

source

6.2 Base.vcat
Base.vcat — Method.

vcat(dfs::AbstractDataFrame...)

Vertically concatenate AbstractDataFrames.

Column names in all passed data frames must be the same, but they can have
different order. In such cases the order of names in the first passed DataFrame
is used.
Example

julia> df1 = DataFrame(A=1:3, B=1:3);

julia> df2 = DataFrame(A=4:6, B=4:6);
julia> vcat(df1, df2)
62 DataFrames.DataFrame
Row A B

1 1 1
2 2 2
3 3 3
4 4 4
5 5 5
6 6 6

julia> df1 = DataFrame(A=1:3, B=1:3);

julia> df2 = DataFrame(B=4:6, A=4:6);
julia> vcat(df1, df2)
62 DataFrames.DataFrame
Row A B

1 1 1
2 2 2
3 3 3
6.3. DATAFRAMES.AGGREGATE 167

4 4 4
5 5 5
6 6 6

source

6.3 DataFrames.aggregate
DataFrames.aggregate — Method.
Split-apply-combine that applies a set of functions over columns of an Ab-
stractDataFrame or GroupedDataFrame
[] aggregate(d::AbstractDataFrame, cols, fs) aggregate(gd::GroupedDataFrame,
fs)
Arguments

• d : an AbstractDataFrame

• gd : a GroupedDataFrame

• cols : a column indicator (Symbol, Int, Vector{Symbol}, etc.)

• fs : a function or vector of functions to be applied to vectors within

groups; expects each argument to be a column vector

Each fs should return a value or vector. All returns must be the same
length.
Returns

• ::DataFrame

Examples
[] df = DataFrame(a = repeat([1, 2, 3, 4], outer=[2]), b = repeat([2, 1],
outer=[4]), c = randn(8)) aggregate(df, :a, sum) aggregate(df, :a, [sum, x-
¿mean(skipmissing(x))]) aggregate(groupby(df, :a), [sum, x-¿mean(skipmissing(x))])

source

6.4 DataFrames.allowmissing!
DataFrames.allowmissing! — Function.

allowmissing!(df::DataFrame)

Convert all columns of a df from element type T to Union{T, Missing} to

support missing values.

allowmissing!(df::DataFrame, col::Union{Integer, Symbol})

168 CHAPTER 6. DATAFRAMES

Convert a single column of a df from element type T to Union{T, Missing}

to support missing values.

allowmissing!(df::DataFrame, cols::AbstractVector{<:Union{Integer, Symbol}})

Convert multiple columns of a df from element type T to Union{T, Missing}

to support missing values.
source

6.5 DataFrames.by
DataFrames.by — Method.
Split-apply-combine in one step; apply f to each grouping in d based on
columns col
[] by(d::AbstractDataFrame, cols, f::Function; sort::Bool = false) by(f::Function,
d::AbstractDataFrame, cols; sort::Bool = false)
Arguments

• d : an AbstractDataFrame

• cols : a column indicator (Symbol, Int, Vector{Symbol}, etc.)

• f : a function to be applied to groups; expects each argument to be an

AbstractDataFrame

• sort: sort row groups (no sorting by default)

f can return a value, a vector, or a DataFrame. For a value or vector, these

are merged into a column along with the cols keys. For a DataFrame, cols are
combined along columns with the resulting DataFrame. Returning a DataFrame
is the clearest because it allows column labeling.
A method is defined with f as the first argument, so do-block notation can
be used.
by(d, cols, f) is equivalent to combine(map(f, groupby(d, cols))).
Returns

• ::DataFrame

Examples
[] df = DataFrame(a = repeat([1, 2, 3, 4], outer=[2]), b = repeat([2,
1], outer=[4]), c = randn(8)) by(df, :a, d -¿ sum(d[:c])) by(df, :a, d -¿ 2 *
skipmissing(d[:c])) by(df, :a, d -¿ DataFrame(cs um = sum(d[: c]), cm ean =
mean(skipmissing(d[: c]))))by(df, : a, d− > DataF rame(c = d[: c], cm ean = mean(skipmissing(d[: c])))

source
6.6. DATAFRAMES.COLWISE 169

6.6 DataFrames.colwise
DataFrames.colwise — Method.
Apply a function to each column in an AbstractDataFrame or Grouped-
DataFrame
[] colwise(f::Function, d) colwise(d)
Arguments

• f : a function or vector of functions

• d : an AbstractDataFrame of GroupedDataFrame

If d is not provided, a curried version of groupby is given.

Returns

• various, depending on the call

Examples
[] df = DataFrame(a = repeat([1, 2, 3, 4], outer=[2]), b = repeat([2, 1],
outer=[4]), c = randn(8)) colwise(sum, df) colwise([sum, length], df) col-
wise((minimum, maximum), df) colwise(sum, groupby(df, :a))
source

6.7 DataFrames.combine
DataFrames.combine — Method.
Combine a GroupApplied object (rudimentary)
[] combine(ga::GroupApplied)
Arguments

• ga : a GroupApplied

Returns

• ::DataFrame

Examples
[] df = DataFrame(a = repeat([1, 2, 3, 4], outer=[2]), b = repeat([2, 1],
outer=[4]), c = randn(8)) gd = groupby(df, :a) combine(map(d -¿ mean(skipmissing(d[:c])),
gd))
source
170 CHAPTER 6. DATAFRAMES

6.8 DataFrames.completecases
DataFrames.completecases — Method.
Indexes of complete cases (rows without missing values)
[] completecases(df::AbstractDataFrame)
Arguments

• df : the AbstractDataFrame

Result

• ::Vector{Bool} : indexes of complete cases

See also dropmissing and dropmissing!.

Examples
[] df = DataFrame(i = 1:10, x = Vector{Union{Missing, Float64}}(rand(10)),
y = Vector{Union{Missing, String}}(rand([”a”, ”b”, ”c”], 10))) df[[1,4,5], :x] =
missing df[[9,10], :y] = missing completecases(df)
source

6.9 DataFrames.disallowmissing!
DataFrames.disallowmissing! — Function.

disallowmissing!(df::DataFrame)

Convert all columns of a df from element type Union{T, Missing} to T to

drop support for missing values.

disallowmissing!(df::DataFrame, col::Union{Integer, Symbol})

Convert a single column of a df from element type Union{T, Missing} to

T to drop support for missing values.

disallowmissing!(df::DataFrame, cols::AbstractVector{<:Union{Integer, Symbol}})

Convert multiple columns of a df from element type Union{T, Missing} to

T to drop support for missing values.
source

6.10 DataFrames.dropmissing!
DataFrames.dropmissing! — Method.
Remove rows with missing values in-place.
[] dropmissing!(df::AbstractDataFrame)
Arguments
6.11. DATAFRAMES.DROPMISSING 171

• df : the AbstractDataFrame

Result

• ::AbstractDataFrame : the updated version

See also dropmissing and completecases.

Examples
[] df = DataFrame(i = 1:10, x = Vector{Union{Missing, Float64}}(rand(10)),
y = Vector{Union{Missing, String}}(rand([”a”, ”b”, ”c”], 10))) df[[1,4,5], :x] =
missing df[[9,10], :y] = missing dropmissing!(df)
source

6.11 DataFrames.dropmissing
DataFrames.dropmissing — Method.
Remove rows with missing values.
[] dropmissing(df::AbstractDataFrame)
Arguments

• df : the AbstractDataFrame

Result

• ::AbstractDataFrame : the updated copy

See also completecases and dropmissing!.

Examples
[] df = DataFrame(i = 1:10, x = Vector{Union{Missing, Float64}}(rand(10)),
y = Vector{Union{Missing, String}}(rand([”a”, ”b”, ”c”], 10))) df[[1,4,5], :x] =
missing df[[9,10], :y] = missing dropmissing(df)
source

6.12 DataFrames.eachrow
DataFrames.eachrow — Method.

eachrow(df) => DataFrames.DFRowIterator

Iterate a DataFrame row by row, with each row represented as a DataFrameRow,

which is a view that acts like a one-row DataFrame.
source
172 CHAPTER 6. DATAFRAMES

6.13 DataFrames.eltypes
DataFrames.eltypes — Method.
Return element types of columns
[] eltypes(df::AbstractDataFrame)
Arguments

• df : the AbstractDataFrame

Result

• ::Vector{Type} : the element type of each column

Examples
[] df = DataFrame(i = 1:10, x = rand(10), y = rand([”a”, ”b”, ”c”], 10))
eltypes(df)
source

6.14 DataFrames.groupby
DataFrames.groupby — Method.
A view of an AbstractDataFrame split into row groups
[] groupby(d::AbstractDataFrame, cols; sort = false, skipmissing = false)
groupby(cols; sort = false, skipmissing = false)
Arguments

• d : an AbstractDataFrame to split (optional, see Returns)

• cols : data table columns to group by
• sort: whether to sort rows according to the values of the grouping columns
cols
• skipmissing: whether to skip rows with missing values in one of the
grouping columns cols

Returns

• ::GroupedDataFrame : a grouped view into d

• ::Function: a function x -> groupby(x, cols) (if d is not specified)

Details
An iterator over a GroupedDataFrame returns a SubDataFrame view for each
grouping into d. A GroupedDataFrame also supports indexing by groups and
map.
See the following for additional split-apply-combine operations:

• by : split-apply-combine using functions

6.15. DATAFRAMES.HEAD 173

• aggregate : split-apply-combine; applies functions in the form of a cross

product
• combine : combine (obviously)
• colwise : apply a function to each column in an AbstractDataFrame or
GroupedDataFrame

Examples
[] df = DataFrame(a = repeat([1, 2, 3, 4], outer=[2]), b = repeat([2, 1],
outer=[4]), c = randn(8)) gd = groupby(df, :a) gd[1] last(gd) vcat([g[:b] for g in
gd]...) for g in gd println(g) end map(d -¿ mean(skipmissing(d[:c])), gd) returns
a GroupApplied object combine(map(d -¿ mean(skipmissing(d[:c])), gd))
source

6.15 DataFrames.head
DataFrames.head — Function.
Show the first or last part of an AbstractDataFrame
[] head(df::AbstractDataFrame, r::Int = 6) tail(df::AbstractDataFrame, r::Int
= 6)
Arguments

• df : the AbstractDataFrame
• r : the number of rows to show

Result

• ::AbstractDataFrame : the first or last part of df

Examples
[] df = DataFrame(i = 1:10, x = rand(10), y = rand([”a”, ”b”, ”c”], 10))
head(df) tail(df)
source

6.16 DataFrames.melt
DataFrames.melt — Method.
Stacks a DataFrame; convert from a wide to long format; see stack.
source

6.17 DataFrames.meltdf
DataFrames.meltdf — Method.
A stacked view of a DataFrame (long format); see stackdf
source
174 CHAPTER 6. DATAFRAMES

6.18 DataFrames.names!
DataFrames.names! — Method.
Set column names
[] names!(df::AbstractDataFrame, vals)
Arguments

• df : the AbstractDataFrame
• vals : column names, normally a Vector{Symbol} the same length as the
number of columns in df
• makeunique : if false (the default), an error will be raised if duplicate
names are found; if true, duplicate names will be suffixed with i (i
starting at 1 for the first duplicate).

Result

• ::AbstractDataFrame : the updated result

Examples
[] df = DataFrame(i = 1:10, x = rand(10), y = rand([”a”, ”b”, ”c”], 10))
names!(df, [:a, :b, :c]) names!(df, [:a, :b, :a]) throws ArgumentError names!(df,
[:a, :b, :a], makeunique=true) renames second :a to :a1
source

6.19 DataFrames.nonunique
DataFrames.nonunique — Method.
Indexes of duplicate rows (a row that is a duplicate of a prior row)
[] nonunique(df::AbstractDataFrame) nonunique(df::AbstractDataFrame, cols)

Arguments

• df : the AbstractDataFrame
• cols : a column indicator (Symbol, Int, Vector{Symbol}, etc.) specifying
the column(s) to compare

Result

• ::Vector{Bool} : indicates whether the row is a duplicate of some prior

row

See also unique and unique!.

Examples
[] df = DataFrame(i = 1:10, x = rand(10), y = rand([”a”, ”b”, ”c”], 10)) df
= vcat(df, df) nonunique(df) nonunique(df, 1)
source
6.20. DATAFRAMES.READTABLE 175

6.20 DataFrames.readtable
DataFrames.readtable — Method.
Read data from a tabular-file format (CSV, TSV, . . . )
[] readtable(filename, [keyword options])
Arguments

• filename::AbstractString : the filename to be read

Keyword Arguments

• header::Bool – Use the information from the file’s header line to deter-
mine column names. Defaults to true.

• separator::Char – Assume that fields are split by the separator char-

acter. If not specified, it will be guessed from the filename: .csv defaults
to ,, .tsv defaults to , .wsv defaults to .

• quotemark::Vector{Char} – Assume that fields contained inside of two

quotemark characters are quoted, which disables processing of separators
and linebreaks. Set to Char[] to disable this feature and slightly improve
performance. Defaults to ["].

• decimal::Char – Assume that the decimal place in numbers is written

using the decimal character. Defaults to ..

• nastrings::Vector{String} – Translate any of the strings into this vec-

tor into a missing. Defaults to ["", "NA"].

• truestrings::Vector{String} – Translate any of the strings into this

vector into a Boolean true. Defaults to ["T", "t", "TRUE", "true"].

• falsestrings::Vector{String} – Translate any of the strings into this

vector into a Boolean false. Defaults to ["F", "f", "FALSE", "false"].

• makefactors::Bool – Convert string columns into PooledDataVector’s

for use as factors. Defaults to false.

• nrows::Int – Read only nrows from the file. Defaults to -1, which indi-
cates that the entire file should be read.

• names::Vector{Symbol} – Use the values in this array as the names for

all columns instead of or in lieu of the names in the file’s header. Defaults
to [], which indicates that the header should be used if present or that
numeric names should be invented if there is no header.

• eltypes::Vector – Specify the types of all columns. Defaults to [].

• allowcomments::Bool – Ignore all text inside comments. Defaults to

false.
176 CHAPTER 6. DATAFRAMES

• commentmark::Char – Specify the character that starts comments. De-

faults to #.
• ignorepadding::Bool – Ignore all whitespace on left and right sides of a
field. Defaults to true.
• skipstart::Int – Specify the number of initial rows to skip. Defaults to
0.
• skiprows::Vector{Int} – Specify the indices of lines in the input to
ignore. Defaults to [].
• skipblanks::Bool – Skip any blank lines in input. Defaults to true.
• encoding::Symbol – Specify the file’s encoding as either :utf8 or :latin1.
Defaults to :utf8.
• normalizenames::Bool – Ensure that column names are valid Julia iden-
tifiers. For instance this renames a column named "a b" to "a b" which
can then be accessed with :a b instead of Symbol("a b"). Defaults to
true.

Result

• ::DataFrame

Examples
[] df = readtable(”data.csv”) df = readtable(”data.tsv”) df = readtable(”data.wsv”)
df = readtable(”data.txt”, separator = ’ ’) df = readtable(”data.txt”, header =
false)
source

6.21 DataFrames.rename
DataFrames.rename — Function.
Rename columns
[] rename!(df::AbstractDataFrame, (from =¿ to)::Pair{Symbol, Symbol}...)
rename!(df::AbstractDataFrame, d::AbstractDict{Symbol,Symbol}) rename!(df::AbstractDataFrame,
d::AbstractArray{Pair{Symbol,Symbol}}) rename!(f::Function, df::AbstractDataFrame)
rename(df::AbstractDataFrame, (from =¿ to)::Pair{Symbol, Symbol}...) re-
name(df::AbstractDataFrame, d::AbstractDict{Symbol,Symbol}) rename(df::AbstractDataFrame,
d::AbstractArray{Pair{Symbol,Symbol}}) rename(f::Function, df::AbstractDataFrame)

Arguments

• df : the AbstractDataFrame
• d : an Associative type or an AbstractArray of pairs that maps the original
names to new names
6.22. DATAFRAMES.RENAME! 177

• f : a function which for each column takes the old name (a Symbol) and
returns the new name (a Symbol)
Result
• ::AbstractDataFrame : the updated result
New names are processed sequentially. A new name must not already exist
in the DataFrame at the moment an attempt to rename a column is performed.
Examples
[] df = DataFrame(i = 1:10, x = rand(10), y = rand([”a”, ”b”, ”c”], 10))
rename(df, :i =¿ :A, :x =¿ :X) rename(df, [:i =¿ :A, :x =¿ :X]) rename(df, Dict(:i
=¿ :A, :x =¿ :X)) rename(x -¿ Symbol(uppercase(string(x))), df) rename!(df,
Dict(:i =¿: A, :x =¿ :X))
source

6.22 DataFrames.rename!
DataFrames.rename! — Function.
Rename columns
[] rename!(df::AbstractDataFrame, (from =¿ to)::Pair{Symbol, Symbol}...)
rename!(df::AbstractDataFrame, d::AbstractDict{Symbol,Symbol}) rename!(df::AbstractDataFrame,
d::AbstractArray{Pair{Symbol,Symbol}}) rename!(f::Function, df::AbstractDataFrame)
rename(df::AbstractDataFrame, (from =¿ to)::Pair{Symbol, Symbol}...) re-
name(df::AbstractDataFrame, d::AbstractDict{Symbol,Symbol}) rename(df::AbstractDataFrame,
d::AbstractArray{Pair{Symbol,Symbol}}) rename(f::Function, df::AbstractDataFrame)

Arguments
• df : the AbstractDataFrame
• d : an Associative type or an AbstractArray of pairs that maps the original
names to new names
• f : a function which for each column takes the old name (a Symbol) and
returns the new name (a Symbol)
Result
• ::AbstractDataFrame : the updated result
New names are processed sequentially. A new name must not already exist
in the DataFrame at the moment an attempt to rename a column is performed.
Examples
[] df = DataFrame(i = 1:10, x = rand(10), y = rand([”a”, ”b”, ”c”], 10))
rename(df, :i =¿ :A, :x =¿ :X) rename(df, [:i =¿ :A, :x =¿ :X]) rename(df, Dict(:i
=¿ :A, :x =¿ :X)) rename(x -¿ Symbol(uppercase(string(x))), df) rename!(df,
Dict(:i =¿: A, :x =¿ :X))
source
178 CHAPTER 6. DATAFRAMES

6.23 DataFrames.stack
DataFrames.stack — Method.
Stacks a DataFrame; convert from a wide to long format
[] stack(df::AbstractDataFrame, [measurev ars], [idv ars]; variablen ame ::Symbol=: variable, valuen am

Arguments

• df : the AbstractDataFrame to be stacked

• measure vars : the columns to be stacked (the measurement variables),

a normal column indexing type, like a Symbol, Vector{Symbol}, Int,
etc.; for melt, defaults to all variables that are not id vars. If neither
measure vars or id vars are given, measure vars defaults to all floating
point columns.

• id vars : the identifier columns that are repeated during stacking, a

normal column indexing type; for stack defaults to all variables that are
not measure vars

• variable name : the name of the new stacked column that shall hold the
names of each of measure vars

• value name : the name of the new stacked column containing the values
from each of measure vars

Result

• ::DataFrame : the long-format DataFrame with column :value holding

the values of the stacked columns (measure vars), with column :variable
a Vector of Symbols with the measure vars name, and with columns for
each of the id vars.

See also stackdf and meltdf for stacking methods that return a view into
the original DataFrame. See unstack for converting from long to wide format.
Examples
[] d1 = DataFrame(a = repeat([1:3;], inner = [4]), b = repeat([1:4;], inner
= [3]), c = randn(12), d = randn(12), e = map(string, ’a’:’l’))
d1s = stack(d1, [:c, :d]) d1s2 = stack(d1, [:c, :d], [:a]) d1m = melt(d1, [:a,
:b, :e]) d1sn ame = melt(d1, [: a, : b, : e], variablen ame =: somemeasure)
source

6.24 DataFrames.stackdf
DataFrames.stackdf — Method.
A stacked view of a DataFrame (long format)
Like stack and melt, but a view is returned rather than data copies.
6.25. DATAFRAMES.TAIL 179

[] stackdf(df::AbstractDataFrame, [measurev ars], [idv ars]; variablen ame ::Symbol=: variable, valuen ame ::Symbo

Arguments
• df : the wide AbstractDataFrame
• measure vars : the columns to be stacked (the measurement variables),
a normal column indexing type, like a Symbol, Vector{Symbol}, Int, etc.;
for melt, defaults to all variables that are not id vars
• id vars : the identifier columns that are repeated during stacking, a
normal column indexing type; for stack defaults to all variables that are
not measure vars
Result
• ::DataFrame : the long-format DataFrame with column :value holding
the values of the stacked columns (measure vars), with column :variable
a Vector of Symbols with the measure vars name, and with columns for
each of the id vars.
The result is a view because the columns are special AbstractVectors that
return indexed views into the original DataFrame.
Examples
[] d1 = DataFrame(a = repeat([1:3;], inner = [4]), b = repeat([1:4;], inner
= [3]), c = randn(12), d = randn(12), e = map(string, ’a’:’l’))
d1s = stackdf(d1, [:c, :d]) d1s2 = stackdf(d1, [:c, :d], [:a]) d1m = meltdf(d1,
[:a, :b, :e])
source

6.25 DataFrames.tail
DataFrames.tail — Function.
Show the first or last part of an AbstractDataFrame
[] head(df::AbstractDataFrame, r::Int = 6) tail(df::AbstractDataFrame, r::Int
= 6)
Arguments
• df : the AbstractDataFrame
• r : the number of rows to show
Result
• ::AbstractDataFrame : the first or last part of df
Examples
[] df = DataFrame(i = 1:10, x = rand(10), y = rand([”a”, ”b”, ”c”], 10))
head(df) tail(df)
source
180 CHAPTER 6. DATAFRAMES

6.26 DataFrames.unique!
DataFrames.unique! — Function.
Delete duplicate rows
[] unique(df::AbstractDataFrame) unique(df::AbstractDataFrame, cols) unique!(df::AbstractDataFram
unique!(df::AbstractDataFrame, cols)
Arguments

• df : the AbstractDataFrame
• cols : column indicator (Symbol, Int, Vector{Symbol}, etc.)

specifying the column(s) to compare.

Result

• ::AbstractDataFrame : the updated version of df with unique rows.

When cols is specified, the return DataFrame contains complete rows, re-
taining in each case the first instance for which df[cols] is unique.
See also nonunique.
Examples
[] df = DataFrame(i = 1:10, x = rand(10), y = rand([”a”, ”b”, ”c”], 10)) df
= vcat(df, df) unique(df) doesn’t modify df unique(df, 1) unique!(df) modifies
df
source

6.27 DataFrames.unstack
DataFrames.unstack — Method.
Unstacks a DataFrame; convert from a long to wide format
[] unstack(df::AbstractDataFrame, rowkeys::Union{Symbol, Integer}, colkey::Union{Symbol,
Integer}, value::Union{Symbol, Integer}) unstack(df::AbstractDataFrame, rowkeys::AbstractVector{¡:Un
Integer}}, colkey::Union{Symbol, Integer}, value::Union{Symbol, Integer}) un-
stack(df::AbstractDataFrame, colkey::Union{Symbol, Integer}, value::Union{Symbol,
Integer}) unstack(df::AbstractDataFrame)
Arguments

• df : the AbstractDataFrame to be unstacked

• rowkeys : the column(s) with a unique key for each row, if not given, find
a key by grouping on anything not a colkey or value
• colkey : the column holding the column names in wide format, defaults
to :variable
• value : the value column, defaults to :value

Result
6.28. DATAFRAMES.WRITETABLE 181

• ::DataFrame : the wide-format DataFrame

If colkey contains missing values then they will be skipped and a warning
will be printed.
If combination of rowkeys and colkey contains duplicate entries then last
value will be retained and a warning will be printed.
Examples
[] wide = DataFrame(id = 1:12, a = repeat([1:3;], inner = [4]), b = re-
peat([1:4;], inner = [3]), c = randn(12), d = randn(12))
long = stack(wide) wide0 = unstack(long) wide1 = unstack(long, :variable,
:value) wide2 = unstack(long, :id, :variable, :value) wide3 = unstack(long, [:id,
:a], :variable, :value)
Note that there are some differences between the widened results above.
source

6.28 DataFrames.writetable
DataFrames.writetable — Method.
Write data to a tabular-file format (CSV, TSV, . . . )
[] writetable(filename, df, [keyword options])
Arguments
• filename::AbstractString : the filename to be created
• df::AbstractDataFrame : the AbstractDataFrame to be written
Keyword Arguments
• separator::Char – The separator character that you would like to use.
Defaults to the output of getseparator(filename), which uses commas
for files that end in .csv, tabs for files that end in .tsv and a single space
for files that end in .wsv.
• quotemark::Char – The character used to delimit string fields. Defaults
to ".
• header::Bool – Should the file contain a header that specifies the column
names from df. Defaults to true.
• nastring::AbstractString – What to write in place of missing data.
Defaults to "NA".
Result
• ::DataFrame
Examples
[] df = DataFrame(A = 1:10) writetable(”output.csv”, df) writetable(”output.dat”,
df, separator = ’,’, header = false) writetable(”output.dat”, df, quotemark =
”’, separator = ’,’) writetable(”output.dat”, df, header = false)
source
182 CHAPTER 6. DATAFRAMES

6.29 StatsBase.describe
StatsBase.describe — Method.
Summarize the columns of an AbstractDataFrame
[] describe(df::AbstractDataFrame) describe(io, df::AbstractDataFrame)
Arguments

• df : the AbstractDataFrame
• io : optional output descriptor

Result

• nothing

Details
If the column’s base type derives from Number, compute the minimum, first
quantile, median, mean, third quantile, and maximum. Missings are filtered
and reported separately.
For boolean columns, report trues, falses, and missings.
For other types, show column characteristics and number of missings.
Examples
[] df = DataFrame(i = 1:10, x = rand(10), y = rand([”a”, ”b”, ”c”], 10))
describe(df)
source
Chapter 7

Knet

7.1 Knet.accuracy
Knet.accuracy — Method.

accuracy(model, data, predict; average=true)

Compute accuracy(predict(model,x), y) for (x,y) in data and return

the ratio (if average=true) or the count (if average=false) of correct answers.
source

7.2 Knet.accuracy
Knet.accuracy — Method.

accuracy(scores, answers, d=1; average=true)

Given an unnormalized scores matrix and an Integer array of correct

answers, return the ratio of instances where the correct answer has the max-
imum score. d=1 means instances are in columns, d=2 means instances are in
rows. Use average=false to return the number of correct answers instead of
the ratio.
source

7.3 Knet.batchnorm
Knet.batchnorm — Function.
batchnorm(x[, moments, params]; kwargs...) performs batch normal-
ization to x with optional scaling factor and bias stored in params.
2d, 4d and 5d inputs are supported. Mean and variance are computed over
dimensions (2,), (1,2,4) and (1,2,3,5) for 2d, 4d and 5d arrays, respectively.

183
184 CHAPTER 7. KNET

moments stores running mean and variance to be used in testing. It is op-

tional in the training mode, but mandatory in the test mode. Training and test
modes are controlled by the training keyword argument.
params stores the optional affine parameters gamma and beta. bnparams
function can be used to initialize params.
Example

# Inilization, C is an integer
moments = bnmoments()
params = bnparams(C)
...
# size(x) -> (H, W, C, N)
y = batchnorm(x, moments, params)
# size(y) -> (H, W, C, N)

Keywords
eps=1e-5: The epsilon parameter added to the variance to avoid division by
0.
training: When training is true, the mean and variance of x are used and
moments argument is modified if it is provided. When training is false, mean
and variance stored in the moments argument are used. Default value is true
when at least one of x and params is AutoGrad.Rec, false otherwise.
source

7.4 Knet.bilinear
Knet.bilinear — Method.
Bilinear interpolation filter weights; used for initializing deconvolution layers.
Adapted from https://github.com/shelhamer/fcn.berkeleyvision.org/blob/master/surgery.py#L33
Arguments:
T : Data Type
fw: Width upscale factor
fh: Height upscale factor
IN: Number of input filters
ON: Number of output filters
Example usage:
w = bilinear(Float32,2,2,128,128)
source

7.5 Knet.bnmoments
Knet.bnmoments — Method.
bnmoments(;momentum=0.1, mean=nothing, var=nothing, meaninit=zeros,
varinit=ones) can be used directly load moments from data. meaninit and
varinit are called if mean and var are nothing. Type and size of the mean and
7.6. KNET.BNPARAMS 185

var are determined automatically from the inputs in the batchnorm calls. A
BNMoments object is returned.
BNMoments
A high-level data structure used to store running mean and running variance
of batch normalization with the following fields:
momentum::AbstractFloat: A real number between 0 and 1 to be used as
the scale of last mean and variance. The existing running mean or variance is
multiplied by (1-momentum).
mean: The running mean.
var: The running variance.
meaninit: The function used for initialize the running mean. Should either
be nothing or of the form (eltype, dims...)->data. zeros is a good option.
varinit: The function used for initialize the running variance. Should either
be nothing or (eltype, dims...)->data. ones is a good option.
source

7.6 Knet.bnparams
Knet.bnparams — Method.
bnparams(etype, channels) creates a single 1d array that contains both
scale and bias of batchnorm, where the first half is scale and the second half is
bias.
bnparams(channels) calls bnparams with etype=Float64, following Julia
convention
source

7.7 Knet.conv4
Knet.conv4 — Method.

conv4(w, x; kwargs...)

Execute convolutions or cross-correlations using filters specified with w over

tensor x.
Currently KnetArray{Float32/64,4/5} and Array{Float32/64,4} are sup-
ported as w and x. If w has dimensions (W1,W2,...,I,O) and x has dimensions
(X1,X2,...,I,N), the result y will have dimensions (Y1,Y2,...,O,N) where

Yi=1+floor((Xi+2*padding[i]-Wi)/stride[i])

Here I is the number of input channels, O is the number of output channels,

N is the number of instances, and Wi,Xi,Yi are spatial dimensions. padding
and stride are keyword arguments that can be specified as a single number
(in which case they apply to all dimensions), or an array/tuple with entries for
each spatial dimension.
Keywords
186 CHAPTER 7. KNET

• padding=0: the number of extra zeros implicitly concatenated at the start

and at the end of each dimension.
• stride=1: the number of elements to slide to reach the next filtering
window.
• upscale=1: upscale factor for each dimension.
• mode=0: 0 for convolution and 1 for cross-correlation.
• alpha=1: can be used to scale the result.
• handle: handle to a previously created cuDNN context. Defaults to a
Knet allocated handle.

source

7.8 Knet.deconv4
Knet.deconv4 — Method.

y = deconv4(w, x; kwargs...)

Simulate 4-D deconvolution by using transposed convolution operation. Its

forward pass is equivalent to backward pass of a convolution (gradients with
respect to input tensor). Likewise, its backward pass (gradients with respect to
input tensor) is equivalent to forward pass of a convolution. Since it swaps for-
ward and backward passes of convolution operation, padding and stride options
belong to output tensor. See this report for further explanation.
Currently KnetArray{Float32/64,4} and Array{Float32/64,4} are supported
as w and x. If w has dimensions (W1,W2,...,O,I) and x has dimensions
(X1,X2,...,I,N), the result y will have dimensions (Y1,Y2,...,O,N) where
Yi = Wi+stride[i](Xi-1)-2 padding[i]
Here I is the number of input channels, O is the number of output channels,
N is the number of instances, and Wi,Xi,Yi are spatial dimensions. padding
and stride are keyword arguments that can be specified as a single number (in
which case they apply to all dimensions), or an array/tuple with entries for each
spatial dimension.
Keywords

• padding=0: the number of extra zeros implicitly concatenated at the start

and at the end of each dimension.
• stride=1: the number of elements to slide to reach the next filtering
window.
• mode=0: 0 for convolution and 1 for cross-correlation.
• alpha=1: can be used to scale the result.
7.9. KNET.DROPOUT 187

• handle: handle to a previously created cuDNN context. Defaults to a

Knet allocated handle.
source

7.9 Knet.dropout
Knet.dropout — Method.
dropout(x, p)
Given an array x and probability 0<=p<=1, just return x if p==0, or re-
turn an array y in which each element is 0 with probability p or x[i]/(1-p)
with probability 1-p. Use seed::Number to set the random number seed for
reproducible results. See (Srivastava et al. 2014) for a reference.
source

7.10 Knet.gaussian
Knet.gaussian — Method.
gaussian(a...; mean=0.0, std=0.01)
Return a Gaussian array with a given mean and standard deviation. The a
arguments are passed to randn.
source

7.11 Knet.goldensection
Knet.goldensection — Method.
goldensection(f,n;kwargs) => (fmin,xmin)
Find the minimum of f using concurrent golden section search in n dimen-
sions. See Knet.goldensection demo() for an example.
f is a function from a Vector{Float64} of length n to a Number. It can
return NaN for out of range inputs. Goldensection will always start with a
zero vector as the initial input to f, and the initial step size will be 1 in each
dimension. The user should define f to scale and shift this input range into a
vector meaningful for their application. For positive inputs like learning rate or
hidden size, you can use a transformation such as x0*exp(x) where x is a value
goldensection passes to f and x0 is your initial guess for this value. This will
effectively start the search at x0, then move with multiplicative steps.
I designed this algorithm combining ideas from Golden Section Search and
Hill Climbing Search. It essentially runs golden section search concurrently in
each dimension, picking the next step based on estimated gain.
Keyword arguments
188 CHAPTER 7. KNET

• dxmin=0.1: smallest step size.

• accel=: acceleration rate. Golden ratio =1.618... is best.
• verbose=false: use true to print individual steps.
• history=[]: cache of [(x,f(x)),...] function evaluations.
source

7.12 Knet.gpu
Knet.gpu — Function.
gpu() returns the id of the active GPU device or -1 if none are active.
gpu(true) resets all GPU devices and activates the one with the most avail-
able memory.
gpu(false) resets and deactivates all GPU devices.
gpu(d::Int) activates the GPU device d if 0 <= d < gpuCount(), other-
wise deactivates devices.
gpu(true/false) resets all devices. If there are any allocated KnetArrays
their pointers will be left dangling. Thus gpu(true/false) should only be used
during startup. If you want to suspend GPU use temporarily, use gpu(-1).
gpu(d::Int) does not reset the devices. You can select a previous device
and find allocated memory preserved. However trying to operate on arrays of
an inactive device will result in error.
source

7.13 Knet.hyperband
Knet.hyperband — Function.
hyperband(getconfig, getloss, maxresource=27, reduction=3)
Hyperparameter optimization using the hyperband algorithm from (Lisha et
al. 2016). You can try a simple MNIST example using Knet.hyperband demo().
Arguments
• getconfig() returns random configurations with a user defined type and
distribution.
• getloss(c,n) returns loss for configuration c and number of resources
(e.g. epochs) n.
• maxresource is the maximum number of resources any one configuration
should be given.
• reduction is an algorithm parameter (see paper), 3 is a good value.
source
7.14. KNET.INVX 189

7.14 Knet.invx
Knet.invx — Function.
invx(x) = (1./x)
source

7.15 Knet.knetgc
Knet.knetgc — Function.

knetgc(dev=gpu())

cudaFree all pointers allocated on device dev that were previously allocated
and garbage collected. Normally Knet holds on to all garbage collected pointers
for reuse. Try this if you run out of GPU memory.
source

7.16 Knet.logp
Knet.logp — Method.

logp(x,[dims])

Treat entries in x as as unnormalized log probabilities and return normalized

log probabilities.
dims is an optional argument, if not specified the normalization is over the
whole x, otherwise the normalization is performed over the given dimensions.
In particular, if x is a matrix, dims=1 normalizes columns of x and dims=2
normalizes rows of x.
source

7.17 Knet.logsumexp
Knet.logsumexp — Method.

logsumexp(x,[dims])

Compute log(sum(exp(x),dims)) in a numerically stable manner.

dims is an optional argument, if not specified the summation is over the
whole x, otherwise the summation is performed over the given dimensions. In
particular if x is a matrix, dims=1 sums columns of x and dims=2 sums rows of
x.
source
190 CHAPTER 7. KNET

7.18 Knet.mat
Knet.mat — Method.
mat(x)
Reshape x into a two-dimensional matrix.
This is typically used when turning the output of a 4-D convolution re-
sult into a 2-D input for a fully connected layer. For 1-D inputs returns
reshape(x, (length(x),1)). For inputs with more than two dimensions of
size (X1,X2,...,XD), returns
reshape(x, (X1*X2*...*X[D-1],XD))
source

7.19 Knet.minibatch
Knet.minibatch — Method.
minibatch(x, y, batchsize; shuffle, partial, xtype, ytype)
Return an iterable of minibatches [(xi,yi). . . ] given data tensors x, y and
batchsize. The last dimension of x and y should match and give the number of
instances. Keyword arguments:
• shuffle=false: Shuffle the instances before minibatching.
• partial=false: If true include the last partial minibatch < batchsize.
• xtype=typeof(x): Convert xi in minibatches to this type.
• ytype=typeof(y): Convert yi in minibatches to this type.
source

7.20 Knet.minibatch
Knet.minibatch — Method.
minibatch(x, batchsize; shuffle, partial, xtype, ytype)
Return an iterable of minibatches [x1,x2,. . . ] given data tensor x and batch-
size. The last dimension of x gives the number of instances. Keyword arguments:
• shuffle=false: Shuffle the instances before minibatching.
• partial=false: If true include the last partial minibatch < batchsize.
• xtype=typeof(x): Convert xi in minibatches to this type.
source
7.21. KNET.NLL 191

7.21 Knet.nll
Knet.nll — Method.
nll(model, data, predict; average=true)
Compute nll(predict(model,x), y) for (x,y) in data and return the
per-instance average (if average=true) or total (if average=false) negative log
likelihood.
source

7.22 Knet.nll
Knet.nll — Method.
nll(scores, answers, d=1; average=true)
Given an unnormalized scores matrix and an Integer array of correct
answers, return the per-instance negative log likelihood. d=1 means instances
are in columns, d=2 means instances are in rows. Use average=false to return
the sum instead of per-instance average.
source

7.23 Knet.optimizers
Knet.optimizers — Method.
optimizers(model, otype; options...)
Given parameters of a model, initialize and return corresponding optimiza-
tion parameters for a given optimization type otype and optimization options
options. This is useful because each numeric array in model needs its own dis-
tinct optimization parameter. optimizers makes the creation of optimization
parameters that parallel model parameters easy when all of them use the same
type and options.
source

7.24 Knet.pool
Knet.pool — Method.
pool(x; kwargs...)
Compute pooling of input values (i.e., the maximum or average of several
adjacent values) to produce an output with smaller height and/or width.
Currently 4 or 5 dimensional KnetArrays with Float32 or Float64 entries
are supported. If x has dimensions (X1,X2,...,I,N), the result y will have
dimensions (Y1,Y2,...,I,N) where
192 CHAPTER 7. KNET

Yi=1+floor((Xi+2*padding[i]-window[i])/stride[i])

Here I is the number of input channels, N is the number of instances, and

Xi,Yi are spatial dimensions. window, padding and stride are keyword argu-
ments that can be specified as a single number (in which case they apply to all
dimensions), or an array/tuple with entries for each spatial dimension.
Keywords:

• window=2: the pooling window size for each dimension.

• padding=0: the number of extra zeros implicitly concatenated at the start
and at the end of each dimension.
• stride=window: the number of elements to slide to reach the next pooling
window.
• mode=0: 0 for max, 1 for average including padded values, 2 for average
excluding padded values.
• maxpoolingNanOpt=0: Nan numbers are not propagated if 0, they are
propagated if 1.
• alpha=1: can be used to scale the result.
• handle: Handle to a previously created cuDNN context. Defaults to a
Knet allocated handle.

source

7.25 Knet.relu
Knet.relu — Function.
relu(x) = max(0,x)
source

7.26 Knet.rnnforw
Knet.rnnforw — Method.

rnnforw(r, w, x[, hx, cx]; batchSizes, hy, cy)

Returns a tuple (y,hyout,cyout,rs) given rnn r, weights w, input x and op-

tionally the initial hidden and cell states hx and cx (cx is only used in LSTMs).
r and w should come from a previous call to rnninit. Both hx and cx are
optional, they are treated as zero arrays if not provided. The output y contains
the hidden states of the final layer for each time step, hyout and cyout give the
final hidden and cell states for all layers, rs is a buffer the RNN needs for its
gradient calculation.
7.27. KNET.RNNINIT 193

The boolean keyword arguments hy and cy control whether hyout and cyout
will be output. By default hy = (hx!=nothing) and cy = (cx!=nothing &&
r.mode==2), i.e. a hidden state will be output if one is provided as input and
for cell state we also require an LSTM. If hy/cy is false, hyout/cyout will be
nothing. batchSizes can be an integer array that specifies non-uniform batch
sizes as explained below. By default batchSizes=nothing and the same batch
size, size(x,2), is used for all time steps.
The input and output dimensions are:

• x: (X,[B,T])
• y: (H/2H,[B,T])
• hx,cx,hyout,cyout: (H,B,L/2L)
• batchSizes: nothing or Vector{Int}(T)

where X is inputSize, H is hiddenSize, B is batchSize, T is seqLength, L is

numLayers. x can be 1, 2, or 3 dimensional. If batchSizes==nothing, a 1-D x
represents a single instance, a 2-D x represents a single minibatch, and a 3-D x
represents a sequence of identically sized minibatches. If batchSizes is an array
of (non-increasing) integers, it gives us the batch size for each time step in the se-
quence, in which case sum(batchSizes) should equal div(length(x),size(x,1)).
y has the same dimensionality as x, differing only in its first dimension, which
is H if the RNN is unidirectional, 2H if bidirectional. Hidden vectors hx, cx,
hyout, cyout all have size (H,B1,L) for unidirectional RNNs, and (H,B1,2L) for
bidirectional RNNs where B1 is the size of the first minibatch.
source

7.27 Knet.rnninit
Knet.rnninit — Method.

rnninit(inputSize, hiddenSize; opts...)

Return an (r,w) pair where r is a RNN struct and w is a single weight array
that includes all matrices and biases for the RNN. Keyword arguments:

• rnnType=:lstm Type of RNN: One of :relu, :tanh, :lstm, :gru.

• numLayers=1: Number of RNN layers.
• bidirectional=false: Create a bidirectional RNN if true.
• dropout=0.0: Dropout probability. Ignored if numLayers==1.
• skipInput=false: Do not multiply the input with a matrix if true.
• dataType=Float32: Data type to use for weights.
194 CHAPTER 7. KNET

• algo=0: Algorithm to use, see CUDNN docs for details.

• seed=0: Random number seed. Uses time() if 0.
• winit=xavier: Weight initialization method for matrices.
• binit=zeros: Weight initialization method for bias vectors.
• ‘usegpu=(gpu()>=0): GPU used by default if one exists.

RNNs compute the output h[t] for a given iteration from the recurrent input
h[t-1] and the previous layer input x[t] given matrices W, R and biases bW, bR
from the following equations:
:relu and :tanh: Single gate RNN with activation function f:

h[t] = f(W * x[t] .+ R * h[t-1] .+ bW .+ bR)

:gru: Gated recurrent unit:

i[t] = sigm(Wi * x[t] .+ Ri * h[t-1] .+ bWi .+ bRi) # input gate

r[t] = sigm(Wr * x[t] .+ Rr * h[t-1] .+ bWr .+ bRr) # reset gate
n[t] = tanh(Wn * x[t] .+ r[t] .* (Rn * h[t-1] .+ bRn) .+ bWn) # new gate
h[t] = (1 - i[t]) .* n[t] .+ i[t] .* h[t-1]

:lstm: Long short term memory unit with no peephole connections:

i[t] = sigm(Wi * x[t] .+ Ri * h[t-1] .+ bWi .+ bRi) # input gate

f[t] = sigm(Wf * x[t] .+ Rf * h[t-1] .+ bWf .+ bRf) # forget gate
o[t] = sigm(Wo * x[t] .+ Ro * h[t-1] .+ bWo .+ bRo) # output gate
n[t] = tanh(Wn * x[t] .+ Rn * h[t-1] .+ bWn .+ bRn) # new gate
c[t] = f[t] .* c[t-1] .+ i[t] .* n[t] # cell output
h[t] = o[t] .* tanh(c[t])

source

7.28 Knet.rnnparam
Knet.rnnparam — Method.

rnnparam{T}(r::RNN, w::KnetArray{T}, layer, id, param)

Return a single weight matrix or bias vector as a slice of w.

Valid layer values:

• For unidirectional RNNs 1:numLayers

• For bidirectional RNNs 1:2*numLayers, forw and back layers alternate.

Valid id values:
7.29. KNET.RNNPARAMS 195

• For RELU and TANH RNNs, input = 1, hidden = 2.

• For GRU reset = 1,4; update = 2,5; newmem = 3,6; 1:3 for input, 4:6 for
hidden

• For LSTM inputgate = 1,5; forget = 2,6; newmem = 3,7; output = 4,8;
1:4 for input, 5:8 for hidden

Valid param values:

• Return the weight matrix (transposed!) if param==1.

• Return the bias vector if param==2.

The effect of skipInput: Let I=1 for RELU/TANH, 1:3 for GRU, 1:4 for
LSTM

• For skipInput=false (default), rnnparam(r,w,1,I,1) is a (inputSize,hiddenSize)

matrix.

• For skipInput=true, rnnparam(r,w,1,I,1) is nothing.

• For bidirectional, the same applies to rnnparam(r,w,2,I,1): the first back

layer.

source

7.29 Knet.rnnparams
Knet.rnnparams — Method.

rnnparams(r::RNN, w)

Split w into individual parameters and return them as an array.

The order of params returned (subject to change):

• All weight matrices come before all bias vectors.

• Matrices and biases are sorted lexically based on (layer,id).

• See @doc rnnparam for valid layer and id values.

• Input multiplying matrices are nothing if r.inputMode = 1.

source
196 CHAPTER 7. KNET

7.30 Knet.setseed
Knet.setseed — Method.

setseed(n::Integer)

Run srand(n) on both cpu and gpu.

source

7.31 Knet.sigm
Knet.sigm — Function.
sigm(x) = (1./(1+exp(-x)))
source

7.32 Knet.unpool
Knet.unpool — Method.
Unpooling; reverse of pooling.

x == pool(unpool(x;o...); o...)

source

7.33 Knet.update!
Knet.update! — Function.

update!(weights, gradients, params)

update!(weights, gradients; lr=0.001, gclip=0)

Update the weights using their gradients and the optimization algorithm
parameters specified by params. The 2-arg version defaults to the Sgd algorithm
with learning rate lr and gradient clip gclip. gclip==0 indicates no clipping.
The weights and possibly gradients and params are modified in-place.
weights can be an individual numeric array or a collection of arrays repre-
sented by an iterator or dictionary. In the individual case, gradients should
be a similar numeric array of size(weights) and params should be a single
object. In the collection case, each individual weight array should have a cor-
responding params object. This way different weight arrays can have their own
optimization state, different learning rates, or even different optimization algo-
rithms running in parallel. In the iterator case, gradients and params should
be iterators of the same length as weights with corresponding elements. In the
dictionary case, gradients and params should be dictionaries with the same
keys as weights.
7.34. KNET.XAVIER 197

Individual optimization parameters can be one of the following types. The

keyword arguments for each type’s constructor and their default values are listed
as well.

• Sgd(;lr=0.001, gclip=0)

• Momentum(;lr=0.001, gclip=0, gamma=0.9)

• Nesterov(;lr=0.001, gclip=0, gamma=0.9)

• Rmsprop(;lr=0.001, gclip=0, rho=0.9, eps=1e-6)

• Adagrad(;lr=0.1, gclip=0, eps=1e-6)

• Adadelta(;lr=0.01, gclip=0, rho=0.9, eps=1e-6)

• Adam(;lr=0.001, gclip=0, beta1=0.9, beta2=0.999, eps=1e-8)

Example:

w = rand(d) # an individual weight array

g = lossgradient(w) # gradient g has the same shape as w
update!(w, g) # update w in-place with Sgd()
update!(w, g; lr=0.1) # update w in-place with Sgd(lr=0.1)
update!(w, g, Sgd(lr=0.1)) # update w in-place with Sgd(lr=0.1)

w = (rand(d1), rand(d2)) # a tuple of weight arrays

g = lossgradient2(w) # g will also be a tuple
p = (Adam(), Sgd()) # p has params for each w[i]
update!(w, g, p) # update each w[i] in-place with g[i],p[i]

w = Any[rand(d1), rand(d2)] # any iterator can be used

g = lossgradient3(w) # g will be similar to w
p = Any[Adam(), Sgd()] # p should be an iterator of same length
update!(w, g, p) # update each w[i] in-place with g[i],p[i]

w = Dict(:a => rand(d1), :b => rand(d2)) # dictionaries can be used

g = lossgradient4(w)
p = Dict(:a => Adam(), :b => Sgd())
update!(w, g, p)

source

7.34 Knet.xavier
Knet.xavier — Method.

xavier(a...)
198 CHAPTER 7. KNET

Xavier initialization. The a arguments are passed to rand. See (Glorot

and Bengio 2010) for a description. Caffe implements this slightly differently.
Lasagne calls it GlorotUniform.
source
Chapter 8

BinaryProvider

8.1 Base.merge
Base.merge — Method.
merge(collector::OutputCollector; colored::Bool = false)
Merge the stdout and stderr streams of the OutputCollector on a per-
line basis, returning a single string containing all collected lines, interleaved by
capture time. If colored is set to true, embeds terminal color codes to print
stderr in red.
source

8.2 BinaryProvider.activate
BinaryProvider.activate — Method.
activate(prefix::Prefix)
Prepends paths to environment variables so that binaries and libraries are
available to Julia.
source

8.3 BinaryProvider.activate
BinaryProvider.activate — Method.
activate(func::Function, prefix::Prefix)
Prepends paths to environment variables so that binaries and libraries are
available to Julia, calls the user function func, then deactivate()’s the ‘prefix“
again.
source

199
200 CHAPTER 8. BINARYPROVIDER

8.4 BinaryProvider.arch
BinaryProvider.arch — Method.

arch(p::Platform)

Get the architecture for the given Platform object as a Symbol.

Examples

julia> arch(Linux(:aarch64))
:aarch64

julia> arch(MacOS())
:x86_64

source

8.5 BinaryProvider.bindir
BinaryProvider.bindir — Method.

bindir(prefix::Prefix)

Returns the binary directory for the given prefix.

source

8.6 BinaryProvider.collect stderr

BinaryProvider.collect stderr — Method.

collect_stderr(collector::OutputCollector)

Returns all stderr lines collected by this collector so far.

source

8.7 BinaryProvider.collect stdout

BinaryProvider.collect stdout — Method.

collect_stdout(collector::OutputCollector)

Returns all stdout lines collected by this collector so far.

source
8.8. BINARYPROVIDER.DEACTIVATE 201

8.8 BinaryProvider.deactivate
BinaryProvider.deactivate — Method.

deactivate(prefix::Prefix)

Removes paths added to environment variables by activate()

source

8.9 BinaryProvider.download verify

BinaryProvider.download verify — Method.

download_verify(url::AbstractString, hash::AbstractString,
dest::AbstractString; verbose::Bool = false,
force::Bool = false, quiet_download::Bool = false)

Download file located at url, verify it matches the given hash, and throw
an error if anything goes wrong. If dest already exists, just verify it. If force
is set to true, overwrite the given file if it exists but does not match the given
hash.
This method returns true if the file was downloaded successfully, false if
an existing file was removed due to the use of force, and throws an error if
force is not set and the already-existent file fails verification, or if force is set,
verification fails, and then verification fails again after redownloading the file.
If quiet download is set to false (the default), this method will print to
stdout when downloading a new file. If it is set to true (and verbose is set
to false) the downloading process will be completely silent. If verbose is set
to true, messages about integrity verification will be printed in addition to
messages regarding downloading.
source

8.10 BinaryProvider.download verify unpack

BinaryProvider.download verify unpack — Method.

download_verify_unpack(url::AbstractString, hash::AbstractString,
dest::AbstractString; verbose::Bool = false,
force::Bool = false)

Helper method to download tarball located at url, verify it matches the

given hash, then unpack it into folder dest. In general, the method install()
should be used to download and install tarballs into a Prefix; this method
should only be used if the extra functionality of install() is undesired.
If tarball path is specified, the given url will be downloaded to tarball path,
and it will not be removed after downloading and verification is complete. If
202 CHAPTER 8. BINARYPROVIDER

it is not specified, the tarball will be downloaded to a temporary location, and

removed after verification is complete.
If force is specified, a verification failure will cause tarball path to be
deleted (if it exists), the dest folder to be removed (if it exists) and the tarball
to be redownloaded and reverified. If the verification check is failed a second
time, an exception is raised. If force is not specified, a verification failure will
result in an immediate raised exception.
source

8.11 BinaryProvider.extract platform key

BinaryProvider.extract platform key — Method.

extract_platform_key(path::AbstractString)

Given the path to a tarball, return the platform key of that tarball. If none
can be found, prints a warning and return the current platform suffix.
source

8.12 BinaryProvider.gen download cmd

BinaryProvider.gen download cmd — Function.

gen_download_cmd(url::AbstractString, out_path::AbstractString)

Return a Cmd that will download resource located at url and store it at the
location given by out path.
This method is initialized by probe platform engines(), which should be
automatically called upon first import of BinaryProvider.
source

8.13 BinaryProvider.gen list tarball cmd

BinaryProvider.gen list tarball cmd — Function.

gen_list_tarball_cmd(tarball_path::AbstractString)

Return a Cmd that will list the files contained within the tarball located at
tarball path. The list will not include directories contained within the tarball.
This method is initialized by probe platform engines(), which should be
automatically called upon first import of BinaryProvider.
source
8.14. BINARYPROVIDER.GEN PACKAGE CMD 203

8.14 BinaryProvider.gen package cmd

BinaryProvider.gen package cmd — Function.

gen_package_cmd(in_path::AbstractString, tarball_path::AbstractString)

Return a Cmd that will package up the given in path directory into a tarball
located at tarball path.
This method is initialized by probe platform engines(), which should be
automatically called upon first import of BinaryProvider.
source

8.15 BinaryProvider.gen sh cmd

BinaryProvider.gen sh cmd — Function.

gen_sh_cmd(cmd::Cmd)

Runs a command using sh. On Unices, this will default to the first sh found
on the PATH, however on Windows if that is not found it will fall back to the sh
provided by the busybox.exe shipped with Julia.
This method is initialized by probe platform engines(), which should be
automatically called upon first import of BinaryProvider.
source

8.16 BinaryProvider.gen unpack cmd

BinaryProvider.gen unpack cmd — Function.

gen_unpack_cmd(tarball_path::AbstractString, out_path::AbstractString)

Return a Cmd that will unpack the given tarball path into the given out path.
If out path is not already a directory, it will be created.
This method is initialized by probe platform engines(), which should be
automatically called upon first import of BinaryProvider.
source

8.17 BinaryProvider.includedir
BinaryProvider.includedir — Method.

includedir(prefix::Prefix)

Returns the include directory for the given prefix

source
204 CHAPTER 8. BINARYPROVIDER

8.18 BinaryProvider.install
BinaryProvider.install — Method.

install(tarball_url::AbstractString,
hash::AbstractString;
prefix::Prefix = global_prefix,
force::Bool = false,
ignore_platform::Bool = false,
verbose::Bool = false)

Given a prefix, a tarball url and a hash, download that tarball into the
prefix, verify its integrity with the hash, and install it into the prefix. Also
save a manifest of the files into the prefix for uninstallation later.
This will not overwrite any files within prefix unless force=true is set. If
force=true is set, installation will overwrite files as needed, and it will also
delete any files previously installed for tarball url as listed in a pre-existing
manifest (if any).
By default, this will not install a tarball that does not match the platform
of the current host system, this can be overridden by setting ignore platform.
source

8.19 BinaryProvider.isinstalled
BinaryProvider.isinstalled — Method.

isinstalled(tarball_url::AbstractString,
hash::AbstractString;
prefix::Prefix = global_prefix)

Given a prefix, a tarball url and a hash, check whether the tarball with
that hash has been installed into prefix.
In particular, it checks for the tarball, matching hash file, and manifest
installed by install, and checks that the files listed in the manifest are installed
and are not older than the tarball.
source

8.20 BinaryProvider.libdir
BinaryProvider.libdir — Function.

libdir(prefix::Prefix, platform = platform_key())

Returns the library directory for the given prefix (note that this differs
between unix systems and windows systems).
source
8.21. BINARYPROVIDER.LIST TARBALL FILES 205

8.21 BinaryProvider.list tarball files

BinaryProvider.list tarball files — Method.

list_tarball_files(path::AbstractString; verbose::Bool = false)

Given a .tar.gz filepath, list the compressed contents.

source

8.22 BinaryProvider.locate
BinaryProvider.locate — Method.
locate(fp::ExecutableProduct; platform::Platform = platform key(),
verbose::Bool = false, isolate::Bool = false)
If the given executable file exists and is executable, return its path.
On all platforms, an ExecutableProduct checks for existence of the file. On
non-Windows platforms, it will check for the executable bit being set. On
Windows platforms, it will check that the file ends with “.exe”, (adding it on
automatically, if it is not already present).
source

8.23 BinaryProvider.locate
BinaryProvider.locate — Method.
locate(fp::FileProduct; platform::Platform = platform key(), verbose::Bool
= false, isolate::Bool = false)
If the given file exists, return its path. The platform argument is ignored
here, but included for uniformity.
source

8.24 BinaryProvider.locate
BinaryProvider.locate — Method.
locate(lp::LibraryProduct; verbose::Bool = false, platform::Platform = plat-
form key())
If the given library exists (under any reasonable name) and is dlopen()able,
(assuming it was built for the current platform) return its location. Note that
the dlopen() test is only run if the current platform matches the given platform
keyword argument, as cross-compiled libraries cannot be dlopen()ed on foreign
platforms.
source
206 CHAPTER 8. BINARYPROVIDER

8.25 BinaryProvider.logdir
BinaryProvider.logdir — Method.

logdir(prefix::Prefix)

Returns the logs directory for the given prefix.

source

8.26 BinaryProvider.manifest for file

BinaryProvider.manifest for file — Method.

manifest_for_file(path::AbstractString; prefix::Prefix = global_prefix)

Returns the manifest file containing the installation receipt for the given
path, throws an error if it cannot find a matching manifest.
source

8.27 BinaryProvider.manifest from url

BinaryProvider.manifest from url — Method.

manifest_from_url(url::AbstractString; prefix::Prefix = global_prefix())

Returns the file path of the manifest file for the tarball located at url.
source

8.28 BinaryProvider.package
BinaryProvider.package — Method.

package(src_dir::AbstractString, tarball_path::AbstractString;
verbose::Bool = false)

Compress src dir into a tarball located at tarball path.

source

8.29 BinaryProvider.package
BinaryProvider.package — Method.

package(prefix::Prefix, output_base::AbstractString,
version::VersionNumber;
platform::Platform = platform_key(),
verbose::Bool = false, force::Bool = false)
8.30. BINARYPROVIDER.PARSE 7Z LIST 207

Build a tarball of the prefix, storing the tarball at output base, appending
a version number, a platform-dependent suffix and a file extension. If no plat-
form is given, defaults to current platform. Runs an audit() on the prefix,
to ensure that libraries can be dlopen()’ed, that all dependencies are located
within the prefix, etc. . . See the audit() documentation for a full list of the
audit steps. Returns the full path to and hash of the generated tarball.
source

8.30 BinaryProvider.parse 7z list

BinaryProvider.parse 7z list — Method.

parse_7z_list(output::AbstractString)

Given the output of 7z l, parse out the listed filenames. This funciton used
by list tarball files.
source

8.31 BinaryProvider.parse tar list

BinaryProvider.parse tar list — Method.

parse_7z_list(output::AbstractString)

Given the output of tar -t, parse out the listed filenames. This funciton
used by list tarball files.
source

8.32 BinaryProvider.parse tarball listing

BinaryProvider.parse tarball listing — Function.

parse_tarball_listing(output::AbstractString)

Parses the result of gen list tarball cmd() into something useful.
This method is initialized by probe platform engines(), which should be
automatically called upon first import of BinaryProvider.
source

8.33 BinaryProvider.platform dlext

BinaryProvider.platform dlext — Method.

platform_dlext(platform::Platform = platform_key())
208 CHAPTER 8. BINARYPROVIDER

Return the dynamic library extension for the given platform, defaulting to
the currently running platform. E.g. returns “so” for a Linux-based platform,
“dll” for a Windows-based platform, etc. . .
source

8.34 BinaryProvider.platform key

BinaryProvider.platform key — Function.

platform_key(machine::AbstractString = Sys.MACHINE)

Returns the platform key for the current platform, or any other though the
the use of the machine parameter.
source

8.35 BinaryProvider.satisfied
BinaryProvider.satisfied — Method.

satisfied(p::Product; platform::Platform = platform_key(),

verbose::Bool = false, isolate::Bool = false)

Given a Product, return true if that Product is satisfied, e.g. whether a

file exists that matches all criteria setup for that Product. If isolate is set
to true, will isolate all checks from the main Julia process in the event that
dlopen()’ing a library might cause issues.
source

8.36 BinaryProvider.tail
BinaryProvider.tail — Method.

tail(collector::OutputCollector; len::Int = 100, colored::Bool = false)

Write out the last len lines, optionally writing colored lines.
source

8.37 BinaryProvider.tee
BinaryProvider.tee — Method.

tee(c::OutputCollector; colored::Bool = false, stream::IO = stdout)

Spawn a background task to incrementally output lines from collector to

the standard output, optionally colored.
source
8.38. BINARYPROVIDER.TEMP PREFIX 209

8.38 BinaryProvider.temp prefix

BinaryProvider.temp prefix — Method.

temp_prefix(func::Function)

Create a temporary prefix, passing the prefix into the user-defined function
so that build/packaging operations can occur within the temporary prefix, which
is then cleaned up after all operations are finished. If the path provided exists
already, it will be deleted.
Usage example:

out_path = abspath("./libfoo")
temp_prefix() do p
# <insert build steps here>

# tarball up the built package

tarball_path, tarball_hash = package(p, out_path)
end

source

8.39 BinaryProvider.triplet
BinaryProvider.triplet — Method.

triplet(platform)

Get the target triplet for the given Platform object as a String.
Examples

julia> triplet(MacOS())
"x86_64-apple-darwin14"

julia> triplet(Windows(:i686))
"i686-w64-mingw32"

julia> triplet(Linux(:armv7l))
"arm-linux-gnueabihf"

source

8.40 BinaryProvider.uninstall
BinaryProvider.uninstall — Method.

uninstall(manifest::AbstractString; verbose::Bool = false)

210 CHAPTER 8. BINARYPROVIDER

Uninstall a package from a prefix by providing the manifest path that was
generated during install(). To find the manifest file for a particular in-
stalled file, use manifest for file(file path; prefix=prefix).
source

8.41 BinaryProvider.unpack
BinaryProvider.unpack — Method.

unpack(tarball_path::AbstractString, dest::AbstractString;
verbose::Bool = false)

Unpack tarball located at file tarball path into directory dest.

source

8.42 BinaryProvider.valid dl path

BinaryProvider.valid dl path — Method.

valid_dl_path(path::AbstractString, platform::Platform)

Return true if the given path ends in a valid dynamic library filename. E.g.
returns true for a path like "usr/lib/libfoo.so.3.5", but returns false for
a path like "libbar.so.f.a".
source

8.43 BinaryProvider.variable name

BinaryProvider.variable name — Method.

variable_name(p::Product)

Return the variable name associated with this Product as a string

source

8.44 BinaryProvider.verify
BinaryProvider.verify — Method.

verify(path::AbstractString, hash::AbstractString;
verbose::Bool = false, report_cache_status::Bool = false)
8.45. BINARYPROVIDER.WORDSIZE 211

Given a file path and a hash, calculate the SHA256 of the file and compare
it to hash. If an error occurs, verify() will throw an error. This method caches
verification results in a "$(path).sha256" file to accelerate re- verification of
files that have been previously verified. If no ".sha256" file exists, a full verifi-
cation will be done and the file will be created, with the calculated hash being
stored within the ".sha256" file.. If a ".sha256" file does exist, its contents
are checked to ensure that the hash contained within matches the given hash
parameter, and its modification time shows that the file located at path has not
been modified since the last verification.
If report cache status is set to true, then the return value will be a
Symbol giving a granular status report on the state of the hash cache, in addition
to the true/false signifying whether verification completed successfully.
source

8.45 BinaryProvider.wordsize
BinaryProvider.wordsize — Method.

wordsize(platform)

Get the word size for the given Platform object.

Examples

julia> wordsize(Linux(:arm7vl))
32

julia> wordsize(MacOS())
64

source

8.46 BinaryProvider.write deps file

BinaryProvider.write deps file — Method.

write_deps_file(depsjl_path::AbstractString, products::Vector{Product};
verbose::Bool = false)

Generate a deps.jl file that contains the variables referred to by the prod-
ucts within products. As an example, running the following code:

fooifier = ExecutableProduct(..., :foo_exe)

libbar = LibraryProduct(..., :libbar)
write_deps_file(joinpath(@__DIR__, "deps.jl"), [fooifier, libbar])
212 CHAPTER 8. BINARYPROVIDER

Will generate a deps.jl file that contains definitions for the two vari-
ables foo exe and libbar. If any Product object cannot be satisfied (e.g.
LibraryProduct objects must be dlopen()-able, FileProduct objects must
exist on the filesystem, etc. . . ) this method will error out. Ensure that you
have used install() to install the binaries you wish to write a deps.jl file for.
The result of this method is a deps.jl file containing variables named as
defined within the Product objects passed in to it, holding the full path to the
installed binaries. Given the example above, it would contain code similar to:

global const foo_exe = "<pkg path>/deps/usr/bin/fooifier"

global const libbar = "<pkg path>/deps/usr/lib/libbar.so"

This deps.jl file is intended to be include()’ed from within the top-level

source of your package. Note that all files are checked for consistency on package
load time, and if an error is discovered, package loading will fail, asking the user
to re-run Pkg.build("package name").
source
Chapter 9

DataStructures

9.1 Base.pop!
Base.pop! — Method.

pop!(sc, k)

Deletes the item with key k in SortedDict or SortedSet sc and returns the
value that was associated with k in the case of SortedDict or k itself in the case
of SortedSet. A KeyError results if k is not in sc. Time: O(c log n)
source

9.2 Base.pop!
Base.pop! — Method.

pop!(sc, k)

Deletes the item with key k in SortedDict or SortedSet sc and returns the
value that was associated with k in the case of SortedDict or k itself in the case
of SortedSet. A KeyError results if k is not in sc. Time: O(c log n)
source

9.3 Base.pop!
Base.pop! — Method.

pop!(ss)

Deletes the item with first key in SortedSet ss and returns the key. A
BoundsError results if ss is empty. Time: O(c log n)
source

213
214 CHAPTER 9. DATASTRUCTURES

9.4 Base.push!
Base.push! — Method.

push!(sc, k)

Argument sc is a SortedSet and k is a key. This inserts the key into the
container. If the key is already present, this overwrites the old value. (This is
not necessarily a no-op; see below for remarks about the customizing the sort
order.) The return value is sc. Time: O(c log n)
source

9.5 Base.push!
Base.push! — Method.

push!(sc, k=>v)

Argument sc is a SortedDict or SortedMultiDict and k=>v is a key-value

pair. This inserts the key-value pair into the container. If the key is already
present, this overwrites the old value. The return value is sc. Time: O(c log n)
source

9.6 Base.push!
Base.push! — Method.

push!(sc, k=>v)

Argument sc is a SortedDict or SortedMultiDict and k=>v is a key-value

pair. This inserts the key-value pair into the container. If the key is already
present, this overwrites the old value. The return value is sc. Time: O(c log n)
source

9.7 DataStructures.back
DataStructures.back — Method.

back(q::Deque)

Returns the last element of the deque q.

source
9.8. DATASTRUCTURES.COMPARE 215

9.8 DataStructures.compare
DataStructures.compare — Method.

compare(m::SAContainer, s::IntSemiToken, t::IntSemiToken)

Determines the relative positions of the data items indexed by (m,s) and
(m,t) in the sorted order. The return value is -1 if (m,s) precedes (m,t), 0 if
they are equal, and 1 if (m,s) succeeds (m,t). s and t are semitokens for the
same container m.
source

9.9 DataStructures.counter
DataStructures.counter — Method.

counter(seq)

Returns an Accumulator object containing the elements from seq.

source

9.10 DataStructures.dec!
DataStructures.dec! — Method.

dec!(ct, x, [v=1])

Decrements the count for x by v (defaulting to one)

source

9.11 DataStructures.deque
DataStructures.deque — Method.

deque(T)

Create a deque of type T.

source

9.12 DataStructures.dequeue!
DataStructures.dequeue! — Method.

dequeue!(pq)

Remove and return the lowest priority key from a priority queue.
216 CHAPTER 9. DATASTRUCTURES

julia> a = PriorityQueue(["a","b","c"],[2,3,1],Base.Order.Forward)
PriorityQueue{String,Int64,Base.Order.ForwardOrdering} with 3 entries:
"c" => 1
"b" => 3
"a" => 2

julia> dequeue!(a)
"c"

julia> a
PriorityQueue{String,Int64,Base.Order.ForwardOrdering} with 2 entries:
"b" => 3
"a" => 2
source

9.13 DataStructures.dequeue!
DataStructures.dequeue! — Method.
dequeue!(s::Queue)
Removes an element from the front of the queue s and returns it.
source

9.14 DataStructures.dequeue pair!

DataStructures.dequeue pair! — Method.
dequeue_pair!(pq)
Remove and return a the lowest priority key and value from a priority queue
as a pair.
julia> a = PriorityQueue(["a","b","c"],[2,3,1],Base.Order.Forward)
PriorityQueue{String,Int64,Base.Order.ForwardOrdering} with 3 entries:
"c" => 1
"b" => 3
"a" => 2

julia> dequeue_pair!(a)
"c" => 1

julia> a
PriorityQueue{String,Int64,Base.Order.ForwardOrdering} with 2 entries:
"b" => 3
"a" => 2
source
9.15. DATASTRUCTURES.ENQUEUE! 217

9.15 DataStructures.enqueue!
DataStructures.enqueue! — Method.
enqueue!(pq, k, v)
Insert the a key k into a priority queue pq with priority v.
source

9.16 DataStructures.enqueue!
DataStructures.enqueue! — Method.
enqueue!(s::Queue, x)
Inserts the value x to the end of the queue s.
source

9.17 DataStructures.enqueue!
DataStructures.enqueue! — Method.
enqueue!(pq, k=>v)
Insert the a key k into a priority queue pq with priority v.
julia> a = PriorityQueue(PriorityQueue("a"=>1, "b"=>2, "c"=>3))
PriorityQueue{String,Int64,Base.Order.ForwardOrdering} with 3 entries:
"c" => 3
"b" => 2
"a" => 1

julia> enqueue!(a, "d"=>4)

PriorityQueue{String,Int64,Base.Order.ForwardOrdering} with 4 entries:
"c" => 3
"b" => 2
"a" => 1
"d" => 4
source

9.18 DataStructures.find root

DataStructures.find root — Method.
find_root{T}(s::DisjointSets{T}, x::T)
Finds the root element of the subset in s which has the element x as a
member.
source
218 CHAPTER 9. DATASTRUCTURES

9.19 DataStructures.findkey
DataStructures.findkey — Method.
findkey(sd, k)
Argument sd is a SortedDict and argument k is a key. This function returns
the semitoken that refers to the item whose key is k, or past-end semitoken if k
is absent. Time: O(c log n)
source

9.20 DataStructures.front
DataStructures.front — Method.
front(q::Deque)
Returns the first element of the deque q.
source

9.21 DataStructures.heapify
DataStructures.heapify — Function.
heapify(v, ord::Ordering=Forward)
Returns a new vector in binary heap order, optionally using the given order-
ing.
julia> a = [1,3,4,5,2];

julia> heapify(a)
5-element Array{Int64,1}:
1
2
4
5
3

julia> heapify(a, Base.Order.Reverse)

5-element Array{Int64,1}:
5
3
4
1
2
source
9.22. DATASTRUCTURES.HEAPIFY! 219

9.22 DataStructures.heapify!
DataStructures.heapify! — Function.
heapify!(v, ord::Ordering=Forward)
In-place heapify.
source

9.23 DataStructures.heappop!
DataStructures.heappop! — Function.
heappop!(v, [ord])
Given a binary heap-ordered array, remove and return the lowest ordered
element. For efficiency, this function does not check that the array is indeed
heap-ordered.
source

9.24 DataStructures.heappush!
DataStructures.heappush! — Function.
heappush!(v, x, [ord])
Given a binary heap-ordered array, push a new element x, preserving the
heap property. For efficiency, this function does not check that the array is
indeed heap-ordered.
source

9.25 DataStructures.in same set

DataStructures.in same set — Method.
in_same_set(s::IntDisjointSets, x::Integer, y::Integer)
Returns true if x and y belong to the same subset in s and false otherwise.
source

9.26 DataStructures.inc!
DataStructures.inc! — Method.
inc!(ct, x, [v=1])
Increments the count for x by v (defaulting to one)
source
220 CHAPTER 9. DATASTRUCTURES

9.27 DataStructures.isheap
DataStructures.isheap — Function.
isheap(v, ord::Ordering=Forward)
Return true if an array is heap-ordered according to the given order.
julia> a = [1,2,3]
3-element Array{Int64,1}:
1
2
3

julia> isheap(a,Base.Order.Forward)
true

julia> isheap(a,Base.Order.Reverse)
false
source

9.28 DataStructures.nlargest
DataStructures.nlargest — Method.
nlargest(acc::Accumulator, [n])
Returns a sorted vector of the n most common elements, with their counts.
If n is omitted, the full sorted collection is returned.
This corresponds to Python’s Counter.most common function.
Example
julia> nlargest(counter("abbbccddddda"))

4-element Array{Pair{Char,Int64},1}:
’d’=>5
’b’=>3
’c’=>2
’a’=>2

julia> nlargest(counter("abbbccddddda"),2)

2-element Array{Pair{Char,Int64},1}:
’d’=>5
’b’=>3
source
9.29. DATASTRUCTURES.NLARGEST 221

9.29 DataStructures.nlargest
DataStructures.nlargest — Method.

nlargest(n, arr)

Return the n largest elements of the array arr.

Equivalent to sort(arr, lt = >)[1:min(n, end)]
source

9.30 DataStructures.nsmallest
DataStructures.nsmallest — Method.

nsmallest(acc::Accumulator, [n])

Returns a sorted vector of the n least common elements, with their counts.
If n is omitted, the full sorted collection is returned.
This is the opposite of the nlargest function. For obvious reasons this will
not include zero counts for items not encountered. (unless those elements are
added to he accumulator directly, eg via ‘acc[foo]=0)
source

9.31 DataStructures.nsmallest
DataStructures.nsmallest — Method.

nsmallest(n, arr)

Return the n smallest elements of the array arr.

Equivalent to sort(arr, lt = <)[1:min(n, end)]
source

9.32 DataStructures.orderobject
DataStructures.orderobject — Method.

orderobject(sc)

Returns the order object used to construct the container. Time: O(1)
source
222 CHAPTER 9. DATASTRUCTURES

9.33 DataStructures.orderobject
DataStructures.orderobject — Method.

orderobject(sc)

Returns the order object used to construct the container. Time: O(1)
source

9.34 DataStructures.orderobject
DataStructures.orderobject — Method.

orderobject(sc)

Returns the order object used to construct the container. Time: O(1)
source

9.35 DataStructures.ordtype
DataStructures.ordtype — Method.

ordtype(sc)

Returns the order type for SortedDict, SortedMultiDict and SortedSet. This
function may also be applied to the type itself. Time: O(1)
source

9.36 DataStructures.ordtype
DataStructures.ordtype — Method.

ordtype(sc)

Returns the order type for SortedDict, SortedMultiDict and SortedSet. This
function may also be applied to the type itself. Time: O(1)
source

9.37 DataStructures.ordtype
DataStructures.ordtype — Method.

ordtype(sc)

Returns the order type for SortedDict, SortedMultiDict and SortedSet. This
function may also be applied to the type itself. Time: O(1)
source
9.38. DATASTRUCTURES.PACKCOPY 223

9.38 DataStructures.packcopy
DataStructures.packcopy — Method.

packcopy(sc)

This returns a copy of sc in which the data is packed. When deletions take
place, the previously allocated memory is not returned. This function can be
used to reclaim memory after many deletions. Time: O(cn log n)
source

9.39 DataStructures.packcopy
DataStructures.packcopy — Method.

packcopy(sc)

This returns a copy of sc in which the data is packed. When deletions take
place, the previously allocated memory is not returned. This function can be
used to reclaim memory after many deletions. Time: O(cn log n)
source

9.40 DataStructures.packcopy
DataStructures.packcopy — Method.

packcopy(sc)

This returns a copy of sc in which the data is packed. When deletions take
place, the previously allocated memory is not returned. This function can be
used to reclaim memory after many deletions. Time: O(cn log n)
source

9.41 DataStructures.packdeepcopy
DataStructures.packdeepcopy — Method.

packdeepcopy(sc)

This returns a packed copy of sc in which the keys and values are deep-
copied. This function can be used to reclaim memory after many deletions.
Time: O(cn log n)
source
224 CHAPTER 9. DATASTRUCTURES

9.42 DataStructures.packdeepcopy
DataStructures.packdeepcopy — Method.
packdeepcopy(sc)
This returns a packed copy of sc in which the keys and values are deep-
copied. This function can be used to reclaim memory after many deletions.
Time: O(cn log n)
source

9.43 DataStructures.packdeepcopy
DataStructures.packdeepcopy — Method.
packdeepcopy(sc)
This returns a packed copy of sc in which the keys and values are deep-
copied. This function can be used to reclaim memory after many deletions.
Time: O(cn log n)
source

9.44 DataStructures.peek
DataStructures.peek — Method.
peek(pq)
Return the lowest priority key from a priority queue without removing that
key from the queue.
source

9.45 DataStructures.reset!
DataStructures.reset! — Method.
reset!(ct::Accumulator, x)
Resets the count of x to zero. Returns its former count.
source

9.46 DataStructures.top
DataStructures.top — Method.
top(h::BinaryHeap)
Returns the element at the top of the heap h.
source
9.47. DATASTRUCTURES.TOP WITH HANDLE 225

9.47 DataStructures.top with handle

DataStructures.top with handle — Method.

top_with_handle(h::MutableBinaryHeap)

Returns the element at the top of the heap h and its handle.
source

9.48 DataStructures.update!
DataStructures.update! — Method.

update!{T}(h::MutableBinaryHeap{T}, i::Int, v::T)

Replace the element at index i in heap h with v. This is equivalent to

h[i]=v.
source
Chapter 10

JDBC

10.1 JDBC.commit
JDBC.commit — Method.
Commit any pending transaction to the database. Throws a JDBCError if
connection is null.
Returns nothing.
source

10.2 JDBC.commit
JDBC.commit — Method.

commit(connection::JConnection)

Commits the transaction

Args

• connection: The connection object

Returns
None
source

10.3 JDBC.connection
JDBC.connection — Method.
Return the corresponding connection for a given cursor.
source

226
10.4. JDBC.CREATESTATEMENT 227

10.4 JDBC.createStatement
JDBC.createStatement — Method.

createStatement(connection::JConnection)

Initializes a Statement
Args

• connection: The connection object

Returns
The JStatement object
source

10.5 JDBC.cursor
JDBC.cursor — Method.
Create a new database cursor.
Returns a JDBCCursor instance.
source

10.6 JDBC.execute!
JDBC.execute! — Method.
Run a query on a database.
The results of the query are not returned by this function but are accessible
through the cursor.
parameters can be any iterable of positional parameters, or of some T<:Associative
for keyword/named parameters.
Throws a JDBCError if query caused an error, cursor is not initialized or
connection is null.
Returns nothing.
source

10.7 JDBC.execute
JDBC.execute — Method.

execute(stmt::JStatement, query::AbstractString)

Executes the auery based on JStatement or any of its sub-types

Args

• stmt: The JStatement object or any of its sub-types

228 CHAPTER 10. JDBC

• query: The query to be executed

Returns
A boolean indicating whether the execution was successful or not
source

10.8 JDBC.execute
JDBC.execute — Method.

execute(stmt::(Union{JPreparedStatement, JCallableStatement}))

Executes the auery based on the Prepared Statement or Callable Statement

Args

• stmt: The Prepared Statement or the Callable Statement object

Returns
A boolean indicating whether the execution was successful or not
source

10.9 JDBC.executeQuery
JDBC.executeQuery — Method.

executeQuery(stmt::JStatement, query::AbstractString)

Executes the auery and returns the results as a JResultSet object.

Args

• stmt: The Statement object

• query: The query to be executed

Returns
The result set as a JResultSet object
source

10.10 JDBC.executeQuery
JDBC.executeQuery — Method.

executeQuery(stmt::Union{JPreparedStatement, JCallableStatement})

Executes the auery based on a JPreparedStatement object or a JCallableState-

ment object
Args
10.11. JDBC.EXECUTEUPDATE 229

• stmt: The JPreparedStatement object or JCallableStatement object

Returns
The result set as a JResultSet object
source

10.11 JDBC.executeUpdate
JDBC.executeUpdate — Method.

executeUpdate(stmt::JStatement, query::AbstractString)

Executes the update auery and returns the status of the execution of the
query
Args

• stmt: The Statement object

• query: The query to be executed

Returns
An integer representing the status of the execution
source

10.12 JDBC.executeUpdate
JDBC.executeUpdate — Method.

executeUpdate(stmt::Union{JPreparedStatement, JCallableStatement})

Executes the update auery based on a JPreparedStatement object or a

JCallableStatement object
Args

• stmt: The JPreparedStatement object or JCallableStatement object

Returns
An integer indicating the status of the execution of the query
source

10.13 JDBC.getColumnCount
JDBC.getColumnCount — Method.

getColumnCount(rsmd::JResultSetMetaData)
230 CHAPTER 10. JDBC

Returns the number of columns based on the JResultSetMetaData object

Args

• rsmd: The JResultSetMetaData object

Returns
The number of columns.
source

10.14 JDBC.getColumnName
JDBC.getColumnName — Method.

getColumnName(rsmd::JResultSetMetaData, col::Integer)

Returns the column’s name based on the JResultSetMetaData object and

the column number
Args

• rsmd: The JResultSetMetaData object

• col: The column number

Returns
The column name
source

10.15 JDBC.getColumnType
JDBC.getColumnType — Method.

getColumnType(rsmd::JResultSetMetaData, col::Integer)

Returns the column’s data type based on the JResultSetMetaData object

and the column number
Args

• rsmd: The JResultSetMetaData object

• col: The column number

Returns
The column type as an integer
source
10.16. JDBC.GETDATE 231

10.16 JDBC.getDate
JDBC.getDate — Method.
getDate(rs::Union{JResultSet, JCallableStatement}, fld::AbstractString)
Returns the Date object based on the result set or a callable statement. The
value is extracted based on the column name.
Args
• stmt: The JResultSet or JCallableStatement object
• fld: The column name
Returns
The Date object.
source

10.17 JDBC.getDate
JDBC.getDate — Method.
getDate(rs::Union{JResultSet, JCallableStatement}, fld::Integer)
Returns the Date object based on the result set or a callable statement. The
value is extracted based on the column number.
Args
• stmt: The JResultSet or JCallableStatement object
• fld: The column number
Returns
The Date object.
source

10.18 JDBC.getMetaData
JDBC.getMetaData — Method.
getMetaData(rs::JResultSet)
Returns information about the types and properties of the columns in the
ResultSet object
Args
• stmt: The JResultSet object
Returns
The JResultSetMetaData object.
source
232 CHAPTER 10. JDBC

10.19 JDBC.getResultSet
JDBC.getResultSet — Method.

getResultSet(stmt::JStatement)

Returns the result set based on the previous execution of the query based
on a JStatement
Args

• stmt: The JStatement object

Returns
The JResultSet object.
source

10.20 JDBC.getTableMetaData
JDBC.getTableMetaData — Method.
Get the metadata (column name and type) for each column of the table in
the result set rs.
Returns an array of (column name, column type) tuples.
source

10.21 JDBC.getTime
JDBC.getTime — Method.

getTime(rs::Union{JResultSet, JCallableStatement}, fld::AbstractString)

Returns the Time object based on the result set or a callable statement. The
value is extracted based on the column name.
Args

• stmt: The JResultSet or JCallableStatement object

• fld: The column name

Returns
The Time object.
source
10.22. JDBC.GETTIME 233

10.22 JDBC.getTime
JDBC.getTime — Method.
getTime(rs::Union{JResultSet, JCallableStatement}, fld::Integer)
Returns the Time object based on the result set or a callable statement. The
value is extracted based on the column number.
Args
• stmt: The JResultSet or JCallableStatement object
• fld: The column number
Returns
The Time object.
source

10.23 JDBC.getTimestamp
JDBC.getTimestamp — Method.
getTimestamp(rs::Union{JResultSet, JCallableStatement}, fld::AbstractString)
Returns the Timestamp object based on the result set or a callable state-
ment. The value is extracted based on the column name.
Args
• stmt: The JResultSet or JCallableStatement object
• fld: The column name
Returns
The Timestamp object.
source

10.24 JDBC.getTimestamp
JDBC.getTimestamp — Method.
getTimestamp(rs::@Union{JResultSet, JCallableStatement}, fld::Integer)
Returns the Timestamp object based on the result set or a callable state-
ment. The value is extracted based on the column number.
Args
• stmt: The JResultSet or JCallableStatement object
• fld: The column number
Returns
The Timestamp object.
source
234 CHAPTER 10. JDBC

10.25 JDBC.isopen
JDBC.isopen — Method.
Returns a boolean indicating whether connection conn is open.
source

10.26 JDBC.prepareCall
JDBC.prepareCall — Method.

prepareCall(connection::JConnection, query::AbstractString)

Prepares the Callable Statement for the given query

Args

• connection: The connection object

• query: The query string

Returns
The JCallableStatement object
source

10.27 JDBC.prepareStatement
JDBC.prepareStatement — Method.

prepareStatement(connection::JConnection, query::AbstractString)

Prepares the Statement for the given query

Args

• connection: The connection object

• query: The query string

Returns
The JPreparedStatement object
source

10.28 JDBC.rollback
JDBC.rollback — Method.
Roll back to the start of any pending transaction. Throws a JDBCError if
connection is null.
Returns nothing.
source
10.29. JDBC.ROLLBACK 235

10.29 JDBC.rollback
JDBC.rollback — Method.

rollback(connection::JConnection)

Rolls back the transactions.

Args

• connection: The connection object

Returns
None
source

10.30 JDBC.rows
JDBC.rows — Method.
Create a row iterator.
This method returns an instance of an iterator type which returns one row
on each iteration. Each row returnes a Tuple{. . . }.
Throws a JDBCError if execute! was not called on the cursor or connection
is null.
Returns a JDBCRowIterator instance.
source

10.31 JDBC.setAutoCommit
JDBC.setAutoCommit — Method.

setAutoCommit(connection::JConnection, x::Bool)

Set the Auto Commit flag to either true or false. If set to false, commit has
to be called explicitly
Args

• connection: The connection object

Returns
None
source
236 CHAPTER 10. JDBC

10.32 JDBC.setFetchSize
JDBC.setFetchSize — Method.

setFetchSize(stmt::Union{JStatement, JPreparedStatement, JCallableStatement }, x::Integ

Sets the fetch size in a JStatement or a JPreparedStatement object or a

JCallableStatement object. The number of records that are returned in subse-
quent query executions are determined by what is set here.
Args

• stmt: The JPreparedStatement object or JCallableStatement object

• x: The number of records to be returned

Returns
None
source
Chapter 11

ImageCore

11.1 ImageCore.assert timedim last

ImageCore.assert timedim last — Method.

assert_timedim_last(img)

Throw an error if the image has a time dimension that is not the last di-
mension.
source

11.2 ImageCore.channelview
ImageCore.channelview — Method.

channelview(A)

returns a view of A, splitting out (if necessary) the color channels of A into a
new first dimension. This is almost identical to ChannelView(A), except that if
A is a ColorView, it will simply return the parent of A, or will use reinterpret
when appropriate. Consequently, the output may not be a ChannelView array.
Of relevance for types like RGB and BGR, the channels of the returned array
will be in constructor-argument order, not memory order (see reinterpret if
you want to use memory order).
source

11.3 ImageCore.clamp01
ImageCore.clamp01 — Method.

clamp01(x) -> y

237
238 CHAPTER 11. IMAGECORE

Produce a value y that lies between 0 and 1, and equal to x when x is already
in this range. Equivalent to clamp(x, 0, 1) for numeric values. For colors,
this function is applied to each color channel separately.
See also: clamp01nan.
source

11.4 ImageCore.clamp01nan
ImageCore.clamp01nan — Method.

clamp01nan(x) -> y

Similar to clamp01, except that any NaN values are changed to 0.

See also: clamp01.
source

11.5 ImageCore.colorsigned
ImageCore.colorsigned — Method.

colorsigned()
colorsigned(colorneg, colorpos) -> f
colorsigned(colorneg, colorcenter, colorpos) -> f

Define a function that maps negative values (in the range [-1,0]) to the linear
colormap between colorneg and colorcenter, and positive values (in the range
[0,1]) to the linear colormap between colorcenter and colorpos.
The default colors are:

• colorcenter: white

• colorneg: green1

• colorpos: magenta

See also: scalesigned.

source

11.6 ImageCore.colorview
ImageCore.colorview — Method.

colorview(C, gray1, gray2, ...) -> imgC

11.7. IMAGECORE.COLORVIEW 239

Combine numeric/grayscale images gray1, gray2, etc., into the separate

color channels of an array imgC with element type C<:Colorant.
As a convenience, the constant zeroarray fills in an array of matched size
with all zeros.
Example
[] imgC = colorview(RGB, r, zeroarray, b)
creates an image with r in the red chanel, b in the blue channel, and nothing
in the green channel.
See also: StackedView.
source

11.7 ImageCore.colorview
ImageCore.colorview — Method.

colorview(C, A)

returns a view of the numeric array A, interpreting successive elements of A as

if they were channels of Colorant C. This is almost identical to ColorView{C}(A),
except that if A is a ChannelView, it will simply return the parent of A, or
use reinterpret when appropriate. Consequently, the output may not be a
ColorView array.
Of relevance for types like RGB and BGR, the elements of A are interpreted
in constructor-argument order, not memory order (see reinterpret if you want
to use memory order).
Example
[] A = rand(3, 10, 10) img = colorview(RGB, A)
source

11.8 ImageCore.coords spatial

ImageCore.coords spatial — Method.
coords spatial(img)
Return a tuple listing the spatial dimensions of img.
Note that a better strategy may be to use ImagesAxes and take slices along
the time axis.
source

11.9 ImageCore.float32
ImageCore.float32 — Function.

float32.(img)
240 CHAPTER 11. IMAGECORE

converts the raw storage type of img to Float32, without changing the color
space.
source

11.10 ImageCore.float64
ImageCore.float64 — Function.

float64.(img)

converts the raw storage type of img to Float64, without changing the color
space.
source

11.11 ImageCore.indices spatial

ImageCore.indices spatial — Method.

indices_spatial(img)

Return a tuple with the indices of the spatial dimensions of the image. De-
faults to the same as indices, but using ImagesAxes you can mark some axes
as being non-spatial.
source

11.12 ImageCore.n0f16
ImageCore.n0f16 — Function.

n0f16.(img)

converts the raw storage type of img to N0f16, without changing the color
space.
source

11.13 ImageCore.n0f8
ImageCore.n0f8 — Function.

n0f8.(img)

converts the raw storage type of img to N0f8, without changing the color
space.
source
11.14. IMAGECORE.N2F14 241

11.14 ImageCore.n2f14
ImageCore.n2f14 — Function.

n2f14.(img)

converts the raw storage type of img to N2f14, without changing the color
space.
source

11.15 ImageCore.n4f12
ImageCore.n4f12 — Function.

n4f12.(img)

converts the raw storage type of img to N4f12, without changing the color
space.
source

11.16 ImageCore.n6f10
ImageCore.n6f10 — Function.

n6f10.(img)

converts the raw storage type of img to N6f10, without changing the color
space.
source

11.17 ImageCore.nimages
ImageCore.nimages — Method.

nimages(img)

Return the number of time-points in the image array. Defaults to

1. Use ImagesAxes if you want to use an explicit time dimension.

source
242 CHAPTER 11. IMAGECORE

11.18 ImageCore.normedview
ImageCore.normedview — Method.

normedview([T], img::AbstractArray{Unsigned})

returns a “view” of img where the values are interpreted in terms of Normed
number types. For example, if img is an Array{UInt8}, the view will act like an
Array{N0f8}. Supply T if the element type of img is UInt16, to specify whether
you want a N6f10, N4f12, N2f14, or N0f16 result.
source

11.19 ImageCore.permuteddimsview
ImageCore.permuteddimsview — Method.

permuteddimsview(A, perm)

returns a “view” of A with its dimensions permuted as specified by perm.

This is like permutedims, except that it produces a view rather than a copy of
A; consequently, any manipulations you make to the output will be mirrored in
A. Compared to the copy, the view is much faster to create, but generally slower
to use.
source

11.20 ImageCore.pixelspacing
ImageCore.pixelspacing — Method.

pixelspacing(img) -> (sx, sy, ...)

Return a tuple representing the separation between adjacent pixels along

each axis of the image. Defaults to (1,1,. . . ). Use ImagesAxes for images with
anisotropic spacing or to encode the spacing using physical units.
source

11.21 ImageCore.rawview
ImageCore.rawview — Method.

rawview(img::AbstractArray{FixedPoint})

returns a “view” of img where the values are interpreted in terms of their
raw underlying storage. For example, if img is an Array{N0f8}, the view will
act like an Array{UInt8}.
source
11.22. IMAGECORE.SCALEMINMAX 243

11.22 ImageCore.scaleminmax
ImageCore.scaleminmax — Method.

scaleminmax(min, max) -> f

scaleminmax(T, min, max) -> f

Return a function f which maps values less than or equal to min to 0, values
greater than or equal to max to 1, and uses a linear scale in between. min and
max should be real values.
Optionally specify the return type T. If T is a colorant (e.g., RGB), then
scaling is applied to each color channel.
Examples
Example 1
[] julia¿ f = scaleminmax(-10, 10) (::9) (generic function with 1 method)
julia¿ f(10) 1.0
julia¿ f(-10) 0.0
julia¿ f(5) 0.75
Example 2
[] julia¿ c = RGB(255.0,128.0,0.0) RGB{Float64}(255.0,128.0,0.0)
julia¿ f = scaleminmax(RGB, 0, 255) (::13) (generic function with 1 method)
julia¿ f(c) RGB{Float64}(1.0,0.5019607843137255,0.0)
See also: takemap.
source

11.23 ImageCore.scalesigned
ImageCore.scalesigned — Method.

scalesigned(maxabs) -> f

Return a function f which scales values in the range [-maxabs, maxabs]

(clamping values that lie outside this range) to the range [-1, 1].
See also: colorsigned.
source

11.24 ImageCore.scalesigned
ImageCore.scalesigned — Method.

scalesigned(min, center, max) -> f

Return a function f which scales values in the range [min, center] to

[-1,0] and [center,max] to [0,1]. Values smaller than min/max get clamped
to min/max, respectively.
See also: colorsigned.
source
244 CHAPTER 11. IMAGECORE

11.25 ImageCore.sdims
ImageCore.sdims — Method.

sdims(img)

Return the number of spatial dimensions in the image. Defaults to the same
as ndims, but with ImagesAxes you can specify that some axes correspond to
other quantities (e.g., time) and thus not included by sdims.
source

11.26 ImageCore.size spatial

ImageCore.size spatial — Method.

size_spatial(img)

Return a tuple listing the sizes of the spatial dimensions of the image. De-
faults to the same as size, but using ImagesAxes you can mark some axes as
being non-spatial.
source

11.27 ImageCore.spacedirections
ImageCore.spacedirections — Method.

spacedirections(img) -> (axis1, axis2, ...)

Return a tuple-of-tuples, each axis[i] representing the displacement vector

between adjacent pixels along spatial axis i of the image array, relative to some
external coordinate system (“physical coordinates”).
By default this is computed from pixelspacing, but you can set this man-
ually using ImagesMeta.
source

11.28 ImageCore.takemap
ImageCore.takemap — Function.

takemap(f, A) -> fnew

takemap(f, T, A) -> fnew

Given a value-mapping function f and an array A, return a “concrete” map-

ping function fnew. When applied to elements of A, fnew should return valid
values for storage or display, for example in the range from 0 to 1 (for grayscale)
11.28. IMAGECORE.TAKEMAP 245

or valid colorants. fnew may be adapted to the actual values present in A, and
may not produce valid values for any inputs not in A.
Optionally one can specify the output type T that fnew should produce.
Example:
[] julia¿ A = [0, 1, 1000];
julia¿ f = takemap(scaleminmax, A) (::7) (generic function with 1 method)
julia¿ f.(A) 3-element Array{Float64,1}: 0.0 0.001 1.0
source
Chapter 12

Reactive

12.1 Base.filter
Base.filter — Method.

filter(f, default, signal)

remove updates from the signal where f returns false. The filter will hold
the value default until f(value(signal)) returns true, when it will be updated to
value(signal).
source

12.2 Base.filter
Base.filter — Method.

filter(f, signal)

remove updates from the signal where f returns false. The filter will hold
the current value of the signal until f(value(signal)) returns true.
source

12.3 Base.map
Base.map — Method.

map(f, s::Signal...) -> signal

Transform signal s by applying f to each element. For multiple signal argu-

ments, apply f elementwise.
source

246
12.4. BASE.MERGE 247

12.4 Base.merge
Base.merge — Method.

merge(inputs...)

Merge many signals into one. Returns a signal which updates when any of
the inputs update. If many signals update at the same time, the value of the
youngest (most recently created) input signal is taken.
source

12.5 Base.push!
Base.push! — Function.
push!(signal, value, onerror=Reactive.print error)
Queue an update to a signal. The update will be propagated when all
currently queued updates are done processing.
The third (optional) argument, onerror, is a callback triggered when the
update ends in an error. The callback receives 4 arguments, onerror(sig,
val, node, capex), where sig and val are the Signal and value that push!
was called with, respectively, node is the Signal whose action triggered the error,
and capex is a CapturedException with the fields ex which is the original
exception object, and processed bt which is the backtrace of the exception.
The default error callback will print the error and backtrace to STDERR.
source

12.6 Reactive.async map

Reactive.async map — Method.

tasks, results = async_map(f, init, input...;typ=typeof(init), onerror=Reactive.print_error)

Spawn a new task to run a function when input signal updates. Returns
a signal of tasks and a results signal which updates asynchronously with the
results. init will be used as the default value of results. onerror is the
callback to be called when an error occurs, by default it is set to a callback
which prints the error to STDERR. It’s the same as the onerror argument to
push! but is run in the spawned task.
source

12.7 Reactive.bind!
Reactive.bind! — Function.

‘bind!(dest, src, twoway=true; initial=true)‘

248 CHAPTER 12. REACTIVE

for every update to src also update dest with the same value and, if twoway
is true, vice-versa. If initial is false, dest will only be updated to src’s value
when src next updates, otherwise (if initial is true) both dest and src will
take src’s value immediately.
source

12.8 Reactive.bindmap!
Reactive.bindmap! — Function.

‘bindmap!(dest::Signal, src2dest::Function, src::Signal, dest2src=nothing; initial=true

for every update to src also update dest with a modified value (using the
function src2dest) and, if dest2src is specified, a two-way update will hold.
If initial is false, dest will only be updated to src’s modified value when src
next updates, otherwise (if initial is true) both dest and src will take their
respective modified values immediately.
source

12.9 Reactive.bound dests

Reactive.bound dests — Method.
bound dests(src::Signal) returns a vector of all signals that will update
when src updates, that were bound using bind!(dest, src)
source

12.10 Reactive.bound srcs

Reactive.bound srcs — Method.
bound srcs(dest::Signal) returns a vector of all signals that will cause an
update to dest when they update, that were bound using bind!(dest, src)
source

12.11 Reactive.debounce
Reactive.debounce — Method.

debounce(dt, input, f=(acc,x)->x, init=value(input), reinit=x->x;

typ=typeof(init), name=auto_name!(string("debounce ",dt,"s"), input))

Creates a signal that will delay updating until dt seconds have passed since
the last time input has updated. By default, the debounce signal holds the last
update of the input signal since the debounce signal last updated.
12.12. REACTIVE.DELAY 249

This behavior can be changed by the f, init and reinit arguments. The
init and f functions are similar to init and f in foldp. reinit is called after
the debounce sends an update, to reinitialize the initial value for accumulation,
it gets one argument, the previous accumulated value.
For example y = debounce(0.2, x, push!, Int[], ->Int[]) will accu-
mulate a vector of updates to the integer signal x and push it after x is inactive
(doesn’t update) for 0.2 seconds.
source

12.12 Reactive.delay
Reactive.delay — Method.

delay(input, default=value(input))

Schedule an update to happen after the current update propagates through-

out the signal graph.
Returns the delayed signal.
source

12.13 Reactive.droprepeats
Reactive.droprepeats — Method.

droprepeats(input)

Drop updates to input whenever the new value is the same as the previous
value of the signal.
source

12.14 Reactive.every
Reactive.every — Method.

every(dt)

A signal that updates every dt seconds to the current timestamp. Consider

using fpswhen or fps if you want specify the timing signal by frequency, rather
than delay.
source
250 CHAPTER 12. REACTIVE

12.15 Reactive.filterwhen
Reactive.filterwhen — Method.

filterwhen(switch::Signal{Bool}, default, input)

Keep updates to input only when switch is true.

If switch is false initially, the specified default value is used.
source

12.16 Reactive.flatten
Reactive.flatten — Method.

flatten(input::Signal{Signal}; typ=Any)

Flatten a signal of signals into a signal which holds the value of the current
signal. The typ keyword argument specifies the type of the flattened signal. It
is Any by default.
source

12.17 Reactive.foldp
Reactive.foldp — Method.

foldp(f, init, inputs...)

Fold over past values.

Accumulate a value as the input signals change. init is the initial value of
the accumulator. f should take 1 + length(inputs) arguments: the first is
the current accumulated value and the rest are the current input signal values.
f will be called when one or more of the inputs updates. It should return the
next accumulated value.
source

12.18 Reactive.fps
Reactive.fps — Method.

fps(rate)

Same as fpswhen(Input(true), rate)

source
12.19. REACTIVE.FPSWHEN 251

12.19 Reactive.fpswhen
Reactive.fpswhen — Method.

fpswhen(switch, rate)

returns a signal which when switch signal is true, updates rate times every
second. If rate is not possible to attain because of slowness in computing
dependent signal values, the signal will self adjust to provide the best possible
rate.
source

12.20 Reactive.preserve
Reactive.preserve — Method.

preserve(signal::Signal)

prevents signal from being garbage collected as long as any of its parents
are around. Useful for when you want to do some side effects in a signal. e.g.
preserve(map(println, x)) - this will continue to print updates to x, until x
goes out of scope. foreach is a shorthand for map with preserve.
source

12.21 Reactive.previous
Reactive.previous — Method.

previous(input, default=value(input))

Create a signal which holds the previous value of input. You can optionally
specify a different initial value.
source

12.22 Reactive.remote map

Reactive.remote map — Method.

remoterefs, results = remote_map(procid, f, init, input...;typ=typeof(init), onerror=Reactive.pri

Spawn a new task on process procid to run a function when input signal
updates. Returns a signal of remote refs and a results signal which updates
asynchronously with the results. init will be used as the default value of
results. onerror is the callback to be called when an error occurs, by default
it is set to a callback which prints the error to STDERR. It’s the same as the
onerror argument to push! but is run in the spawned task.
source
252 CHAPTER 12. REACTIVE

12.23 Reactive.rename!
Reactive.rename! — Method.
rename!(s::Signal, name::String)
Change a Signal’s name
source

12.24 Reactive.sampleon
Reactive.sampleon — Method.
sampleon(a, b)
Sample the value of b whenever a updates.
source

12.25 Reactive.throttle
Reactive.throttle — Method.
throttle(dt, input, f=(acc,x)->x, init=value(input), reinit=x->x;
typ=typeof(init), name=auto_name!(string("throttle ",dt,"s"), input), leadi
Throttle a signal to update at most once every dt seconds. By default, the
throttled signal holds the last update of the input signal during each dt second
time window.
This behavior can be changed by the f, init and reinit arguments. The
init and f functions are similar to init and f in foldp. reinit is called when a
new throttle time window opens to reinitialize the initial value for accumulation,
it gets one argument, the previous accumulated value.
For example y = throttle(0.2, x, push!, Int[], ->Int[]) will cre-
ate vectors of updates to the integer signal x which occur within 0.2 second time
windows.
If leading is true, the first update from input will be sent immediately by
the throttle signal. If it is false, the first update will happen dt seconds after
input’s first update
New in v0.4.1: throttle’s behaviour from previous versions is now available
with the debounce signal type.
source

12.26 Reactive.unbind!
Reactive.unbind! — Function.
‘unbind!(dest, src, twoway=true)‘
remove a link set up using bind!
source
12.27. REACTIVE.UNPRESERVE 253

12.27 Reactive.unpreserve
Reactive.unpreserve — Method.

unpreserve(signal::Signal)

allow signal to be garbage collected. See also preserve.

source
Chapter 13

Combinatorics

13.1 Base.factorial
Base.factorial — Method.

factorial(n, k)

Compute n!/k!.
source

13.2 Combinatorics.bellnum
Combinatorics.bellnum — Method.

bellnum(n)

Compute the nth Bell number.

source

13.3 Combinatorics.catalannum
Combinatorics.catalannum — Method.

catalannum(n)

Compute the nth Catalan number.

source

254
13.4. COMBINATORICS.CHARACTER 255

13.4 Combinatorics.character
Combinatorics.character — Method.
character(::Partition, ::Partition)
Compute the character χλ(µ) of the partition in the th irreducible represen-
tation (“irrep”) of the symmetric group Sn .
Implements the Murnaghan-Nakayama algorithm as described in:
Dan Bernstein,
"The computational complexity of rules for the character table of Sn",
Journal of Symbolic Computation, vol. 37 iss. 6 (2004), pp 727-748.
doi:10.1016/j.jsc.2003.11.001
source

13.5 Combinatorics.combinations
Combinatorics.combinations — Method.
combinations(a, n)
Generate all combinations of n elements from an indexable object a. Because
the number of combinations can be very large, this function returns an iterator
object. Use collect(combinations(a, n)) to get an array of all combinations.
source

13.6 Combinatorics.combinations
Combinatorics.combinations — Method.
combinations(a)
Generate combinations of the elements of a of all orders. Chaining of order
iterators is eager, but the sequence at each order is lazy.
source

13.7 Combinatorics.derangement
Combinatorics.derangement — Method.
derangement(n)
Compute the number of permutations of n with no fixed points, also known
as the subfactorial. An alias subfactorial for this function is provided for
convenience.
source
256 CHAPTER 13. COMBINATORICS

13.8 Combinatorics.integer partitions

Combinatorics.integer partitions — Method.

integer_partitions(n)

List the partitions of the integer n.

!!! note The order of the resulting array is consistent with that produced by
the computational discrete algebra software GAP.
source

13.9 Combinatorics.isrimhook
Combinatorics.isrimhook — Method.

isrimhook(::SkewDiagram)
isrimhook(::Partition, ::Partition)

Check whether the given skew diagram is a rim hook.

source

13.10 Combinatorics.isrimhook
Combinatorics.isrimhook — Method.

isrimhook(a::Int, b::Int)

Take two elements of a partition sequence, with a to the left of b.

source

13.11 Combinatorics.lassallenum
Combinatorics.lassallenum — Method.

lassallenum(n)

Compute the nth entry in Lassalle’s sequence, OEIS entry A180874.

source

13.12 Combinatorics.leglength
Combinatorics.leglength — Method.

leglength(::SkewDiagram)
leglength(::Partition, ::Partition)
13.13. COMBINATORICS.LEVICIVITA 257

Compute the leg length for the given skew diagram.

!!! note Strictly speaking, the leg length is defined for rim hooks only, but
here we define it for all skew diagrams.
source

13.13 Combinatorics.levicivita
Combinatorics.levicivita — Method.

levicivita(p)

Compute the Levi-Civita symbol of a permutation p. Returns 1 if the per-

mutation is even, -1 if it is odd, and 0 otherwise.
The parity is computed by using the fact that a permutation is odd if and
only if the number of even-length cycles is odd.
source

13.14 Combinatorics.multiexponents
Combinatorics.multiexponents — Method.

multiexponents(m, n)

Returns the exponents in the multinomial expansion (x + x + . . . + x).

For example, the expansion (x + x + x) = x + xx + xx + . . . has the
exponents:

julia> collect(multiexponents(3, 2))

6-element Array{Any,1}:
[2, 0, 0]
[1, 1, 0]
[1, 0, 1]
[0, 2, 0]
[0, 1, 1]
[0, 0, 2]

source

13.15 Combinatorics.multinomial
Combinatorics.multinomial — Method.

multinomial(k...)

Multinomial coefficient where n = sum(k).

source
258 CHAPTER 13. COMBINATORICS

13.16 Combinatorics.multiset combinations

Combinatorics.multiset combinations — Method.

multiset_combinations(a, t)

Generate all combinations of size t from an array a with possibly duplicated

elements.
source

13.17 Combinatorics.multiset permutations

Combinatorics.multiset permutations — Method.

multiset_permutations(m, f, t)

Generate all permutations of size t from an array a with possibly duplicated

elements.
source

13.18 Combinatorics.nthperm!
Combinatorics.nthperm! — Method.

nthperm!(a, k)

In-place version of nthperm; the array a is overwritten.

source

13.19 Combinatorics.nthperm
Combinatorics.nthperm — Method.

nthperm(p)

Return the integer k that generated permutation p. Note that nthperm(nthperm([1:n],

k)) == k for 1 <= k <= factorial(n).
source

13.20 Combinatorics.nthperm
Combinatorics.nthperm — Method.

nthperm(a, k)

Compute the kth lexicographic permutation of the vector a.

source
13.21. COMBINATORICS.PARITY 259

13.21 Combinatorics.parity
Combinatorics.parity — Method.

parity(p)

Compute the parity of a permutation p using the levicivita function,

permitting calls such as iseven(parity(p)). If p is not a permutation then an
error is thrown.
source

13.22 Combinatorics.partitions
Combinatorics.partitions — Method.

partitions(s::AbstractVector, m::Int)

Generate all set partitions of the elements of an array s into exactly m subsets,
represented as arrays of arrays. Because the number of partitions can be very
large, this function returns an iterator object. Use collect(partitions(s,
m)) to get an array of all partitions. The number of partitions into m subsets is
equal to the Stirling number of the second kind, and can be efficiently computed
using length(partitions(s, m)).
source

13.23 Combinatorics.partitions
Combinatorics.partitions — Method.

partitions(s::AbstractVector)

Generate all set partitions of the elements of an array s, represented as

arrays of arrays. Because the number of partitions can be very large, this
function returns an iterator object. Use collect(partitions(s)) to get an
array of all partitions. The number of partitions to generate can be efficiently
computed using length(partitions(s)).
source

13.24 Combinatorics.partitions
Combinatorics.partitions — Method.

partitions(n, m)
260 CHAPTER 13. COMBINATORICS

Generate all arrays of m integers that sum to n. Because the number of

partitions can be very large, this function returns an iterator object. Use
collect(partitions(n, m)) to get an array of all partitions. The number of
partitions to generate can be efficiently computed using length(partitions(n,
m)).
source

13.25 Combinatorics.partitions
Combinatorics.partitions — Method.
partitions(n)
Generate all integer arrays that sum to n. Because the number of partitions
can be very large, this function returns an iterator object. Use collect(partitions(n))
to get an array of all partitions. The number of partitions to generate can be
efficiently computed using length(partitions(n)).
source

13.26 Combinatorics.partitionsequence
Combinatorics.partitionsequence — Method.
partitionsequence(lambda::Partition)
Compute essential part of the partition sequence of lambda.
source

13.27 Combinatorics.permutations
Combinatorics.permutations — Method.
permutations(a, t)
Generate all size t permutations of an indexable object a.
source

13.28 Combinatorics.permutations
Combinatorics.permutations — Method.
permutations(a)
Generate all permutations of an indexable object a. Because the number of
permutations can be very large, this function returns an iterator object. Use
collect(permutations(a)) to get an array of all permutations.
source
13.29. COMBINATORICS.POWERSET 261

13.29 Combinatorics.powerset
Combinatorics.powerset — Function.

powerset(a, min=0, max=length(a))

Generate all subsets of an indexable object a including the empty set, with
cardinality bounded by min and max. Because the number of subsets can be
very large, this function returns an iterator object. Use collect(powerset(a,
min, max)) to get an array of all subsets.
source

13.30 Combinatorics.prevprod
Combinatorics.prevprod — Method.

prevprod(a::Vector{Int}, x)

Previous integer not greater than x that can be written as kipi for integers
Q
p1 , p2 , etc.
For integers i1 , i2 , i3 , this is equivalent to finding the largest x such that

in1 1 in2 2 in3 3 ≤ x

for integers n1 , n2 , n3 .
source

13.31 Combinatorics.with replacement combinations

Combinatorics.with replacement combinations — Method.

with_replacement_combinations(a, t)

Generate all combinations with replacement of size t from an array a.

source
Chapter 14

JuliaDB

14.1 Dagger.compute
Dagger.compute — Method.

compute(t::DNDSparse; allowoverlap, closed)

Computes any delayed-evaluations in the DNDSparse. The computed data

is left on the worker processes. Subsequent operations on the results will reuse
the chunks.
If allowoverlap is false then the computed data is re-sorted if required to
have no chunks with overlapping index ranges if necessary.
If closed is true then the computed data is re-sorted if required to have no
chunks with overlapping OR continuous boundaries.
See also collect.
!!! warning compute(t) requires at least as much memory as the size of
the result of the computing t. You usually don’t need to do this for the whole
dataset. If the result is expected to be big, try compute(save(t, "output dir"))
instead. See save for more.
source

14.2 Dagger.distribute
Dagger.distribute — Function.

distribute(itable::NDSparse, nchunks::Int=nworkers())

Distributes an NDSparse object into a DNDSparse of nchunks chunks of

approximately equal size.
Returns a DNDSparse.
source

262
14.3. DAGGER.DISTRIBUTE 263

14.3 Dagger.distribute
Dagger.distribute — Method.

distribute(t::Table, chunks)

Distribute a table in chunks pieces. Equivalent to table(t, chunks=chunks).

source

14.4 Dagger.distribute
Dagger.distribute — Method.

distribute(itable::NDSparse, rowgroups::AbstractArray)

Distributes an NDSparse object into a DNDSparse by splitting it up into

chunks of rowgroups elements. rowgroups is a vector specifying the number of
rows in the chunks.
Returns a DNDSparse.
source

14.5 Dagger.load
Dagger.load — Method.
load(dir::AbstractString; tomemory)
Load a saved DNDSparse from dir directory. Data can be saved using the
save function.
source

14.6 Dagger.save
Dagger.save — Method.
save(t::Union{DNDSparse, DNDSparse}, outputdir::AbstractString)
Saves a distributed dataset to disk. Saved data can be loaded with load.
source

14.7 IndexedTables.convertdim
IndexedTables.convertdim — Method.

convertdim(x::DNDSparse, d::DimName, xlate; agg::Function, name)

264 CHAPTER 14. JULIADB

Apply function or dictionary xlate to each index in the specified dimension.

If the mapping is many-to-one, agg is used to aggregate the results. name op-
tionally specifies a name for the new dimension. xlate must be a monotonically
increasing function.
See also reducedim and aggregate
source

14.8 IndexedTables.leftjoin
IndexedTables.leftjoin — Method.

leftjoin(left::DNDSparse, right::DNDSparse, [op::Function])

Keeps only rows with indices in left. If rows of the same index are present
in right, then they are combined using op. op by default picks the value from
right.
source

14.9 IndexedTables.naturaljoin
IndexedTables.naturaljoin — Method.

naturaljoin(op, left::DNDSparse, right::DNDSparse, ascolumns=false)

Returns a new DNDSparse containing only rows where the indices are present
both in left AND right tables. The data columns are concatenated. The data
of the matching rows from left and right are combined using op. If op returns
a tuple or NamedTuple, and ascolumns is set to true, the output table will
contain the tuple elements as separate data columns instead as a single column
of resultant tuples.
source

14.10 IndexedTables.naturaljoin
IndexedTables.naturaljoin — Method.

naturaljoin(left::DNDSparse, right::DNDSparse, [op])

Returns a new DNDSparse containing only rows where the indices are present
both in left AND right tables. The data columns are concatenated.
source
14.11. INDEXEDTABLES.REDUCEDIM VEC 265

14.11 IndexedTables.reducedim vec

IndexedTables.reducedim vec — Method.

reducedim_vec(f::Function, t::DNDSparse, dims)

Like reducedim, except uses a function mapping a vector of values to a scalar

instead of a 2-argument scalar function.
See also reducedim and aggregate vec.
source

14.12 JuliaDB.loadndsparse
JuliaDB.loadndsparse — Method.
loadndsparse(files::Union{AbstractVector,String}; <options>)
Load an NDSparse from CSV files.
files is either a vector of file paths, or a directory name.
Options:

• indexcols::Vector – columns to use as indexed columns. (by default a

1:n implicit index is used.)

• datacols::Vector – non-indexed columns. (defaults to all columns but

indexed columns). Specify this to only load a subset of columns. In place
of the name of a column, you can specify a tuple of names – this will treat
any column with one of those names as the same column, but use the first
name in the tuple. This is useful when the same column changes name
between CSV files. (e.g. vendor id and VendorId)

All other options are identical to those in loadtable

source

14.13 JuliaDB.loadtable
JuliaDB.loadtable — Method.
loadtable(files::Union{AbstractVector,String}; <options>)
Load a table from CSV files.
files is either a vector of file paths, or a directory name.
Options:

• output::AbstractString – directory name to write the table to. By

default data is loaded directly to memory. Specifying this option will
allow you to load data larger than the available memory.

• indexcols::Vector – columns to use as primary key columns. (defaults

to [])
266 CHAPTER 14. JULIADB

• datacols::Vector – non-indexed columns. (defaults to all columns but

indexed columns). Specify this to only load a subset of columns. In place
of the name of a column, you can specify a tuple of names – this will treat
any column with one of those names as the same column, but use the first
name in the tuple. This is useful when the same column changes name
between CSV files. (e.g. vendor id and VendorId)
• distributed::Bool – should the output dataset be loaded as a dis-
tributed table? If true, this will use all available worker processes to
load the data. (defaults to true if workers are available, false if not)
• chunks::Int – number of chunks to create when loading distributed. (de-
faults to number of workers)
• delim::Char – the delimiter character. (defaults to ,). Use spacedelim=true
to split by spaces.
• spacedelim::Bool: parse space-delimited files. delim has no effect if
true.
• quotechar::Char – quote character. (defaults to ")
• escapechar::Char – escape character. (defaults to ")
• filenamecol::Union{Symbol, Pair} – create a column containing the
file names from where each row came from. This argument gives a name
to the column. By default, basename(name) of the name is kept, and
“.csv” suffix will be stripped. To provide a custom function to apply
on the names, use a name => Function pair. By default, no file name
column will be created.
• header exists::Bool – does header exist in the files? (defaults to true)
• colnames::Vector{String} – specify column names for the files, use this
with (header exists=false, otherwise first row is discarded). By default
column names are assumed to be present in the file.
• samecols – a vector of tuples of strings where each tuple contains alterna-
tive names for the same column. For example, if some files have the name
“vendor id” and others have the name “VendorID”, pass samecols=[("VendorID",
"vendor id")].
• colparsers – either a vector or dictionary of data types or an AbstractToken
object from TextParse package. By default, these are inferred automati-
cally. See type detect rows option below.
• type detect rows: number of rows to use to infer the initial colparsers
defaults to 20.
• nastrings::Vector{String} – strings that are to be considered NA. (de-
faults to TextParse.NA STRINGS)
14.14. JULIADB.PARTITIONPLOT 267

• skiplines begin::Char – skip some lines in the beginning of each file.

(doesn’t skip by default)

• usecache::Bool: (vestigial)

source

14.14 JuliaDB.partitionplot
JuliaDB.partitionplot — Function.

partitionplot(table, y; stat=Extrema(), nparts=100, by=nothing, dropmissing=false)

partitionplot(table, x, y; stat=Extrema(), nparts=100, by=nothing, dropmissing=false)

Plot a summary of variable y against x (1:length(y) if not specified). Using

nparts approximately-equal sections along the x-axis, the data in y over each
section is summarized by stat.
source

14.15 JuliaDB.rechunk
JuliaDB.rechunk — Function.
rechunk(t::Union{DNDSparse, DNDSparse}[, by[, select]]; <options>)
Reindex and sort a distributed dataset by keys selected by by.
Optionally select specifies which non-indexed fields are kept. By default
this is all fields not mentioned in by for Table and the value columns for
NDSparse.
Options:

• chunks – how to distribute the data. This can be:

1. An integer – number of chunks to create

2. An vector of k integers – number of elements in each of the k chunks.
sum(k) must be same as length(t)
3. The distribution of another array. i.e. vec.subdomains where vec
is a distributed array.

• merge::Function – a function which merges two sub-table or sub-ndsparse

into one NDSparse. They may have overlaps in their indices.

• splitters::AbstractVector – specify keys to split by. To create n

chunks you would need to pass n-1 splitters and also the chunks=n option.

• chunks sorted::Bool – are the chunks sorted locally? If true, this skips
sorting or re-indexing them.
268 CHAPTER 14. JULIADB

• affinities::Vector{<:Integer} – which processes (Int pid) should each

output chunk be created on. If unspecified all workers are used.
• closed::Bool – if true, the same key will not be present in multiple
chunks (although sorted). true by default.

• nsamples::Integer – number of keys to randomly sample from each

chunk to estimate splitters in the sorting process. (See samplesort). De-
faults to 2000.
• batchsize::Integer – how many chunks at a time from the input should
be loaded into memory at any given time. This will essentially sort in
batches of batchsize chunks.

source

14.16 JuliaDB.tracktime
JuliaDB.tracktime — Method.
tracktime(f)
Track the time spent on different processes in different categories in running
f.
source
Chapter 15

HypothesisTests

15.1 HypothesisTests.ChisqTest
HypothesisTests.ChisqTest — Method.

ChisqTest(x[, y][, theta0 = ones(length(x))/length(x)])

Perform a PowerDivergenceTest with = 1, i.e. in the form of Pearson’s

chi-squared statistic.
If y is not given and x is a matrix with one row or column, or x is a vector,
then a goodness-of-fit test is performed (x is treated as a one-dimensional con-
tingency table). In this case, the hypothesis tested is whether the population
probabilities equal those in theta0, or are all equal if theta0 is not given.
If x is a matrix with at least two rows and columns, it is taken as a two-
dimensional contingency table. Otherwise, x and y must be vectors of the same
length. The contingency table is calculated using counts function from the
StatsBase package. Then the power divergence test is conducted under the
null hypothesis that the joint distribution of the cell counts in a 2-dimensional
contingency table is the product of the row and column marginals.
Note that the entries of x (and y if provided) must be non-negative integers.
Implements: pvalue, confint
source

15.2 HypothesisTests.MannWhitneyUTest
HypothesisTests.MannWhitneyUTest — Method.

MannWhitneyUTest(x::AbstractVector{<:Real}, y::AbstractVector{<:Real})

Perform a Mann-Whitney U test of the null hypothesis that the probability

that an observation drawn from the same population as x is greater than an ob-
servation drawn from the same population as y is equal to the probability that

269
270 CHAPTER 15. HYPOTHESISTESTS

an observation drawn from the same population as y is greater than an obser-

vation drawn from the same population as x against the alternative hypothesis
that these probabilities are not equal.
The Mann-Whitney U test is sometimes known as the Wilcoxon rank-sum
test.
When there are no tied ranks and 50 samples, or tied ranks and 10 samples,
MannWhitneyUTest performs an exact Mann-Whitney U test. In all other cases,
MannWhitneyUTest performs an approximate Mann-Whitney U test. Behavior
may be further controlled by using ExactMannWhitneyUTest or ApproximateMannWhitneyUTest
directly.
Implements: pvalue
source

15.3 HypothesisTests.MultinomialLRT
HypothesisTests.MultinomialLRT — Method.

MultinomialLRT(x[, y][, theta0 = ones(length(x))/length(x)])

Perform a PowerDivergenceTest with = 0, i.e. in the form of the likelihood

ratio test statistic.
If y is not given and x is a matrix with one row or column, or x is a vector,
then a goodness-of-fit test is performed (x is treated as a one-dimensional con-
tingency table). In this case, the hypothesis tested is whether the population
probabilities equal those in theta0, or are all equal if theta0 is not given.
If x is a matrix with at least two rows and columns, it is taken as a two-
dimensional contingency table. Otherwise, x and y must be vectors of the same
length. The contingency table is calculated using counts function from the
StatsBase package. Then the power divergence test is conducted under the
null hypothesis that the joint distribution of the cell counts in a 2-dimensional
contingency table is the product of the row and column marginals.
Note that the entries of x (and y if provided) must be non-negative integers.
Implements: pvalue, confint
source

15.4 HypothesisTests.SignedRankTest
HypothesisTests.SignedRankTest — Method.

SignedRankTest(x::AbstractVector{<:Real})
SignedRankTest(x::AbstractVector{<:Real}, y::AbstractVector{T<:Real})

Perform a Wilcoxon signed rank test of the null hypothesis that the distri-
bution of x (or the difference x - y if y is provided) has zero median against
the alternative hypothesis that the median is non-zero.
15.5. HYPOTHESISTESTS.PVALUE 271

When there are no tied ranks and 50 samples, or tied ranks and 15 sam-
ples, SignedRankTest performs an exact signed rank test. In all other cases,
SignedRankTest performs an approximate signed rank test. Behavior may be
further controlled by using ExactSignedRankTest or ApproximateSignedRankTest
directly.
Implements: pvalue, confint
source

15.5 HypothesisTests.pvalue
HypothesisTests.pvalue — Function.

pvalue(test::HypothesisTest; tail = :both)

Compute the p-value for a given significance test.

If tail is :both (default), then the p-value for the two-sided test is returned.
If tail is :left or :right, then a one-sided test is performed.
source

15.6 HypothesisTests.pvalue
HypothesisTests.pvalue — Method.

pvalue(x::FisherExactTest; tail = :both, method = :central)

Compute the p-value for a given Fisher exact test.

The one-sided p-values are based on Fisher’s non-central hypergeometric
distribution f(i) with odds ratio :
X
p (left) = f(i)
ia
X
(right)
p = f(i)
ia

For tail = :both, possible values for method are:

• :central (default): Central interval, i.e. the p-value is two times the
minimum of the one-sided p-values.
• :minlike: Minimum likelihood interval, i.e. the p-value is computed by
summing all tables with the same marginals that are equally or less prob-
able:
X
p= f(i)
f(i)f(a)

References
272 CHAPTER 15. HYPOTHESISTESTS

• Gibbons, J.D., Pratt, J.W., P-values: Interpretation and Methodology,

American Statistican, 29(1):20-25, 1975.

• Fay, M.P., Supplementary material to “Confidence intervals that match

Fisher’s exact or Blaker’s exact tests”. Biostatistics, Volume 11, Issue 2,
1 April 2010, Pages 373–374, link

source

15.7 HypothesisTests.testname
HypothesisTests.testname — Method.

testname(::HypothesisTest)

Returns the string value, e.g. “Binomial test” or “Sign Test”.

source

15.8 StatsBase.confint
StatsBase.confint — Function.

confint(test::HypothesisTest, alpha = 0.05; tail = :both)

Compute a confidence interval C with coverage 1-alpha.

If tail is :both (default), then a two-sided confidence interval is returned.
If tail is :left or :right, then a one-sided confidence interval is returned.
!!! note Most of the implemented confidence intervals are strongly consistent,
that is, the confidence interval with coverage 1-alpha does not contain the test
statistic under h0 if and only if the corresponding test rejects the null hypothesis
h0 : = 0 :

$$
C (x, 1 ) = \{ : p_ (x) > \},
$$

where $p_$ is the [‘pvalue‘](HypothesisTests.md#HypothesisTests.pvalue) of the correspo

source

15.9 StatsBase.confint
StatsBase.confint — Function.

confint(test::PowerDivergenceTest, alpha = 0.05; tail = :both, method = :sison_glaz)

15.10. STATSBASE.CONFINT 273

Compute a confidence interval with coverage 1-alpha for multinomial pro-

portions using one of the following methods. Possible values for method are:

• :sison glaz (default): Sison-Glaz intervals

• :bootstrap: Bootstrap intervals

• :quesenberry hurst: Quesenberry-Hurst intervals

• :gold: Gold intervals (asymptotic simultaneous intervals)

References

• Agresti, Alan. Categorical Data Analysis, 3rd Edition. Wiley, 2013.

• Sison, C.P and Glaz, J. Simultaneous confidence intervals and sample

size determination for multinomial proportions. Journal of the American
Statistical Association, 90:366-369, 1995.

• Quesensberry, C.P. and Hurst, D.C. Large Sample Simultaneous Confi-

dence Intervals for Multinational Proportions. Technometrics, 6:191-195,
1964.

• Gold, R. Z. Tests Auxiliary to 2 Tests in a Markov Chain. Annals of

Mathematical Statistics, 30:56-74, 1963.

source

15.10 StatsBase.confint
StatsBase.confint — Function.

confint(x::FisherExactTest, alpha::Float64=0.05; tail=:both, method=:central)

Compute a confidence interval with coverage 1 - alpha. One-sided inter-

vals are based on Fisher’s non-central hypergeometric distribution. For tail =
:both, the only method implemented yet is the central interval (:central).
!!! note Since the p-value is not necessarily unimodal, the corresponding
confidence region might not be an interval.
References

• Gibbons, J.D, Pratt, J.W. P-values: Interpretation and Methodology,

American Statistican, 29(1):20-25, 1975.

• Fay, M.P., Supplementary material to “Confidence intervals that match

Fisher’s exact or Blaker’s exact tests”. Biostatistics, Volume 11, Issue 2,
1 April 2010, Pages 373–374, link

source
274 CHAPTER 15. HYPOTHESISTESTS

15.11 StatsBase.confint
StatsBase.confint — Function.

confint(test::BinomialTest, alpha = 0.05; tail = :both, method = :clopper_pearson)

Compute a confidence interval with coverage 1-alpha for a binomial propor-

tion using one of the following methods. Possible values for method are:

• :clopper pearson (default): Clopper-Pearson interval is based on the bi-

nomial distribution. The empirical coverage is never less than the nominal
coverage of 1-alpha; it is usually too conservative.

• :wald: Wald (or normal approximation) interval relies on the standard

approximation of the actual binomial distribution by a normal distribu-
tion. Coverage can be erratically poor for success probabilities close to
zero or one.

• :wilson: Wilson score interval relies on a normal approximation. In con-

trast to :wald, the standard deviation is not approximated by an empirical
estimate, resulting in good empirical coverages even for small numbers of
draws and extreme success probabilities.
• :jeffrey: Jeffreys interval is a Bayesian credible interval obtained by
using a non-informative Jeffreys prior. The interval is very similar to the
Wilson interval.
• :agresti coull: Agresti-Coull interval is a simplified version of the Wil-
son interval; both are centered around the same value. The Agresti Coull
interval has higher or equal coverage.

• :arcsine: Confidence interval computed using the arcsine transformation

to make var(p) independent of the probability p.

References

• Brown, L.D., Cai, T.T., and DasGupta, A. Interval estimation for a bino-
mial proportion. Statistical Science, 16(2):101–117, 2001.

External links

• Binomial confidence interval on Wikipedia

source
Chapter 16

PyCall

16.1 PyCall.PyNULL
PyCall.PyNULL — Method.

PyNULL()

Return a PyObject that has a NULL underlying pointer, i.e. it doesn’t actually
refer to any Python object.
This is useful for initializing PyObject global variables and array elements
before an actual Python object is available. For example, you might do const
myglobal = PyNULL() and later on (e.g. in a module init function), reas-
sign myglobal to point to an actual object with copy!(myglobal, someobject).
This procedure will properly handle Python’s reference counting (so that the
Python object will not be freed until you are done with myglobal).
source

16.2 PyCall.PyReverseDims
PyCall.PyReverseDims — Method.

PyReverseDims(array)

Passes a Julia array to Python as a NumPy row-major array (rather than

Julia’s native column-major order) with the dimensions reversed (e.g. a 234 Julia
array is passed as a 432 NumPy row-major array). This is useful for Python
libraries that expect row-major data.
source

16.3 PyCall.PyTextIO
PyCall.PyTextIO — Method.

275
276 CHAPTER 16. PYCALL

PyTextIO(io::IO)
PyObject(io::IO)

Julia IO streams are converted into Python objects implementing the Raw-
IOBase interface, so they can be used for binary I/O in Python
source

16.4 PyCall.ispynull
PyCall.ispynull — Method.

ispynull(o::PyObject)

Test where o contains a NULL pointer to a Python object, i.e. whether it is

equivalent to a PyNULL() object.
source

16.5 PyCall.pybuiltin
PyCall.pybuiltin — Method.

pybuiltin(s::AbstractString)

Look up a string or symbol s among the global Python builtins. If s is

a string it returns a PyObject, while if s is a symbol it returns the builtin
converted to PyAny.
source

16.6 PyCall.pybytes
PyCall.pybytes — Method.

pybytes(b::Union{String,DenseVector{UInt8}})

Convert b to a Python bytes object. This differs from the default PyObject(b)
conversion of String to a Python string (which may fail if b does not con-
tain valid Unicode), or from the default conversion of a Vector{UInt8} to a
bytearray object (which is mutable, unlike bytes).
source

16.7 PyCall.pycall
PyCall.pycall — Method.

pycall(o::Union{PyObject,PyPtr}, returntype::TypeTuple, args...; kwargs...)

16.8. PYCALL.PYEVAL 277

Call the given Python function (typically looked up from a module) with
the given args. . . (of standard Julia types which are converted automatically
to the corresponding Python types if possible), converting the return value to
returntype (use a returntype of PyObject to return the unconverted Python
object reference, or of PyAny to request an automated conversion)
source

16.8 PyCall.pyeval
PyCall.pyeval — Function.

pyeval(s::AbstractString, returntype::TypeTuple=PyAny, locals=PyDict{AbstractString, PyObject}(),

input_type=Py_eval_input; kwargs...)

This evaluates s as a Python string and returns the result converted to rtype
(which defaults to PyAny). The remaining arguments are keywords that define
local variables to be used in the expression.
For example, pyeval("x + y", x=1, y=2) returns 3.
source

16.9 PyCall.pyfunction
PyCall.pyfunction — Method.

pyfunction(f, argtypes...; kwtypes...)

Create a Python object that wraps around the Julia function (or callable
object) f. Unlike PyObject(f), this allows you to specify the argument types
that the Julia function expects — giving you more control and potentially better
performance.
kwtypes... should be a set of somekeyword=SomeType arguments giving
the desired Julia type for each keyword somekeyword. Unspecified keywords are
converted to PyAny (i.e. auto-detected) by default.
The return value ret = f(...) is still converted back to a Python object by
PyObject(ret). If you want a different return-type conversion than the default
of PyObject(ret), you can instead call pyfunctionret.
source

16.10 PyCall.pyfunctionret
PyCall.pyfunctionret — Method.

pyfunctionret(f, returntype, argtypes...; kwtypes...)

278 CHAPTER 16. PYCALL

Like pyfunction, but also lets you specify the returntype for conversion
back to Python. In particular, if ret = f(...) is the return value of f, then
it is converted to Python via PyObject(returntype(ret)).
If returntype is Any, then ret is not converted to a “native” Python type
at all, and is instead converted to a “wrapped” Julia object in Python. If
returntype is nothing, then the return value is discarded and nothing is re-
turned to Python.
source

16.11 PyCall.pygui
PyCall.pygui — Method.

pygui()

Return the current GUI toolkit as a symbol.

source

16.12 PyCall.pygui start

PyCall.pygui start — Function.

pygui_start(gui::Symbol = pygui())

Start the event loop of a certain toolkit.

The argument gui defaults to the current default GUI, but it could be :wx,
:gtk, :gtk3, :tk, or :qt.
source

16.13 PyCall.pygui stop

PyCall.pygui stop — Function.

pygui_stop(gui::Symbol = pygui())

Stop any running event loop for gui. The gui argument defaults to current
default GUI.
source

16.14 PyCall.pyimport
PyCall.pyimport — Method.

pyimport(s::AbstractString)
16.15. PYCALL.PYIMPORT CONDA 279

Import the Python module s (a string or symbol) and return a pointer to it

(a PyObject). Functions or other symbols in the module may then be looked
up by s[name] where name is a string (for the raw PyObject) or symbol (for
automatic type-conversion). Unlike the @pyimport macro, this does not define
a Julia module and members cannot be accessed with s.name
source

16.15 PyCall.pyimport conda

PyCall.pyimport conda — Function.

pyimport_conda(modulename, condapkg, [channel])

Returns the result of pyimport(modulename) if possible. If the module is not

found, and PyCall is configured to use the Conda Python distro (via the Julia
Conda package), then automatically install condapkg via Conda.add(condapkg)
and then re-try the pyimport. Other Anaconda-based Python installations are
also supported as long as their conda program is functioning.
If PyCall is not using Conda and the pyimport fails, throws an exception
with an error message telling the user how to configure PyCall to use Conda for
automated installation of the module.
The third argument, channel is an optional Anaconda “channel” to use for
installing the package; this is useful for packages that are not included in the
default Anaconda package listing.
source

16.16 PyCall.pyimport e
PyCall.pyimport e — Method.

pyimport_e(s::AbstractString)

Like pyimport(s), but returns an ispynull(o) == true object if the mod-

ule cannot be imported rather than throwing an error.
source

16.17 PyCall.pyrepr
PyCall.pyrepr — Method.

pyrepr(o::PyObject)

Return a string representation of o corresponding to repr(o) in Python.

source
280 CHAPTER 16. PYCALL

16.18 PyCall.pystr
PyCall.pystr — Method.

pystr(o::PyObject)

Return a string representation of o corresponding to str(o) in Python.

source

16.19 PyCall.pystring
PyCall.pystring — Method.

pystring(o::PyObject)

Return a string representation of o. Normally, this corresponds to repr(o)

in Python, but unlike repr it should never fail (falling back to str or to printing
the raw pointer if necessary).
source

16.20 PyCall.pytype mapping

PyCall.pytype mapping — Method.

pytype_mapping(pytype, jltype)

Given a Python type object pytype, tell PyCall to convert it to jltype in

PyAny(object) conversions.
source

16.21 PyCall.pytype query

PyCall.pytype query — Function.

pytype_query(o::PyObject, default=PyObject)

Given a Python object o, return the corresponding native Julia type (de-
faulting to default) that we convert o to in PyAny(o) conversions.
source
16.22. PYCALL.PYWRAP 281

16.22 PyCall.pywrap
PyCall.pywrap — Function.

pywrap(o::PyObject)

This returns a wrapper w that is an anonymous module which provides (read)

access to converted versions of o’s members as w.member.
For example, @pyimport module as name is equivalent to const name =
pywrap(pyimport("module"))
If the Python module contains identifiers that are reserved words in Julia
(e.g. function), they cannot be accessed as w.member; one must instead use
w.pymember(:member) (for the PyAny conversion) or w.pymember(“member”)
(for the raw PyObject).
source
Chapter 17

CategoricalArrays

17.1 CategoricalArrays.categorical
CategoricalArrays.categorical — Function.

categorical{T}(A::AbstractArray{T}[, compress::Bool]; ordered::Bool=false)

Construct a categorical array with the values from A.

If the element type supports it, levels are sorted in ascending order; else,
they are kept in their order of appearance in A. The ordered keyword argument
determines whether the array values can be compared according to the ordering
of levels or not (see isordered).
If compress is provided and set to true, the smallest reference type able
to hold the number of unique values in A will be used. While this will reduce
memory use, passing this parameter will also introduce a type instability which
can affect performance inside the function where the call is made. Therefore,
use this option with caution (the one-argument version does not suffer from this
problem).

categorical{T}(A::CategoricalArray{T}[, compress::Bool]; ordered::Bool=isordered(A))

If A is already a CategoricalArray, its levels are preserved; the same applies

to the ordered property, and to the reference type unless compress is passed.
source

17.2 CategoricalArrays.compress
CategoricalArrays.compress — Method.

compress(A::CategoricalArray)

282
17.3. CATEGORICALARRAYS.CUT 283

Return a copy of categorical array A using the smallest reference type able
to hold the number of levels of A.
While this will reduce memory use, this function is type-unstable, which can
affect performance inside the function where the call is made. Therefore, use it
with caution.
source

17.3 CategoricalArrays.cut
CategoricalArrays.cut — Method.

cut(x::AbstractArray, ngroups::Integer;
labels::AbstractVector=String[])

Cut a numeric array into ngroups quantiles, determined using quantile.

source

17.4 CategoricalArrays.cut
CategoricalArrays.cut — Method.

cut(x::AbstractArray, breaks::AbstractVector;
extend::Bool=false, labels::AbstractVector=[], allow_missing::Bool=false)

Cut a numeric array into intervals and return an ordered CategoricalArray

indicating the interval into which each entry falls. Intervals are of the form
[lower, upper), i.e. the lower bound is included and the upper bound is ex-
cluded.
If x accepts missing values (i.e. eltype(x) >: Missing) the returned array
will also accept them.
Arguments

• extend::Bool=false: when false, an error is raised if some values in x

fall outside of the breaks; when true, breaks are automatically added to
include all values in x, and the upper bound is included in the last interval.

• labels::AbstractVector=[]: a vector of strings giving the names to use

for the intervals; if empty, default labels are used.

• allow missing::Bool=true: when true, values outside of breaks result

in missing values. only supported when x accepts missing values.

source
284 CHAPTER 17. CATEGORICALARRAYS

17.5 CategoricalArrays.decompress
CategoricalArrays.decompress — Method.
decompress(A::CategoricalArray)
Return a copy of categorical array A using the default reference type (UInt32).
If A is using a small reference type (such as UInt8 or UInt16) the decompressed
array will have room for more levels.
To avoid the need to call decompress, ensure compress is not called when
creating the categorical array.
source

17.6 CategoricalArrays.isordered
CategoricalArrays.isordered — Method.
isordered(A::CategoricalArray)
Test whether entries in A can be compared using <, > and similar operators,
using the ordering of levels.
source

17.7 CategoricalArrays.levels!
CategoricalArrays.levels! — Method.
levels!(A::CategoricalArray, newlevels::Vector; allow_missing::Bool=false)
Set the levels categorical array A. The order of appearance of levels will be
respected by levels, which may affect display of results in some operations; if
A is ordered (see isordered), it will also be used for order comparisons using <,
> and similar operators. Reordering levels will never affect the values of entries
in the array.
If A accepts missing values (i.e. eltype(A) >: Missing) and allow missing=true,
entries corresponding to omitted levels will be set to missing. Else, newlevels
must include all levels which appear in the data.
source

17.8 CategoricalArrays.ordered!
CategoricalArrays.ordered! — Method.
ordered!(A::CategoricalArray, ordered::Bool)
Set whether entries in A can be compared using <, > and similar operators,
using the ordering of levels. Return the modified A.
source
17.9. CATEGORICALARRAYS.RECODE 285

17.9 CategoricalArrays.recode
CategoricalArrays.recode — Function.

recode(a::AbstractArray[, default::Any], pairs::Pair...)

Return a copy of a, replacing elements matching a key of pairs with the

corresponding value. The type of the array is chosen so that it can hold all
recoded elements (but not necessarily original elements from a).
For each Pair in pairs, if the element is equal to (according to isequal) or
in the key (first item of the pair), then the corresponding value (second item) is
used. If the element matches no key and default is not provided or nothing, it
is copied as-is; if default is specified, it is used in place of the original element.
If an element matches more than one key, the first match is used.

recode(a::CategoricalArray[, default::Any], pairs::Pair...)

If a is a CategoricalArray then the ordering of resulting levels is determined

by the order of passed pairs and default will be the last level if provided.
Examples

julia> using CategoricalArrays

julia> recode(1:10, 1=>100, 2:4=>0, [5; 9:10]=>-1)

10-element Array{Int64,1}:
100
0
0
0
-1
6
7
8
-1
-1

recode(a::AbstractArray{>:Missing}[, default::Any], pairs::Pair...)

If a contains missing values, they are never replaced with default: use
missing in a pair to recode them. If that’s not the case, the returned array will
accept missing values.
Examples

julia> using CategoricalArrays, Missings

julia> recode(1:10, 1=>100, 2:4=>0, [5; 9:10]=>-1, 6=>missing)

10-element Array{Union{Int64, Missings.Missing},1}:
100
286 CHAPTER 17. CATEGORICALARRAYS

0
0
0
-1
missing
7
8
-1
-1

source

17.10 CategoricalArrays.recode!
CategoricalArrays.recode! — Function.

recode!(dest::AbstractArray, src::AbstractArray[, default::Any], pairs::Pair...)

Fill dest with elements from src, replacing those matching a key of pairs
with the corresponding value.
For each Pair in pairs, if the element is equal to (according to isequal))
the key (first item of the pair) or to one of its entries if it is a collection, then
the corresponding value (second item) is copied to dest. If the element matches
no key and default is not provided or nothing, it is copied as-is; if default
is specified, it is used in place of the original element. dest and src must be
of the same length, but not necessarily of the same type. Elements of src as
well as values from pairs will be converted when possible on assignment. If
an element matches more than one key, the first match is used.

recode!(dest::CategoricalArray, src::AbstractArray[, default::Any], pairs::Pair...)

If dest is a CategoricalArray then the ordering of resulting levels is de-

termined by the order of passed pairs and default will be the last level if
provided.

recode!(dest::AbstractArray, src::AbstractArray{>:Missing}[, default::Any], pairs::Pair

If src contains missing values, they are never replaced with default: use
missing in a pair to recode them.
source

17.11 CategoricalArrays.recode!
CategoricalArrays.recode! — Method.

recode!(a::AbstractArray[, default::Any], pairs::Pair...)

17.12. MISSINGS.LEVELS 287

Convenience function for in-place recoding, equivalent to recode!(a, a,

...).
Examples

julia> using CategoricalArrays

julia> x = collect(1:10);

julia> recode!(x, 1=>100, 2:4=>0, [5; 9:10]=>-1);

julia> x
10-element Array{Int64,1}:
100
0
0
0
-1
6
7
8
-1
-1

source

17.12 Missings.levels
Missings.levels — Method.

levels(A::CategoricalArray)

Return the levels of categorical array A. This may include levels which do
not actually appear in the data (see droplevels!).
source
Chapter 18

GLM

18.1 GLM.canonicallink
GLM.canonicallink — Function.

canonicallink(D::Distribution)

Return the canonical link for distribution D, which must be in the exponential
family.
Examples

julia> canonicallink(Bernoulli())
GLM.LogitLink()

source

18.2 GLM.delbeta!
GLM.delbeta! — Function.

delbeta!(p::LinPred, r::Vector)

Evaluate and return p.delbeta the increment to the coefficient vector from
residual r
source

18.3 GLM.devresid
GLM.devresid — Function.

devresid(D, y, )

288
18.4. GLM.FTEST 289

Return the squared deviance residual of from y for distribution D

The deviance of a GLM can be evaluated as the sum of the squared de-
viance residuals. This is the principal use for these values. The actual deviance
residual, say for plotting, is the signed square root of this value
[] sign(y - ) * sqrt(devresid(D, y, ))
Examples

julia> showcompact(devresid(Normal(), 0, 0.25)) # abs2(y - )

0.0625
julia> showcompact(devresid(Bernoulli(), 1, 0.75)) # -2log() when y == 1
0.575364
julia> showcompact(devresid(Bernoulli(), 0, 0.25)) # -2log1p(-) = -2log(1-) when y == 0
0.575364

source

18.4 GLM.ftest
GLM.ftest — Method.

ftest(mod::LinearModel...; atol=0::Real)

For each sequential pair of linear predictors in mod, perform an F-test to

determine if the first one fits significantly better than the next.
A table is returned containing residual degrees of freedom (DOF), degrees of
freedom, difference in DOF from the preceding model, sum of squared residuals
(SSR), difference in SSR from the preceding model, R, difference in R from the
preceding model, and F-statistic and p-value for the comparison between the
two models.
!!! note This function can be used to perform an ANOVA by testing the
relative fit of two models to the data
Optional keyword argument atol controls the numerical tolerance when test-
ing whether the models are nested.
Examples
Suppose we want to compare the effects of two or more treatments on some
result. Because this is an ANOVA, our null hypothesis is that Result ~ 1 fits
the data as well as Result ~ 1 + Treatment.

julia> dat = DataFrame(Treatment=[1, 1, 1, 2, 2, 2, 1, 1, 1, 2, 2, 2.],

Result=[1.1, 1.2, 1, 2.2, 1.9, 2, .9, 1, 1, 2.2, 2, 2],
Other=[1, 1, 2, 1, 2, 1, 3, 1, 1, 2, 2, 1]);

julia> mod = lm(@formula(Result ~ 1 + Treatment), dat);

julia> nullmod = lm(@formula(Result ~ 1), dat);

290 CHAPTER 18. GLM

julia> bigmod = lm(@formula(Result ~ 1 + Treatment + Other), dat);

julia> ft = ftest(mod.model, nullmod.model)

Res. DOF DOF DOF SSR SSR R R F* p(>F)
Model 1 10 3 0.1283 0.9603
Model 2 11 2 -1 3.2292 -3.1008 0.0000 0.9603 241.6234 <1e-7

julia> ftest(bigmod.model, mod.model, nullmod.model)

Res. DOF DOF DOF SSR SSR R R F* p(>F)
Model 1 9 4 0.1038 0.9678
Model 2 10 3 -1 0.1283 -0.0245 0.9603 0.0076 2.1236 0.1790
Model 3 11 2 -1 3.2292 -3.1008 0.0000 0.9603 241.6234 <1e-7

source

18.5 GLM.glmvar
GLM.glmvar — Function.

glmvar(D::Distribution, )

Return the value of the variance function for D at

The variance of D at is the product of the dispersion parameter, , which
does not depend on and the value of glmvar. In other words glmvar returns
the factor of the variance that depends on .
Examples

julia> = inv(6):inv(3):1; showcompact(collect())

[0.166667, 0.5, 0.833333]
julia> showcompact(glmvar.(Normal(), )) # constant for Normal()
[1.0, 1.0, 1.0]
julia> showcompact(glmvar.(Bernoulli(), )) # * (1 - ) for Bernoulli()
[0.138889, 0.25, 0.138889]
julia> showcompact(glmvar.(Poisson(), )) # for Poisson()
[0.166667, 0.5, 0.833333]

source

18.6 GLM.inverselink
GLM.inverselink — Function.

inverselink(L::Link, )
18.7. GLM.LINKFUN 291

Return a 3-tuple of the inverse link, the derivative of the inverse link, and
when appropriate, the variance function *(1 - ).
The variance function is returned as NaN unless the range of is (0, 1)
Examples

julia> showcompact(inverselink(LogitLink(), 0.0))

(0.5, 0.25, 0.25)
julia> showcompact(inverselink(CloglogLink(), 0.0))
(0.632121, 0.367879, 0.232544)
julia> showcompact(inverselink(LogLink(), 2.0))
(7.38906, 7.38906, NaN)

source

18.7 GLM.linkfun
GLM.linkfun — Function.

linkfun(L::Link, )

Return , the value of the linear predictor for link L at mean .

Examples

julia> = inv(10):inv(5):1
0.1:0.2:0.9

julia> show(linkfun.(LogitLink(), ))
[-2.19722, -0.847298, 0.0, 0.847298, 2.19722]

source

18.8 GLM.linkinv
GLM.linkinv — Function.

linkinv(L::Link, )

Return , the mean value, for link L at linear predictor value .

Examples

julia> = inv(10):inv(5):1; showcompact(collect())

[0.1, 0.3, 0.5, 0.7, 0.9]
julia> = logit.(); showcompact()
[-2.19722, -0.847298, 0.0, 0.847298, 2.19722]
julia> showcompact(linkinv.(LogitLink(), ))
[0.1, 0.3, 0.5, 0.7, 0.9]

source
292 CHAPTER 18. GLM

18.9 GLM.linpred
GLM.linpred — Function.
linpred(p::LinPred, f::Read=1.0)
Return the linear predictor p.X * (p.beta0 .+ f * p.delbeta)
source

18.10 GLM.linpred!
GLM.linpred! — Function.
linpred!(out, p::LinPred, f::Real=1.0)
Overwrite out with the linear predictor from p with factor f
The effective coefficient vector, p.scratchbeta, is evaluated as p.beta0 .+
f * p.delbeta, and out is updated to p.X * p.scratchbeta
source

18.11 GLM.lm
GLM.lm — Function.
lm(X, y, allowrankdeficient::Bool=false)
An alias for fit(LinearModel, X, y, allowrankdeficient)
The arguments X and y can be a Matrix and a Vector or a Formula and a
DataFrame.
source

18.12 GLM.mueta
GLM.mueta — Function.
mueta(L::Link, )
Return the derivative of linkinv, d/d, for link L at linear predictor value .
Examples
julia> showcompact(mueta(LogitLink(), 0.0))
0.25
julia> showcompact(mueta(CloglogLink(), 0.0))
0.367879
julia> showcompact(mueta(LogLink(), 2.0))
7.38906
source
18.13. GLM.MUSTART 293

18.13 GLM.mustart
GLM.mustart — Function.
mustart(D::Distribution, y, wt)
Return a starting value for .
For some distributions it is appropriate to set = y to initialize the IRLS
algorithm but for others, notably the Bernoulli, the values of y are not allowed
as values of and must be modified.
Examples
julia> showcompact(mustart(Bernoulli(), 0.0, 1))
0.25
julia> showcompact(mustart(Bernoulli(), 1.0, 1))
0.75
julia> showcompact(mustart(Binomial(), 0.0, 10))
0.0454545
julia> showcompact(mustart(Normal(), 0.0, 1))
0.0
source

18.14 GLM.update!
GLM.update! — Function.
update!{T<:FPVector}(r::GlmResp{T}, linPr::T)
Update the mean, working weights and working residuals, in r given a value
of the linear predictor, linPr.
source

18.15 GLM.wrkresp
GLM.wrkresp — Method.
wrkresp(r::GlmResp)
The working response, r.eta + r.wrkresid - r.offset.
source

18.16 StatsBase.deviance
StatsBase.deviance — Method.
deviance(obj::LinearModel)
For linear models, the deviance is equal to the residual sum of squares (RSS).
source
294 CHAPTER 18. GLM

18.17 StatsBase.nobs
StatsBase.nobs — Method.

nobs(obj::LinearModel)
nobs(obj::GLM)

For linear and generalized linear models, returns the number of rows, or,
when prior weights are specified, the sum of weights.
source

18.18 StatsBase.nulldeviance
StatsBase.nulldeviance — Method.

nulldeviance(obj::LinearModel)

For linear models, the deviance of the null model is equal to the total sum
of squares (TSS).
source

18.19 StatsBase.predict
StatsBase.predict — Function.

predict(mm::LinearModel, newx::AbstractMatrix, interval_type::Symbol, level::Real = 0.9

Specifying interval type will return a 3-column matrix with the prediction
and the lower and upper confidence bounds for a given level (0.95 equates
alpha = 0.05). Valid values of interval type are :confint delimiting the
uncertainty of the predicted relationship, and :predint delimiting estimated
bounds for new data points.
source

18.20 StatsBase.predict
StatsBase.predict — Method.

predict(mm::AbstractGLM, newX::AbstractMatrix; offset::FPVector=Vector{eltype(newX)}(0)

Form the predicted response of model mm from covariate values newX and,
optionally, an offset.
source
Chapter 19

Roots

19.1 Roots.D
Roots.D — Function.
Take derivative of order k of a function.
Arguments:

• f::Function: a mathematical function from R to R.

• k::Int=1: A non-negative integer specifying the order of the derivative.

Values larger than 8 can be slow to compute.

Wrapper around derivative function in ForwardDiff

source

19.2 Roots.find zero

Roots.find zero — Method.

find_zero(fs, x0, method; kwargs...)

Interface to one of several methods for find zeros of a univariate function.

Initial starting value
For most methods, x0 is a scalar value indicating the initial value in the
iterative procedure. Values must be a subtype of Number and have methods for
float, real, and oneunit defined.
May also be a bracketing interval, specified as a tuple or a vector. A brack-
eting interval, (a,b), is one where f(a) and f(b) have different signs.
Specifying a method
A method is specified to indicate which algorithm to employ:

• There are methods for bisection where a bracket is specified: Bisection

295
296 CHAPTER 19. ROOTS

• There are methods for guarded bisection where a bracket is specified:

FalsePosition
• There are several derivative-free methods: cf. Order0, Order1 (secant
method), Order2 (Steffensen), Order5, Order8, and Order16, where the
number indicates the order of the convergence.

• There are some classical methods where a derivative is assumed or com-

puted using ForwardDiff: Newton, Halley. (The are not exported, so
they need qualification, e.g., Roots.Newton().

For more detail, see the help page for each method (e.g., ?Order5).
If no method is specified, the default method depends on x0:

• If x0 is a scalar, the default is the slower, but more robust Order0 method.
• If x0 is a tuple or vector indicating a bracketing interval, the Bisection
method is use. (this method specializes on floating point values, but oth-
erwise uses an algorithm of Alefeld, Potra, and Shi.)

Specifying the function

The function(s) are passed as the first argument.
For the few methods that use a derivative (Newton, Halley, and optionally
Order5) a tuple of functions is used. For methods requiring a derivative and
second derivative, a tuple of three functions is used. If the derivative functions
are not specified, automatic differentiation via the ForwardDiff package will be
employed (for Newton and Halley).
Optional arguments (tolerances, limit evaluations, tracing)

• xatol - absolute tolerance for x values. Passed to isapprox(x n, x {n-1})

• xrtol - relative tolerance for x values. Passed to isapprox(x n, x {n-1})

• atol - absolute tolerance for f(x) values. Passed to isapprox(f(x n),

zero(f(x n))
• rtol - relative tolerance for f(x) values. Passed to isapprox(f(x n),
zero(f(x n))

• maxevals - limit on maximum number of iterations

• maxfnevals - limit on maximum number of function evaluations
• verbose - if true a trace of the algorithm will be shown on successful
completion.

Convergence
For most methods there are several heuristics used for convergence:

• if f(x n) 0, using the tolerances atol and rtol, convergence is declared

19.2. ROOTS.FIND ZERO 297

• if x n x {n-1}, using the tolerances xatol and xrtol, and f(x n) 0 with
a relaxed tolerance then convergence is declared.

• if the algorithm has an issue (say a value of NaN appears) and f(x n) 0
with a relaxed tolerance then convergence is declared, otherwise a failure
to converge is declared

• if the number of iterations exceeds maxevals or the number of function

evaluations exceeds maxfnevals a failure to converge is declared

• if x n is NaN or f(x n) is infinite a failure to converge is declared

In general, with floating point numbers, convergence must be understood as

not an absolute statement. Even if mathematically x is an answer the floating
point realization, say xstar, may have f(xstar) - f(x) = f(xstar) f’(x) eps(x), so
tolerances must be appreciated, and at times specified.
For the Bisection methods, convergence is guaranteed, so the tolerances
are set to be 0 by default.
Examples:

# default methods
find_zero(sin, 3) # use Order0()
find_zero(sin, (3,4)) # use Bisection()

# specifying a method
find_zero(sin, 3.0, Order2()) # Use Steffensen method
find_zero(sin, big(3.0), Order16()) # rapid convergence
find_zero(sin, (3, 4), FalsePosition()) # fewer function calls than Bisection(), in this case
find_zero(sin, (3, 4), FalsePosition(8)) # 1 or 12 possible algorithms for false position
find_zero((sin,cos), 3.0, Roots.Newton()) # use Newton’s method
find_zero(sin, 3.0, Roots.Newton()) # use Newton’s method with automatic f’(x)
find_zero((sin, cos, x->-sin(x)), 3.0, Roots.Halley()) # use Halley’s method

# changing tolerances
fn, x0, xstar = (x -> (2x*cos(x) + x^2 - 3)^10/(x^2 + 1), 3.0, 2.9806452794385368)
find_zero(fn, x0, Order2()) - xstar # 0.011550654688925466
find_zero(fn, x0, Order2(), atol=0.0, rtol=0.0) # error: x_n x_{n-1}; just f(x_n) 0
fn, x0, xstar = (x -> (sin(x)*cos(x) - x^3 + 1)^9, 1.0, 1.117078770687451)
find_zero(fn, x0, Order2()) # 1.1122461983100858
find_zero(fn, x0, Order2(), maxevals=10) # Roots.ConvergenceFailed: 26 iterations needed

# tracing output
find_zero(x->sin(x), 3.0, Order2(), verbose=true) # 3 iterations
find_zero(x->sin(x)^5, 3.0, Order2(), verbose=true) # 23 iterations

source
298 CHAPTER 19. ROOTS

19.3 Roots.find zeros

Roots.find zeros — Method.
Searches for zeros of f in an interval [a, b].
Basic algorithm used:

• split interval [a,b] into no pts subintervals.

• For each bracketing interval finds a bracketed zero.

• For other subintervals does a quick search with a derivative-free method.

If there are many zeros relative to the number of points, the process is re-
peated with more points, in hopes of finding more zeros for oscillating functions.
Called by fzeros or Roots.find zeros.
source

19.4 Roots.fzero
Roots.fzero — Method.
Find zero of a function within a bracket
Uses a modified bisection method for non big arguments
Arguments:

• f A scalar function or callable object

• a left endpont of interval

• b right endpont of interval

• xtol optional additional tolerance on sub-bracket size.

For a bracket to be valid, it must be that f(a)*f(b) < 0.

For Float64 values, the answer is such that f(prevfloat(x)) * f(nextfloat(x))
< 0 unless a non-zero value of xtol is specified in which case, it stops when the
sub-bracketing produces an bracket with length less than max(xtol, abs(x1)*xtolrel).
For Big values, which defaults to the algorithm of Alefeld, Potra, and Shi,
a default tolerance is used for the sub-bracket length that can be enlarged with
xtol.
If a==-Inf it is replaced with nextfloat(a); if b==Inf it is replaced with
prevfloat(b).
Example:

‘fzero(sin, 3, 4)‘ # find pi

‘fzero(sin, [big(3), 4]) find pi with more digits

source
19.5. ROOTS.FZERO 299

19.5 Roots.fzero
Roots.fzero — Method.
Find zero of a function using an iterative algorithm

• f: a scalar function or callable object

• x0: an initial guess, finite valued.

Keyword arguments:

• ftol: tolerance for a guess abs(f(x)) < ftol + max(1, |xnew|) * ftolrel

• ftolrel: relative tolerance for convergence towards 0 of f(x)

• xtol: stop if abs(xold - xnew) <= xtol + max(1, |xnew|)*xtolrel

• xtolrel: see xtol

• maxeval: maximum number of steps

• verbose: Boolean. Set true to trace algorithm

• order: Can specify order of algorithm. 0 is most robust, also 1, 2, 5, 8,

16.

• kwargs... passed on to different algorithms. There are maxfneval when

order is 1,2,5,8, or 16 and beta for orders 2,5,8,16,

This is a polyalgorithm redirecting to different algorithms based on the value

of order.
(The tolerance arguments can also be given through atol, rtol, xatol, and
xrtol, as is done with find zero.)
source

19.6 Roots.fzero
Roots.fzero — Method.
Find zero using Newton’s method.
source

19.7 Roots.fzero
Roots.fzero — Method.
Find a zero within a bracket with an initial guess to possibly speed things
along.
source
300 CHAPTER 19. ROOTS

19.8 Roots.fzeros
Roots.fzeros — Method.
fzeros(f, a, b)
Attempt to find all zeros of f within an interval [a,b].
Simple algorithm that splits [a,b] into subintervals and checks each for a
root. For bracketing subintervals, bisection is used. Otherwise, a derivative-
free method is used. If there are a large number of zeros found relative to the
number of subintervals, the number of subintervals is increased and the process
is re-run.
There are possible issues with close-by zeros and zeros which do not cross the
origin (non-simple zeros). Answers should be confirmed graphically, if possible.
source

19.9 Roots.halley
Roots.halley — Method.
Implementation of Halley’s method. xn1 = xn - 2f(xn)*f(xn) / (2*f(xn)^2
- f(xn) * f(xn))
Arguments:

• f::Function – function to find zero of

• fp::Function=D(f) – derivative of f. Defaults to automatic derivative

• fpp:Function=D(f,2) – second derivative of f.

• x0::Number – initial guess

Keyword arguments are passed to find zero.

source

19.10 Roots.newton
Roots.newton — Method.
Implementation of Newton’s method: x n1 = x n - f(x n)/ f(x n)
Arguments:

• f::Function – function to find zero of

• fp::Function=D(f) – derivative of f. Defaults to automatic derivative

• x0::Number – initial guess. For Newton’s method this may be complex.

Keyword arguments are passed to find zero.

source
19.11. ROOTS.SECANT METHOD 301

19.11 Roots.secant method

Roots.secant method — Method.

secant_method(f, x0, x1; [kwargs...])

Solve for zero of f(x) = 0 using the secant method.

Not exported. Use find zero with Order1().
source
Chapter 20

Documenter

20.1 Documenter.deploydocs
Documenter.deploydocs — Method.

deploydocs(
root = "<current-directory>",
target = "site",
repo = "<required>",
branch = "gh-pages",
latest = "master",
osname = "linux",
julia = "<required>",
deps = <Function>,
make = <Function>,
)

Converts markdown files generated by makedocs to HTML and pushes them

to repo. This function should be called from within a package’s docs/make.jl
file after the call to makedocs, like so
[] using Documenter, PACKAGEN AM Emakedocs(options...)deploydocs(repo =”github.com/...”)

Required keyword arguments

julia is the version of Julia that will be used to deploy generated documen-
tation. This value must be one of those specified in the julia: section of the
.travis.yml configuration file.
repo is the remote repository where generated HTML content should be
pushed to. Do not specify any protocol - “https://” or “git@” should not be
present. This keyword must be set and will throw an error when left undefined.
For example this package uses the following repo value:
[] repo = ”github.com/JuliaDocs/Documenter.jl.git”
Optional keyword arguments

302
20.2. DOCUMENTER.HIDE 303

root has the same purpose as the root keyword for makedocs.
target is the directory, relative to root, where generated HTML content
should be written to. This directory must be added to the repository’s .gitignore
file. The default value is "site".
branch is the branch where the generated documentation is pushed. If the
branch does not exist, a new orphaned branch is created automatically. It
defaults to "gh-pages".
latest is the branch that “tracks” the latest generated documentation. By
default this value is set to "master".
osname is the operating system which will be used to deploy generated docu-
mentation. This defaults to "linux". This value must be one of those specified
in the os: section of the .travis.yml configuration file.
deps is the function used to install any dependencies needed to build the
documentation. By default this function installs pygments and mkdocs using
the Deps.pip function:
[] deps = Deps.pip(”pygments”, ”mkdocs”)
make is the function used to convert the markdown files to HTML. By default
this just runs mkdocs build which populates the target directory.
See Also
The Hosting Documentation section of the manual provides a step-by-step
guide to using the deploydocs function to automatically generate docs and
push them to GitHub.
source

20.2 Documenter.hide
Documenter.hide — Method.
[] hide(root, children)
Allows a subsection of pages to be hidden from the navigation menu. root
will be linked to in the navigation menu, with the title determined as usual.
children should be a list of pages (note that it can not be hierarchical).
Usage
[] makedocs( ..., pages = [ ..., hide(”Hidden section” =¿ ”hiddeni ndex.md”, [”hidden1.md”,”Hidden2”=>”hidde

source

20.3 Documenter.hide
Documenter.hide — Method.
[] hide(page)
Allows a page to be hidden in the navigation menu. It will only show up if
it happens to be the current page. The hidden page will still be present in the
linear page list that can be accessed via the previous and next page links. The
title of the hidden page can be overriden using the => operator as usual.
304 CHAPTER 20. DOCUMENTER

Usage
[] makedocs( ..., pages = [ ..., hide(”page1.md”), hide(”Title” =¿
”page2.md”) ] )
source

20.4 Documenter.makedocs
Documenter.makedocs — Method.

makedocs(
root = "<current-directory>",
source = "src",
build = "build",
clean = true,
doctest = true,
modules = Module[],
repo = "",
)

Combines markdown files and inline docstrings into an interlinked document.

In most cases makedocs should be run from a make.jl file:
[] using Documenter makedocs( keywords... )
which is then run from the command line with:
[] juliamake.jl
The folder structure that makedocs expects looks like:

docs/
build/
src/
make.jl

Keywords
root is the directory from which makedocs should run. When run from a
make.jl file this keyword does not need to be set. It is, for the most part,
needed when repeatedly running makedocs from the Julia REPL like so:

julia> makedocs(root = Pkg.dir("MyPackage", "docs"))

source is the directory, relative to root, where the markdown source files
are read from. By convention this folder is called src. Note that any non-
markdown files stored in source are copied over to the build directory when
makedocs is run.
build is the directory, relative to root, into which generated files and folders
are written when makedocs is run. The name of the build directory is, by
convention, called build, though, like with source, users are free to change
this to anything else to better suit their project needs.
20.4. DOCUMENTER.MAKEDOCS 305

clean tells makedocs whether to remove all the content from the build
folder prior to generating new content from source. By default this is set to
true.
doctest instructs makedocs on whether to try to test Julia code blocks
that are encountered in the generated document. By default this keyword is
set to true. Doctesting should only ever be disabled when initially setting
up a newly developed package where the developer is just trying to get their
package and documentation structure correct. After that, it’s encouraged to
always make sure that documentation examples are runnable and produce the
expected results. See the Doctests manual section for details about running
doctests.
modules specifies a vector of modules that should be documented in source.
If any inline docstrings from those modules are seen to be missing from the
generated content then a warning will be printed during execution of makedocs.
By default no modules are passed to modules and so no warnings will appear.
This setting can be used as an indicator of the “coverage” of the generated
documentation. For example Documenter’s make.jl file contains:
[] makedocs( modules = [Documenter], ... )
and so any docstring from the module Documenter that is not spliced into
the generated documentation in build will raise a warning.
repo specifies a template for the “link to source” feature. If you are using
GitHub, this is automatically generated from the remote. If you are using a
different host, you can use this option to tell Documenter how URLs should be
generated. The following placeholders will be replaced with the respective value
of the generated link:

• {commit} Git branch or tag name, or commit hash

• {path} Path to the file in the repository

• {line} Line (or range of lines) in the source file

For example if you are using GitLab.com, you could use

[] makedocs(repo = ”https://gitlab.com/user/project/blob/{commit}{path}{line}”)

Experimental keywords
In addition to standard arguments there is a set of non-finalized experimental
keyword arguments. The behaviour of these may change or they may be removed
without deprecation when a minor version changes (i.e. except in patch releases).
checkdocs instructs makedocs to check whether all names within the mod-
ules defined in the modules keyword that have a docstring attached have the
docstring also listed in the manual (e.g. there’s a @docs blocks with that doc-
string). Possible values are :all (check all names) and :exports (check only
exported names). The default value is :none, in which case no checks are per-
formed. If strict is also enabled then the build will fail if any missing docstrings
are encountered.
306 CHAPTER 20. DOCUMENTER

linkcheck – if set to true makedocs uses curl to check the status codes of
external-pointing links, to make sure that they are up-to-date. The links and
their status codes are printed to the standard output. If strict is also enabled
then the build will fail if there are any broken (400+ status code) links. Default:
false.
linkcheck ignore allows certain URLs to be ignored in linkcheck. The
values should be a list of strings (which get matched exactly) or Regex objects.
By default nothing is ignored.
strict – makedocs fails the build right before rendering if it encountered
any errors with the document in the previous build phases.
Non-MkDocs builds
Documenter also has (experimental) support for native HTML and LaTeX
builds. These can be enabled using the format keyword and they generally re-
quire additional keywords be defined, depending on the format. These keywords
are also currently considered experimental.
format allows the output format to be specified. Possible values are :html,
:latex and :markdown (default).
Other keywords related to non-MkDocs builds (assets, sitename, analytics,
authors, pages, version) should be documented at the respective *Writer
modules (Writers.HTMLWriter, Writers.LaTeXWriter).
See Also
A guide detailing how to document a package using Documenter’s makedocs
is provided in the Usage section of the manual.
source
Chapter 21

ColorTypes

21.1 ColorTypes.alpha
ColorTypes.alpha — Method.
alpha(p) extracts the alpha component of a color. For a color without an
alpha channel, it will always return 1.
source

21.2 ColorTypes.alphacolor
ColorTypes.alphacolor — Function.
alphacolor(RGB) returns ARGB, i.e., the corresponding transparent color
type with storage order (alpha, color).
source

21.3 ColorTypes.base color type

ColorTypes.base color type — Method.
base color type is similar to color type, except it “strips off” the element
type. For example,

color_type(RGB{N0f8}) == RGB{N0f8}
base_color_type(RGB{N0f8}) == RGB

This can be very handy if you want to switch element types. For example:

c64 = base_color_type(c){Float64}(color(c))

converts c into a Float64 representation (potentially discarding any alpha-

channel information).
source

307
308 CHAPTER 21. COLORTYPES

21.4 ColorTypes.base colorant type

ColorTypes.base colorant type — Method.
base colorant type is similar to base color type, but it preserves the
“alpha” portion of the type.
For example,

base_color_type(ARGB{N0f8}) == RGB
base_colorant_type(ARGB{N0f8}) == ARGB

If you just want to switch element types, this is the safest default and the
easiest to use:

c64 = base_colorant_type(c){Float64}(c)

source

21.5 ColorTypes.blue
ColorTypes.blue — Method.
blue(c) returns the blue component of an AbstractRGB opaque or trans-
parent color.
source

21.6 ColorTypes.ccolor
ColorTypes.ccolor — Method.
ccolor (“concrete color”) helps write flexible methods. The idea is that
users may write convert(HSV, c) or even convert(Array{HSV}, A) without
specifying the element type explicitly (e.g., convert(Array{HSV{Float32}},
A)). ccolor implements the logic “choose the user’s eltype if specified, otherwise
retain the eltype of the source object.” However, when the source object has
FixedPoint element type, and the destination only supports AbstractFloat, we
choose Float32.
Usage:

ccolor(desttype, srctype) -> concrete desttype

Example:

convert{C<:Colorant}(::Type{C}, p::Colorant) = cnvt(ccolor(C,typeof(p)), p)

where cnvt is the function that performs explicit conversion.

source
21.7. COLORTYPES.COLOR 309

21.7 ColorTypes.color
ColorTypes.color — Method.
color(c) extracts the opaque color component from a Colorant (e.g., omits
the alpha channel, if present).
source

21.8 ColorTypes.color type

ColorTypes.color type — Method.
color type(c) or color type(C) (c being a color instance and C being the
type) returns the type of the Color object (without alpha channel). This, and
related functions like base color type, base colorant type, and ccolor are
useful for manipulating types for writing generic code.
For example,

color_type(RGB) == RGB
color_type(RGB{Float32}) == RGB{Float32}
color_type(ARGB{N0f8}) == RGB{N0f8}

source

21.9 ColorTypes.coloralpha
ColorTypes.coloralpha — Function.
coloralpha(RGB) returns RGBA, i.e., the corresponding transparent color
type with storage order (color, alpha).
source

21.10 ColorTypes.comp1
ColorTypes.comp1 — Method.
comp1(c) extracts the first component you’d pass to the constructor of the
corresponding object. For most color types without an alpha channel, this is
just the first field, but for types like BGR that reverse the internal storage order
this provides the value that you’d use to reconstruct the color.
Specifically, for any Color{T,3},

c == typeof(c)(comp1(c), comp2(c), comp3(c))

returns true.
source
310 CHAPTER 21. COLORTYPES

21.11 ColorTypes.comp2
ColorTypes.comp2 — Method.
comp2(c) extracts the second constructor argument (see comp1).
source

21.12 ColorTypes.comp3
ColorTypes.comp3 — Method.
comp3(c) extracts the third constructor argument (see comp1).
source

21.13 ColorTypes.gray
ColorTypes.gray — Method.
gray(c) returns the gray component of a grayscale opaque or transparent
color.
source

21.14 ColorTypes.green
ColorTypes.green — Method.
green(c) returns the green component of an AbstractRGB opaque or trans-
parent color.
source

21.15 ColorTypes.mapc
ColorTypes.mapc — Method.
mapc(f, rgb) -> rgbf
mapc(f, rgb1, rgb2) -> rgbf
mapc applies the function f to each color channel of the input color(s), re-
turning an output color in the same colorspace.
Examples:
julia> mapc(x->clamp(x,0,1), RGB(-0.2,0.3,1.2))
RGB{Float64}(0.0,0.3,1.0)

julia> mapc(max, RGB(0.1,0.8,0.3), RGB(0.5,0.5,0.5))

RGB{Float64}(0.5,0.8,0.5)

julia> mapc(+, RGB(0.1,0.8,0.3), RGB(0.5,0.5,0.5))

RGB{Float64}(0.6,1.3,0.8)
21.16. COLORTYPES.MAPREDUCEC 311

source

21.16 ColorTypes.mapreducec
ColorTypes.mapreducec — Method.

mapreducec(f, op, v0, c)

Reduce across color channels of c with the binary operator op, first applying
f to each channel. v0 is the neutral element used to initiate the reduction. For
grayscale,

mapreducec(f, op, v0, c::Gray) = op(v0, f(comp1(c)))

whereas for RGB

mapreducec(f, op, v0, c::RGB) = op(f(comp3(c)), op(f(comp2(c)), op(v0, f(comp1(c)))))

If c has an alpha channel, it is always the last one to be folded into the
reduction.
source

21.17 ColorTypes.red
ColorTypes.red — Method.
red(c) returns the red component of an AbstractRGB opaque or transparent
color.
source

21.18 ColorTypes.reducec
ColorTypes.reducec — Method.

reducec(op, v0, c)

Reduce across color channels of c with the binary operator op. v0 is the
neutral element used to initiate the reduction. For grayscale,

reducec(op, v0, c::Gray) = op(v0, comp1(c))

whereas for RGB

reducec(op, v0, c::RGB) = op(comp3(c), op(comp2(c), op(v0, comp1(c))))

If c has an alpha channel, it is always the last one to be folded into the
reduction.
source
Chapter 22

FileIO

22.1 Base.info
Base.info — Method.
info(fmt) returns the magic bytes/extension information for DataFormat
fmt.
source

22.2 FileIO.add format

FileIO.add format — Method.
add format(fmt, magic, extention) registers a new DataFormat. For ex-
ample:

add_format(format"PNG", (UInt8[0x4d,0x4d,0x00,0x2b], UInt8[0x49,0x49,0x2a,0x00]), [".ti

add_format(format"PNG", [0x89,0x50,0x4e,0x47,0x0d,0x0a,0x1a,0x0a], ".png")
add_format(format"NRRD", "NRRD", [".nrrd",".nhdr"])

Note that extensions, magic numbers, and format-identifiers are case-sensitive.

source

22.3 FileIO.add loader

FileIO.add loader — Function.
add loader(fmt, :Package) triggers using Package before loading format
fmt
source

312
22.4. FILEIO.ADD SAVER 313

22.4 FileIO.add saver

FileIO.add saver — Function.
add saver(fmt, :Package) triggers using Package before saving format
fmt
source

22.5 FileIO.del format

FileIO.del format — Method.
del format(fmt::DataFormat) deletes fmt from the format registry.
source

22.6 FileIO.file extension

FileIO.file extension — Method.
file extension(file) returns the file extension associated with File file.
source

22.7 FileIO.file extension

FileIO.file extension — Method.
file extension(file) returns a nullable-string for the file extension asso-
ciated with Stream stream.
source

22.8 FileIO.filename
FileIO.filename — Method.
filename(file) returns the filename associated with File file.
source

22.9 FileIO.filename
FileIO.filename — Method.
filename(stream) returns a string of the filename associated with Stream
stream, or nothing if there is no file associated.
source
314 CHAPTER 22. FILEIO

22.10 FileIO.load
FileIO.load — Function.

• load(filename) loads the contents of a formatted file, trying to infer

the format from filename and/or magic bytes in the file.

• load(strm) loads from an IOStream or similar object. In this case,

there is no filename extension, so we rely on the magic bytes for format

identification.

• load(File(format"PNG", filename)) specifies the format directly, and

bypasses inference.

• load(Stream(format"PNG", io)) specifies the format directly, and by-

passes inference.

• load(f; options...) passes keyword arguments on to the loader.

source

22.11 FileIO.loadstreaming
FileIO.loadstreaming — Function.
Some packages may implement a streaming API, where the contents of the
file can be read in chunks and processed, rather than all at once. Reading from
these higher-level streams should return a formatted object, like an image or
chunk of video or audio.

• loadstreaming(filename) loads the contents of a formatted file, trying

to infer

the format from filename and/or magic bytes in the file. It returns a stream-
ing type that can be read from in chunks, rather than loading the whole contents
all at once

• loadstreaming(strm) loads the stream from an IOStream or similar ob-

ject.

In this case, there is no filename extension, so we rely on the magic bytes

for format identification.

• loadstreaming(File(format"WAV",filename)) specifies the format di-

rectly, and

bypasses inference.
22.12. FILEIO.MAGIC 315

• loadstreaming(Stream(format"WAV", io)) specifies the format directly,

and
bypasses inference.
• loadstreaming(f; options...) passes keyword arguments on to the
loader.
source

22.12 FileIO.magic
FileIO.magic — Method.
magic(fmt) returns the magic bytes of format fmt
source

22.13 FileIO.query
FileIO.query — Method.
query(filename) returns a File object with information about the format
inferred from the file’s extension and/or magic bytes.
source

22.14 FileIO.query
FileIO.query — Method.
query(io, [filename]) returns a Stream object with information about
the format inferred from the magic bytes.
source

22.15 FileIO.save
FileIO.save — Function.
• save(filename, data...) saves the contents of a formatted file,
trying to infer the format from filename.
• save(Stream(format"PNG",io), data...) specifies the format directly,
and bypasses inference.
• save(File(format"PNG",filename), data...) specifies the format di-
rectly, and bypasses inference.
• save(f, data...; options...) passes keyword arguments on to the
saver.
source
316 CHAPTER 22. FILEIO

22.16 FileIO.savestreaming
FileIO.savestreaming — Function.
Some packages may implement a streaming API, where the contents of the
file can be written in chunks, rather than all at once. These higher-level streams
should accept formatted objects, like an image or chunk of video or audio.

• savestreaming(filename, data...) saves the contents of a formatted

file,

trying to infer the format from filename.

• savestreaming(File(format"WAV",filename)) specifies the format di-

rectly, and

bypasses inference.

• savestreaming(Stream(format"WAV", io)) specifies the format directly,

and

bypasses inference.

• savestreaming(f, data...; options...) passes keyword arguments

on to the saver.

source

22.17 FileIO.skipmagic
FileIO.skipmagic — Method.
skipmagic(s) sets the position of Stream s to be just after the magic bytes.
For a plain IO object, you can use skipmagic(io, fmt).
source

22.18 FileIO.stream
FileIO.stream — Method.
stream(s) returns the stream associated with Stream s
source

22.19 FileIO.unknown
FileIO.unknown — Method.
unknown(f) returns true if the format of f is unknown.
source
Chapter 23

DataArrays

23.1 DataArrays.PooledDataVecs
DataArrays.PooledDataVecs — Method.

PooledDataVecs(v1, v2) -> (pda1, pda2)

Return a tuple of PooledDataArrays created from the data in v1 and v2,

respectively, but sharing a common value pool.
source

23.2 DataArrays.compact
DataArrays.compact — Method.

compact(d::PooledDataArray)

Return a PooledDataArray with the smallest possible reference type for the
data in d.
!!! note If the reference type is already the smallest possible for the data,
the input array is returned, i.e. the function aliases the input.
Examples

julia> p = @pdata(repeat(["A", "B"], outer=4))

8-element DataArrays.PooledDataArray{String,UInt32,1}:
"A"
"B"
"A"
"B"
"A"
"B"
"A"

317
318 CHAPTER 23. DATAARRAYS

"B"

julia> compact(p) # second type parameter compacts to UInt8 (only need 2 unique values)
8-element DataArrays.PooledDataArray{String,UInt8,1}:
"A"
"B"
"A"
"B"
"A"
"B"
"A"
"B"

source

23.3 DataArrays.cut
DataArrays.cut — Method.

cut(x::AbstractVector, breaks::Vector) -> PooledDataArray

cut(x::AbstractVector, ngroups::Integer) -> PooledDataArray

Divide the range of x into intervals based on the cut points specified in
breaks, or into ngroups intervals of approximately equal length.
Examples

julia> cut([1, 2, 3, 4], [1, 3])

4-element DataArrays.PooledDataArray{String,UInt32,1}:
"[1,3]"
"[1,3]"
"[1,3]"
"(3,4]"

source

23.4 DataArrays.data
DataArrays.data — Method.

data(a::AbstractArray) -> DataArray

Convert a to a DataArray.
Examples
23.5. DATAARRAYS.GETPOOLIDX 319

julia> data([1, 2, 3])

3-element DataArrays.DataArray{Int64,1}:
1
2
3

julia> data(@data [1, 2, missing])

3-element DataArrays.DataArray{Int64,1}:
1
2
missing

source

23.5 DataArrays.getpoolidx
DataArrays.getpoolidx — Method.

getpoolidx(pda::PooledDataArray, val)

Return the index of val in the value pool for pda. If val is not already in
the value pool, pda is modified to include it in the pool.
source

23.6 DataArrays.gl
DataArrays.gl — Method.

gl(n::Integer, k::Integer, l::Integer = n*k) -> PooledDataArray

Generate a PooledDataArray with n levels and k replications, optionally

specifying an output length l. If specified, l must be a multiple of n*k.
Examples

julia> gl(2, 1)
2-element DataArrays.PooledDataArray{Int64,UInt8,1}:
1
2

julia> gl(2, 1, 4)
4-element DataArrays.PooledDataArray{Int64,UInt8,1}:
1
2
1
2

source
320 CHAPTER 23. DATAARRAYS

23.7 DataArrays.padmissing
DataArrays.padmissing — Method.

padmissing(dv::AbstractDataVector, front::Integer, back::Integer) -> DataVector

Pad dv with missing values. front is an integer number of missings to

add at the beginning of the array and back is the number of missings to add
at the end.
Examples

julia> padmissing(@data([1, 2, 3]), 1, 2)

6-element DataArrays.DataArray{Int64,1}:
missing
1
2
3
missing
missing

source

23.8 DataArrays.reorder
DataArrays.reorder — Method.

reorder(x::PooledDataArray) -> PooledDataArray

Return a PooledDataArray containing the same data as x but with the value
pool sorted.
source

23.9 DataArrays.replace!
DataArrays.replace! — Method.

replace!(x::PooledDataArray, from, to)

Replace all occurrences of from in x with to, modifying x in place.

source

23.10 DataArrays.setlevels!
DataArrays.setlevels! — Method.

setlevels!(x::PooledDataArray, newpool::Union{AbstractVector, Dict})

23.11. DATAARRAYS.SETLEVELS 321

Set the value pool for the PooledDataArray x to newpool, modifying x in

place. The values can be replaced using a mapping specified in a Dict or with
an array, since the order of the levels is used to identify values. The pool can
be enlarged to contain values not present in the data, but it cannot be reduced
to exclude present values.
Examples

julia> p = @pdata repeat(["A", "B"], inner=3)

6-element DataArrays.PooledDataArray{String,UInt32,1}:
"A"
"A"
"A"
"B"
"B"
"B"

julia> setlevels!(p, Dict("A"=>"C"));

julia> p # has been modified

6-element DataArrays.PooledDataArray{String,UInt32,1}:
"C"
"C"
"C"
"B"
"B"
"B"

source

23.11 DataArrays.setlevels
DataArrays.setlevels — Method.

setlevels(x::PooledDataArray, newpool::Union{AbstractVector, Dict})

Create a new PooledDataArray based on x but with the new value pool
specified by newpool. The values can be replaced using a mapping specified in
a Dict or with an array, since the order of the levels is used to identify values.
The pool can be enlarged to contain values not present in the data, but it cannot
be reduced to exclude present values.
Examples

julia> p = @pdata repeat(["A", "B"], inner=3)

6-element DataArrays.PooledDataArray{String,UInt32,1}:
"A"
"A"
322 CHAPTER 23. DATAARRAYS

"A"
"B"
"B"
"B"

julia> p2 = setlevels(p, ["C", "D"]) # could also be Dict("A"=>"C", "B"=>"D")

6-element DataArrays.PooledDataArray{String,UInt32,1}:
"C"
"C"
"C"
"D"
"D"
"D"

julia> p3 = setlevels(p2, ["C", "D", "E"])

6-element DataArrays.PooledDataArray{String,UInt32,1}:
"C"
"C"
"C"
"D"
"D"
"D"

julia> p3.pool # the pool can contain values not in the array
3-element Array{String,1}:
"C"
"D"
"E"
source

23.12 Missings.ismissing
Missings.ismissing — Method.
ismissing(a::AbstractArray, i) -> Bool
Determine whether the element of a at index i is missing, i.e. missing.
Examples
julia> X = @data [1, 2, missing];

julia> ismissing(X, 2)
false

julia> ismissing(X, 3)
true
23.12. MISSINGS.ISMISSING 323

source
Chapter 24

Primes

24.1 Primes.factor
Primes.factor — Method.

factor(ContainerType, n::Integer) -> ContainerType

Return the factorization of n stored in a ContainerType, which must be a

subtype of AbstractDict or AbstractArray, a Set, or an BitSet.
[] julia¿ factor(DataStructures.SortedDict, 100) DataStructures.SortedDict{Int64,Int64,Base.Order.Fo
with 2 entries: 2 =¿ 2 5 =¿ 2
When ContainerType <: AbstractArray, this returns the list of all prime
factors of n with multiplicities, in sorted order.
[] julia¿ factor(Vector, 100) 4-element Array{Int64,1}: 2 2 5 5
julia¿ prod(factor(Vector, 100)) == 100 true
When ContainerType == Set, this returns the distinct prime factors as a
set.
[] julia¿ factor(Set, 100) Set([2,5])
source

24.2 Primes.factor
Primes.factor — Method.

factor(n::Integer) -> Primes.Factorization

Compute the prime factorization of an integer n. The returned object, of

type Factorization, is an associative container whose keys correspond to the
factors, in sorted order. The value associated with each key indicates the mul-
tiplicity (i.e. the number of times the factor appears in the factorization).
[] julia¿ factor(100) 2252

324
24.3. PRIMES.ISMERSENNEPRIME 325

For convenience, a negative number n is factored as -1*(-n) (i.e. -1 is

considered to be a factor), and 0 is factored as 0^1:
[] julia¿ factor(-9) -1 32
julia¿ factor(0) 0
julia¿ collect(factor(0)) 1-element Array{Pair{Int64,Int64},1}: 0=¿1
source

24.3 Primes.ismersenneprime
Primes.ismersenneprime — Method.
ismersenneprime(M::Integer; [check::Bool = true]) -> Bool
Lucas-Lehmer deterministic test for Mersenne primes. M must be a Mersenne
number, i.e. of the form M = 2^p - 1, where p is a prime number. Use the key-
word argument check to enable/disable checking whether M is a valid Mersenne
number; to be used with caution. Return true if the given Mersenne number
is prime, and false otherwise.
julia> ismersenneprime(2^11 - 1)
false

julia> ismersenneprime(2^13 - 1)
true
source

24.4 Primes.isprime
Primes.isprime — Function.
isprime(x::BigInt, [reps = 25]) -> Bool
Probabilistic primality test. Returns true if x is prime with high probability
(pseudoprime); and false if x is composite (not prime). The false positive rate
is about 0.25^reps. reps = 25 is considered safe for cryptographic applications
(Knuth, Seminumerical Algorithms).
[] julia¿ isprime(big(3)) true
source

24.5 Primes.isprime
Primes.isprime — Method.
isprime(n::Integer) -> Bool
Returns true if n is prime, and false otherwise.
[] julia¿ isprime(3) true
source
326 CHAPTER 24. PRIMES

24.6 Primes.isrieselprime
Primes.isrieselprime — Method.

isrieselprime(k::Integer, Q::Integer) -> Bool

Lucas-Lehmer-Riesel deterministic test for N of the form N = k * Q, with

0 < k < Q, Q = 2^n - 1 and n > 0, also known as Riesel primes. Returns
true if R is prime, and false otherwise or if the combination of k and n is not
supported.

julia> isrieselprime(1, 2^11 - 1) # == ismersenneprime(2^11 - 1)

false

julia> isrieselprime(3, 2^607 - 1)

true

source

24.7 Primes.nextprime
Primes.nextprime — Function.

nextprime(n::Integer, i::Integer=1)

The i-th smallest prime not less than n (in particular, nextprime(p) == p
if p is prime). If i < 0, this is equivalent to prevprime(n, -i). Note that for
n::BigInt, the returned number is only a pseudo-prime (the function isprime
is used internally). See also prevprime.

julia> nextprime(4)
5

julia> nextprime(5)
5

julia> nextprime(4, 2)
7

julia> nextprime(5, 2)
7

source
24.8. PRIMES.PREVPRIME 327

24.8 Primes.prevprime
Primes.prevprime — Function.

prevprime(n::Integer, i::Integer=1)

The i-th largest prime not greater than n (in particular prevprime(p) == p
if p is prime). If i < 0, this is equivalent to nextprime(n, -i). Note that for
n::BigInt, the returned number is only a pseudo-prime (the function isprime
is used internally). See also nextprime.

julia> prevprime(4)
3

julia> prevprime(5)
5

julia> prevprime(5, 2)
3

source

24.9 Primes.prime
Primes.prime — Method.

prime(::Type{<:Integer}=Int, i::Integer)

The i-th prime number.

julia> prime(1)
2

julia> prime(3)
5

source

24.10 Primes.primes
Primes.primes — Method.

primes([lo,] hi)

Returns a collection of the prime numbers (from lo, if specified) up to hi.

source
328 CHAPTER 24. PRIMES

24.11 Primes.primesmask
Primes.primesmask — Method.

primesmask([lo,] hi)

Returns a prime sieve, as a BitArray, of the positive integers (from lo,

if specified) up to hi. Useful when working with either primes or composite
numbers.
source

24.12 Primes.prodfactors
Primes.prodfactors — Function.

prodfactors(factors)

Compute n (or the radical of n when factors is of type Set or BitSet)

where factors is interpreted as the result of factor(typeof(factors), n).
Note that if factors is of type AbstractArray or Primes.Factorization, then
prodfactors is equivalent to Base.prod.

julia> prodfactors(factor(100))
100

source

24.13 Primes.radical
Primes.radical — Method.

radical(n::Integer)

Compute the radical of n, i.e. the largest square-free divisor of n. This is

equal to the product of the distinct prime numbers dividing n.

julia> radical(2*2*3)
6

source
24.14. PRIMES.TOTIENT 329

24.14 Primes.totient
Primes.totient — Method.

totient(n::Integer) -> Integer

Compute the Euler totient function (n), which counts the number of positive
integers less than or equal to n that are relatively prime to n (that is, the number
of positive integers m n with gcd(m, n) == 1). The totient function of n when
n is negative is defined to be totient(abs(n)).
source

24.15 Primes.totient
Primes.totient — Method.

totient(f::Factorization{T}) -> T

Compute the Euler totient function of the number whose prime factorization
is given by f. This method may be preferable to totient(::Integer) when
the factorization can be reused for other purposes.
source
Chapter 25

ImageTransformations

25.1 ImageTransformations.imresize
ImageTransformations.imresize — Method.

imresize(img, sz) -> imgr

imresize(img, inds) -> imgr

Change img to be of size sz (or to have indices inds). This interpolates the
values at sub-pixel locations. If you are shrinking the image, you risk aliasing
unless you low-pass filter img first. For example:

= map((o,n)->0.75*o/n, size(img), sz)

kern = KernelFactors.gaussian() # from ImageFiltering
imgr = imresize(imfilter(img, kern, NA()), sz)

See also restrict.

source

25.2 ImageTransformations.invwarpedview
ImageTransformations.invwarpedview — Method.

invwarpedview(img, tinv, [indices], [degree = Linear()], [fill = NaN]) -> wv

Create a view of img that lazily transforms any given index I passed to
wv[I] to correspond to img[inv(tinv)(I)]. While technically this approach
is known as backward mode warping, note that InvWarpedView is created by
supplying the forward transformation. The given transformation tinv must
accept a SVector as input and support inv(tinv). A useful package to create
a wide variety of such transformations is CoordinateTransformations.jl.
When invoking wv[I], values for img must be reconstructed at arbitrary lo-
cations inv(tinv)(I). InvWarpedView serves as a wrapper around WarpedView

330
25.3. IMAGETRANSFORMATIONS.RESTRICT 331

which takes care of interpolation and extrapolation. The parameters degree and
fill can be used to specify the b-spline degree and the extrapolation scheme
respectively.
The optional parameter indices can be used to specify the domain of the
resulting wv. By default the indices are computed in such a way that wv contains
all the original pixels in img.
source

25.3 ImageTransformations.restrict
ImageTransformations.restrict — Method.

restrict(img[, region]) -> imgr

Reduce the size of img by two-fold along the dimensions listed in region, or
all spatial coordinates if region is not specified. It anti-aliases the image as it
goes, so is better than a naive summation over 2x2 blocks.
See also imresize.
source

25.4 ImageTransformations.warp
ImageTransformations.warp — Method.

warp(img, tform, [indices], [degree = Linear()], [fill = NaN]) -> imgw

Transform the coordinates of img, returning a new imgw satisfying imgw[I]

= img[tform(I)]. This approach is known as backward mode warping. The
transformation tform must accept a SVector as input. A useful package to
create a wide variety of such transformations is CoordinateTransformations.jl.
Reconstruction scheme
During warping, values for img must be reconstructed at arbitrary locations
tform(I) which do not lie on to the lattice of pixels. How this reconstruction
is done depends on the type of img and the optional parameter degree.
When img is a plain array, then on-grid b-spline interpolation will be used.
It is possible to configure what degree of b-spline to use with the parameter
degree. For example one can use degree = Linear() for linear interpola-
tion, degree = Constant() for nearest neighbor interpolation, or degree =
Quadratic(Flat()) for quadratic interpolation.
In the case tform(I) maps to indices outside the original img, those locations
are set to a value fill (which defaults to NaN if the element type supports it,
and 0 otherwise). The parameter fill also accepts extrapolation schemes, such
as Flat(), Periodic() or Reflect().
For more control over the reconstruction scheme — and how beyond-the-edge
points are handled — pass img as an AbstractInterpolation or AbstractExtrapolation
from Interpolations.jl.
332 CHAPTER 25. IMAGETRANSFORMATIONS

The meaning of the coordinates

The output array imgw has indices that would result from applying inv(tform)
to the indices of img. This can be very handy for keeping track of how pixels
in imgw line up with pixels in img.
If you just want a plain array, you can “strip” the custom indices with
parent(imgw).
Examples: a 2d rotation (see JuliaImages documentation for pic-
tures)

julia> using Images, CoordinateTransformations, TestImages, OffsetArrays

julia> img = testimage("lighthouse");

julia> indices(img)
(Base.OneTo(512),Base.OneTo(768))

# Rotate around the center of ‘img‘

julia> tfm = recenter(RotMatrix(-pi/4), center(img))
AffineMap([0.707107 0.707107; -0.707107 0.707107], [-196.755,293.99])

julia> imgw = warp(img, tfm);

julia> indices(imgw)
(-196:709,-68:837)

# Alternatively, specify the origin in the image itself

julia> img0 = OffsetArray(img, -30:481, -384:383); # origin near top of image

julia> rot = LinearMap(RotMatrix(-pi/4))

LinearMap([0.707107 -0.707107; 0.707107 0.707107])

julia> imgw = warp(img0, rot);

julia> indices(imgw)
(-293:612,-293:611)

julia> imgr = parent(imgw);

julia> indices(imgr)
(Base.OneTo(906),Base.OneTo(905))

source

25.5 ImageTransformations.warpedview
ImageTransformations.warpedview — Method.
25.5. IMAGETRANSFORMATIONS.WARPEDVIEW 333

warpedview(img, tform, [indices], [degree = Linear()], [fill = NaN]) -> wv

Create a view of img that lazily transforms any given index I passed to
wv[I] to correspond to img[tform(I)]. This approach is known as backward
mode warping. The given transformation tform must accept a SVector as
input. A useful package to create a wide variety of such transformations is
CoordinateTransformations.jl.
When invoking wv[I], values for img must be reconstructed at arbitrary
locations tform(I) which do not lie on to the lattice of pixels. How this re-
construction is done depends on the type of img and the optional parameter
degree. When img is a plain array, then on-grid b-spline interpolation will
be used, where the pixel of img will serve as the coeficients. It is possible to
configure what degree of b-spline to use with the parameter degree. The two
possible values are degree = Linear() for linear interpolation, or degree =
Constant() for nearest neighbor interpolation.
In the case tform(I) maps to indices outside the domain of img, those
locations are set to a value fill (which defaults to NaN if the element type
supports it, and 0 otherwise). Additionally, the parameter fill also accepts
extrapolation schemes, such as Flat(), Periodic() or Reflect().
The optional parameter indices can be used to specify the domain of the
resulting WarpedView. By default the indices are computed in such a way that
the resulting WarpedView contains all the original pixels in img. To do this
inv(tform) has to be computed. If the given transformation tform does not
support inv, then the parameter indices has to be specified manually.
warpedview is essentially a non-coping, lazy version of warp. As such,
the two functions share the same interface, with one important difference.
warpedview will insist that the resulting WarpedView will be a view of img
(i.e. parent(warpedview(img, ...)) === img). Consequently, warpedview
restricts the parameter degree to be either Linear() or Constant().
source
Chapter 26

StatsFuns

26.1 StatsFuns.log1mexp
StatsFuns.log1mexp — Method.

log1mexp(x::Real)

Return log(1 - exp(x))

See:

• Martin Maechler (2012) “Accurately Computing log(1 exp( |a|))”, http://cran.r-

project.org/web/packages/Rmpfr/vignettes/log1mexp-note.pdf

Note: different than Maechler (2012), no negation inside parentheses

source

26.2 StatsFuns.log1pexp
StatsFuns.log1pexp — Method.

log1pexp(x::Real)

Return log(1+exp(x)) evaluated carefully for largish x.

This is also called the “softplus” transformation, being a smooth approxi-
mation to max(0,x). Its inverse is logexpm1.
source

26.3 StatsFuns.log1pmx
StatsFuns.log1pmx — Method.

log1pmx(x::Float64)

334
26.4. STATSFUNS.LOG1PSQ 335

Return log(1 + x) - x
Use naive calculation or range reduction outside kernel range. Accurate
˜2ulps for all x.
source

26.4 StatsFuns.log1psq
StatsFuns.log1psq — Method.

log1psq(x::Real)

Return log(1+x^2) evaluated carefully for abs(x) very small or very large.
source

26.5 StatsFuns.log2mexp
StatsFuns.log2mexp — Method.

log2mexp(x::Real)

Return log(2 - exp(x)) evaluated as log1p(-expm1(x))

source

26.6 StatsFuns.logaddexp
StatsFuns.logaddexp — Method.

logaddexp(x::Real, y::Real)

Return log(exp(x) + exp(y)), avoiding intermediate overflow/undeflow,

and handling non-finite values.
source

26.7 StatsFuns.logexpm1
StatsFuns.logexpm1 — Method.

logexpm1(x::Real)

Return log(exp(x) - 1) or the “invsoftplus” function. It is the inverse of

log1pexp (aka “softplus”).
source
336 CHAPTER 26. STATSFUNS

26.8 StatsFuns.logistic
StatsFuns.logistic — Method.

logistic(x::Real)

The logistic sigmoid function mapping a real number to a value in the interval
[0,1],

1 ex
σ(x) = = .
e−x + 1 1 + ex
Its inverse is the logit function.
source

26.9 StatsFuns.logit
StatsFuns.logit — Method.

logit(p::Real)

The logit or log-odds transformation,

 
x
log , where0 < x < 1
1−x
Its inverse is the logistic function.
source

26.10 StatsFuns.logmvgamma
StatsFuns.logmvgamma — Method.

logmvgamma(p::Int, a::Real)

Return the logarithm of multivariate gamma function (DLMF 35.3.1).

source

26.11 StatsFuns.logmxp1
StatsFuns.logmxp1 — Method.

logmxp1(x::Float64)

Return log(x) - x + 1 carefully evaluated.

source
26.12. STATSFUNS.LOGSUMEXP 337

26.12 StatsFuns.logsumexp
StatsFuns.logsumexp — Method.

logsumexp(X)

Compute log(sum(exp, X)), evaluated avoiding intermediate overflow/undeflow.

X should be an iterator of real numbers.
source

26.13 StatsFuns.lstirling asym

StatsFuns.lstirling asym — Function.

lstirling_asym(x)

The remainder term after Stirling’s approximation to lgamma:

log Γ(x) ≈ x log(x) − x + log(2π/x)/2 = log(x) ∗ (x − 1/2) + log(2π)/2 − x

In Julia syntax, this means:

lstirling_asym(x) = lgamma(x) + x - (x-0.5)*log(x) - 0.5*log(2)

For sufficiently large x, this can be approximated using the asymptotic Stir-
ling’s series (DLMF 5.11.1):

1 1 1 1
− + − + ...
12x 360x3 1260x5 1680x7
The truncation error is bounded by the first omitted term, and is of the
same sign.
Relative error of approximation is bounded by (174611/125400 xˆ-19) /
(1/12 xˆ-1 - 1/360 xˆ-3) which is < 1/2 ulp for x >= 10.0, and total numeric
error appears to be < 2 ulps
References

• Temme, N. (1996) Special functions: An introduction to the classical func-

tions of mathematical physics, Wiley, New York, ISBN: 0-471-11313-1,
Chapter 3.6, pp 61-65.

• Weisstein, Eric W. “Stirling’s Series.”. MathWorld.

• OEIS A046968 and OEIS A046969 for the series coefficients

source
338 CHAPTER 26. STATSFUNS

26.14 StatsFuns.softmax!
StatsFuns.softmax! — Method.

softmax(x::AbstractArray{<:Real})

Return the softmax transformation applied to x

source

26.15 StatsFuns.softmax!
StatsFuns.softmax! — Method.

softmax!(r::AbstractArray, x::AbstractArray)

Overwrite r with the softmax (or normalized exponential ) transformation

of x
That is, r is overwritten with exp.(x), normalized to sum to 1.
See the Wikipedia entry
source

26.16 StatsFuns.xlogx
StatsFuns.xlogx — Method.

xlogx(x::Real)

Return x * log(x) for x 0, handling x = 0 by taking the downward limit.

julia> StatsFuns.xlogx(0)
0.0

source

26.17 StatsFuns.xlogy
StatsFuns.xlogy — Method.

xlogy(x::Real, y::Real)

Return x * log(y) for y > 0 with correct limit at x = 0.

source
Chapter 27

Gadfly

27.1 Compose.draw
Compose.draw — Method.

draw(backend::Compose.Backend, p::Plot)

A convenience version of Compose.draw without having to call render

Args

• backend: The Compose.Backend object

• p: The Plot object

source

27.2 Compose.hstack
Compose.hstack — Method.

hstack(ps::Union{Plot,Context}...)
hstack(ps::Vector)

Arrange plots into a horizontal row. Use context() as a placeholder for

an empty panel. Heterogeneous vectors must be typed. See also vstack,
gridstack, subplot grid.
Examples
“‘ p1 = plot(x=[1,2], y=[3,4], Geom.line); p2 = Compose.context(); hstack(p1,
p2) hstack(Union{Plot,Compose.Context}[p1, p2])
source

339
340 CHAPTER 27. GADFLY

27.3 Compose.vstack
Compose.vstack — Method.

vstack(ps::Union{Plot,Context}...)
vstack(ps::Vector)

Arrange plots into a vertical column. Use context() as a placeholder for

an empty panel. Heterogeneous vectors must be typed. See also hstack,
gridstack, subplot grid.
Examples
“‘ p1 = plot(x=[1,2], y=[3,4], Geom.line); p2 = Compose.context(); vs-
tack(p1, p2) vstack(Union{Plot,Compose.Context}[p1, p2])
source

27.4 Gadfly.layer
Gadfly.layer — Method.

layer(data_source::@compat(Union{AbstractDataFrame, (@compat Void)}),

elements::ElementOrFunction...; mapping...)

Creates layers based on elements

Args

• data source: The data source as a dataframe

• elements: The elements
• mapping: mapping

Returns
An array of layers
source

27.5 Gadfly.plot
Gadfly.plot — Method.

function plot(data_source::@compat(Union{(@compat Void), AbstractMatrix, AbstractDataFr

mapping::Dict, elements::ElementOrFunctionOrLayers...)

The old fashioned (pre named arguments) version of plot.

This version takes an explicit mapping dictionary, mapping aesthetics sym-
bols to expressions or columns in the data frame.
Args:

• data source: Data to be bound to aesthetics.

27.6. GADFLY.PLOT 341

• mapping: Dictionary of aesthetics symbols (e.g. :x, :y, :color) to names of

columns in the data frame or other expressions.
• elements: Geometries, statistics, etc.

Returns:
A Plot object.
source

27.6 Gadfly.plot
Gadfly.plot — Method.

function plot(data_source::@compat(Union{AbstractMatrix, AbstractDataFrame}),

elements::ElementOrFunctionOrLayers...; mapping...)

Create a new plot.

Grammar of graphics style plotting consists of specifying a dataset, one
or more plot elements (scales, coordinates, geometries, etc), and binding of
aesthetics to columns or expressions of the dataset.
For example, a simple scatter plot would look something like:
plot(my data, Geom.point, x=“time”, y=“price”)
Where “time” and “price” are the names of columns in my data.
Args:

• data source: Data to be bound to aesthetics.

• elements: Geometries, statistics, etc.
• mapping: Aesthetics symbols (e.g. :x, :y, :color) mapped to names of
columns in the data frame or other expressions.

source

27.7 Gadfly.render
Gadfly.render — Method.

render(plot::Plot)

Render a plot based on the Plot object

Args

• plot: Plot to be rendered.

Returns
A Compose context containing the rendered plot.
source
342 CHAPTER 27. GADFLY

27.8 Gadfly.set default plot format

Gadfly.set default plot format — Method.

set_default_plot_format(fmt::Symbol)

Sets the default plot format

source

27.9 Gadfly.set default plot size

Gadfly.set default plot size — Method.

set_default_plot_size(width::Compose.MeasureOrNumber, height::Compose.MeasureOrNumber)

Sets preferred canvas size when rendering a plot without an explicit call to
draw
source

27.10 Gadfly.spy
Gadfly.spy — Method.

spy(M::AbstractMatrix, elements::ElementOrFunction...; mapping...)

Simple heatmap plots of matrices.

It is a wrapper around the plot() function using the rectbin geometry.
It also applies a sane set of defaults to make sure that the plots look nice by
default. Specifically

• the aspect ratio of the coordinate system is fixed Coord.cartesian(fixed=true),

so that the rectangles become squares

• the axes run from 0.5 to N+0.5, because the first row/column is drawn to

(0.5, 1.5) and the last one to (N-0.5, N+0.5).

• the y-direction is flipped, so that the [1,1] of a matrix is in the top

left corner, as is customary

• NaNs are not drawn. spy leaves “holes” instead into the heatmap.

Args:

• M: A matrix.
27.11. GADFLY.STYLE 343

Returns:
A plot object.
Known bugs:

• If the matrix is only NaNs, then it throws an ArgumentError, because an

empty collection gets passed to the plot function / rectbin geometry.

source

27.11 Gadfly.style
Gadfly.style — Method.
Set some attributes in the current Theme. See Theme for available field.
source

27.12 Gadfly.title
Gadfly.title — Method.

title(ctx::Context, str::String, props::Property...) -> Context

Add a title string to a group of plots, typically created with vstack, hstack,
or gridstack.
Examples

p1 = plot(x=[1,2], y=[3,4], Geom.line);

p2 = plot(x=[1,2], y=[4,3], Geom.line);
title(hstack(p1,p2), "my latest data", Compose.fontsize(18pt), fill(colorant"red"))

source
Chapter 28

IterTools

28.1 IterTools.chain
IterTools.chain — Method.

chain(xs...)

Iterate through any number of iterators in sequence.

julia> for i in chain(1:3, [’a’, ’b’, ’c’])

@show i
end
i = 1
i = 2
i = 3
i = ’a’
i = ’b’
i = ’c’

source

28.2 IterTools.distinct
IterTools.distinct — Method.

distinct(xs)

Iterate through values skipping over those already encountered.

julia> for i in distinct([1,1,2,1,2,4,1,2,3,4])

@show i
end
i = 1

344
28.3. ITERTOOLS.GROUPBY 345

i = 2
i = 4
i = 3

source

28.3 IterTools.groupby
IterTools.groupby — Method.

groupby(f, xs)

Group consecutive values that share the same result of applying f.

julia> for i in groupby(x -> x[1], ["face", "foo", "bar", "book", "baz", "zzz"])
@show i
end
i = String["face","foo"]
i = String["bar","book","baz"]
i = String["zzz"]

source

28.4 IterTools.imap
IterTools.imap — Method.

imap(f, xs1, [xs2, ...])

Iterate over values of a function applied to successive values from one or

more iterators.

julia> for i in imap(+, [1,2,3], [4,5,6])

@show i
end
i = 5
i = 7
i = 9

source

28.5 IterTools.iterate
IterTools.iterate — Method.

iterate(f, x)
346 CHAPTER 28. ITERTOOLS

Iterate over successive applications of f, as in x, f(x), f(f(x)), f(f(f(x))),

...
Use Base.take() to obtain the required number of elements.

julia> for i in take(iterate(x -> 2x, 1), 5)

@show i
end
i = 1
i = 2
i = 4
i = 8
i = 16

julia> for i in take(iterate(sqrt, 100), 6)

@show i
end
i = 100
i = 10.0
i = 3.1622776601683795
i = 1.7782794100389228
i = 1.333521432163324
i = 1.1547819846894583

source

28.6 IterTools.ncycle
IterTools.ncycle — Method.

ncycle(xs, n)

Cycle through iter n times.

julia> for i in ncycle(1:3, 2)

@show i
end
i = 1
i = 2
i = 3
i = 1
i = 2
i = 3

source
28.7. ITERTOOLS.NTH 347

28.7 IterTools.nth
IterTools.nth — Method.

nth(xs, n)

Return the nth element of xs. This is mostly useful for non-indexable col-
lections.

julia> mersenne = Set([3, 7, 31, 127])

Set([7,31,3,127])

julia> nth(mersenne, 3)
3

source

28.8 IterTools.partition
IterTools.partition — Method.

partition(xs, n, [step])

Group values into n-tuples.

julia> for i in partition(1:9, 3)

@show i
end
i = (1,2,3)
i = (4,5,6)
i = (7,8,9)

If the step parameter is set, each tuple is separated by step values.

julia> for i in partition(1:9, 3, 2)

@show i
end
i = (1,2,3)
i = (3,4,5)
i = (5,6,7)
i = (7,8,9)

julia> for i in partition(1:9, 3, 3)

@show i
end
i = (1,2,3)
i = (4,5,6)
348 CHAPTER 28. ITERTOOLS

i = (7,8,9)

julia> for i in partition(1:9, 2, 3)

@show i
end
i = (1,2)
i = (4,5)
i = (7,8)

source

28.9 IterTools.peekiter
IterTools.peekiter — Method.

peekiter(xs)

Lets you peek at the head element of an iterator without updating the state.

julia> it = peekiter(["face", "foo", "bar", "book", "baz", "zzz"])

IterTools.PeekIter{Array{String,1}}(String["face","foo","bar","book","baz","zzz"])

julia> s = start(it)
(2,Nullable{String}("face"))

julia> @show peek(it, s)

peek(it,s) = Nullable{String}("face")
Nullable{String}("face")

julia> @show peek(it, s)

peek(it,s) = Nullable{String}("face")
Nullable{String}("face")

julia> x, s = next(it, s)
("face",(3,Nullable{String}("foo"),false))

julia> @show x
x = "face"
"face"

julia> @show peek(it, s)

peek(it,s) = Nullable{String}("foo")
Nullable{String}("foo")

source
28.10. ITERTOOLS.PRODUCT 349

28.10 IterTools.product
IterTools.product — Method.

product(xs...)

Iterate over all combinations in the Cartesian product of the inputs.

julia> for p in product(1:3,4:5)

@show p
end
p = (1,4)
p = (2,4)
p = (3,4)
p = (1,5)
p = (2,5)
p = (3,5)

source

28.11 IterTools.repeatedly
IterTools.repeatedly — Method.

repeatedly(f, n)

Call function f n times, or infinitely if n is omitted.

[] julia¿ t() = (sleep(0.1); Dates.millisecond(now())) t (generic function with
1 method)
julia¿ collect(repeatedly(t, 5)) 5-element Array{Any,1}: 993 97 200 303 408
source

28.12 IterTools.subsets
IterTools.subsets — Method.

subsets(xs)
subsets(xs, k)

Iterate over every subset of the collection xs. You can restrict the subsets
to a specific size k.

julia> for i in subsets([1, 2, 3])

@show i
end
i = Int64[]
i = [1]
350 CHAPTER 28. ITERTOOLS

i = [2]
i = [1,2]
i = [3]
i = [1,3]
i = [2,3]
i = [1,2,3]

julia> for i in subsets(1:4, 2)

@show i
end
i = [1,2]
i = [1,3]
i = [1,4]
i = [2,3]
i = [2,4]
i = [3,4]

source

28.13 IterTools.takenth
IterTools.takenth — Method.

takenth(xs, n)

Iterate through every nth element of xs.

julia> collect(takenth(5:15,3))
3-element Array{Int64,1}:
7
10
13

source

28.14 IterTools.takestrict
IterTools.takestrict — Method.

takestrict(xs, n::Int)

Like take(), an iterator that generates at most the first n elements of xs,
but throws an exception if fewer than n items are encountered in xs.
28.14. ITERTOOLS.TAKESTRICT 351

julia> a = :1:2:11
1:2:11

julia> collect(takestrict(a, 3))

3-element Array{Int64,1}:
1
3
5

source
Chapter 29

Polynomials

29.1 Polynomials.coeffs
Polynomials.coeffs — Method.

coeffs(p::Poly)

Return the coefficient vector [a 0, a 1, ..., a n] of a polynomial p.

source

29.2 Polynomials.degree
Polynomials.degree — Method.

degree(p::Poly)

Return the degree of the polynomial p, i.e. the highest exponent in the
polynomial that has a nonzero coefficient. The degree of the zero polynomial is
defined to be -1.
source

29.3 Polynomials.poly
Polynomials.poly — Method.

poly(r)

Construct a polynomial from its roots. Compare this to the Poly type
constructor, which constructs a polynomial from its coefficients.
If r is a vector, the constructed polynomial is (x − r1 )(x − r2 ) · · · (x − rn ).
If r is a matrix, the constructed polynomial is (x − e1 ) · · · (x − en ), where ei is
the ith eigenvalue of r.

352
29.4. POLYNOMIALS.POLYDER 353

Examples
[] julia¿ poly([1, 2, 3]) The polynomial (x - 1)(x - 2)(x - 3) Poly(-6 + 11x -
6x2+x3)
julia¿ poly([1 2; 3 4]) The polynomial (x - 5.37228)(x + 0.37228) Poly(-
1.9999999999999998 - 5.0x + 1.0x2)
source

29.4 Polynomials.polyder
Polynomials.polyder — Method.

polyder(p::Poly, k=1)

Compute the kth derivative of the polynomial p.

Examples
[] julia¿ polyder(Poly([1, 3, -1])) Poly(3 - 2x)
julia¿ polyder(Poly([1, 3, -1]), 2) Poly(-2)
source

29.5 Polynomials.polyfit
Polynomials.polyfit — Function.

polyfit(x, y, n=length(x)-1, sym=:x)

Fit a polynomial of degree n through the points specified by x and y, where

n <= length(x) - 1, using least squares fit. When n=length(x)-1 (the de-
fault), the interpolating polynomial is returned. The optional fourth argument
can be used to specify the symbol for the returned polynomial.
Examples
[] julia¿ xs = linspace(0, pi, 5);
julia¿ ys = map(sin, xs);
julia¿ polyfit(xs, ys, 2) Poly(-0.004902082150108854 + 1.242031920509868x
- 0.39535103925413095x2)
source

29.6 Polynomials.polyint
Polynomials.polyint — Method.

polyint(p::Poly, a::Number, b::Number)

Compute the definite integral of the polynomial p over the interval [a,b].
Examples
[] julia¿ polyint(Poly([1, 0, -1]), 0, 1) 0.6666666666666667
source
354 CHAPTER 29. POLYNOMIALS

29.7 Polynomials.polyint
Polynomials.polyint — Method.

polyint(p::Poly, k::Number=0)

Integrate the polynomial p term by term, optionally adding a constant term

k. The order of the resulting polynomial is one higher than the order of p.
Examples
[] julia¿ polyint(Poly([1, 0, -1])) Poly(1.0x - 0.3333333333333333x3)
julia¿ polyint(Poly([1, 0, -1]), 2) Poly(2.0 + 1.0x - 0.3333333333333333x3)
source

29.8 Polynomials.polyval
Polynomials.polyval — Method.

polyval(p::Poly, x::Number)

Evaluate the polynomial p at x using Horner’s method. Poly objects are

callable, using this function.
Examples
[] julia¿ p = Poly([1, 0, -1]) Poly(1 - x2)
julia¿ polyval(p, 1) 0
julia¿ p(1) 0
source

29.9 Polynomials.printpoly
Polynomials.printpoly — Method.

printpoly(io::IO, p::Poly, mimetype = MIME"text/plain"(); descending_powers=false)

Print a human-readable representation of the polynomial p to io. The MIME

types “text/plain” (default), “text/latex”, and “text/html” are supported. By
default, the terms are in order of ascending powers, matching the order in
coeffs(p); specifying descending powers=true reverses the order.
Examples

julia> printpoly(stdout, Poly([1,2,3], :y))

1 + 2*y + 3*y^2
julia> printpoly(stdout, Poly([1,2,3], :y), descending_powers=true)
3*y^2 + 2*y + 1

source
29.10. POLYNOMIALS.ROOTS 355

29.10 Polynomials.roots
Polynomials.roots — Method.

roots(p::Poly)

Return the roots (zeros) of p, with multiplicity. The number of roots re-
turned is equal to the order of p. The returned roots may be real or complex.
Examples
[] julia¿ roots(Poly([1, 0, -1])) 2-element Array{Float64,1}: -1.0 1.0
julia¿ roots(Poly([1, 0, 1])) 2-element Array{Complex{Float64},1}: 0.0+1.0im
0.0-1.0im
julia¿ roots(Poly([0, 0, 1])) 2-element Array{Float64,1}: 0.0 0.0
julia¿ roots(poly([1,2,3,4])) 4-element Array{Float64,1}: 4.0 3.0 2.0 1.0
source

29.11 Polynomials.variable
Polynomials.variable — Method.

variable(p::Poly)
variable([T::Type,] var)
variable()

Return the indeterminate of a polynomial, i.e. its variable, as a Poly object.

When passed no arguments, this is equivalent to variable(Float64, :x).
Examples
[] julia¿ variable(Poly([1, 2], :x)) Poly(x)
julia¿ variable(:y) Poly(1.0y)
julia¿ variable() Poly(1.0x)
julia¿ variable(Float32, :x) Poly(1.0f0x)
source
Chapter 30

Colors

30.1 Base.hex
Base.hex — Method.

hex(c)

Print a color as a RGB hex triple, or a transparent paint as an ARGB hex

quadruplet.
source

30.2 Colors.MSC
Colors.MSC — Method.

MSC(h)
MSC(h, l)

Calculates the most saturated color for any given hue h by finding the cor-
responding corner in LCHuv space. Optionally, the lightness l may also be
specified.
source

30.3 Colors.colordiff
Colors.colordiff — Method.

colordiff(a, b)
colordiff(a, b, metric)

356
30.4. COLORS.COLORMAP 357

Compute an approximate measure of the perceptual difference between col-

ors a and b. Optionally, a metric may be supplied, chosen among DE 2000
(the default), DE 94, DE JPC79, DE CMC, DE BFD, DE AB, DE DIN99, DE DIN99d,
DE DIN99o.
source

30.4 Colors.colormap
Colors.colormap — Function.

colormap(cname, [N; mid, logscale, kvs...])

Returns a predefined sequential or diverging colormap computed using the

algorithm by Wijffelaars, M., et al. (2008). Sequential colormaps cname choices
are Blues, Greens, Grays, Oranges, Purples, and Reds. Diverging colormap
choices are RdBu. Optionally, you can specify the number of colors N (default
100). Keyword arguments include the position of the middle point mid (default
0.5) and the possibility to switch to log scaling with logscale (default false).
source

30.5 Colors.colormatch
Colors.colormatch — Method.

colormatch(wavelength)
colormatch(matchingfunction, wavelength)

Evaluate the CIE standard observer color match function.

Args:

• matchingfunction (optional): a type used to specify the matching func-

tion. Choices include CIE1931 CMF (the default, the CIE 1931 2 match-
ing function), CIE1964 CMF (the CIE 1964 10 color matching function),
CIE1931J CMF (Judd adjustment to CIE1931 CMF), CIE1931JV CMF (Judd-
Vos adjustment to CIE1931 CMF).

• wavelen: Wavelength of stimulus in nanometers.

Returns: XYZ value of perceived color.

source

30.6 Colors.deuteranopic
Colors.deuteranopic — Method.
358 CHAPTER 30. COLORS

deuteranopic(c)
deuteranopic(c, p)
Convert a color to simulate deuteranopic color deficiency (lack of the middle-
wavelength photopigment). See the description of protanopic for detail about
the arguments.
source

30.7 Colors.distinguishable colors

Colors.distinguishable colors — Method.
distinguishable_colors(n, [seed]; [transform, lchoices, cchoices, hchoices])
Generate n maximally distinguishable colors.
This uses a greedy brute-force approach to choose n colors that are maxi-
mally distinguishable. Given seed color(s), and a set of possible hue, chroma,
and lightness values (in LCHab space), it repeatedly chooses the next color as
the one that maximizes the minimum pairwise distance to any of the colors
already in the palette.
Args:
• n: Number of colors to generate.
• seed: Initial color(s) included in the palette. Default is Vector{RGB{N0f8}}(0).
Keyword arguments:
• transform: Transform applied to colors before measuring distance. De-
fault is the identity; other choices include deuteranopic to simulate color-
blindness.
• lchoices: Possible lightness values (default linspace(0,100,15))
• cchoices: Possible chroma values (default linspace(0,100,15))
• hchoices: Possible hue values (default linspace(0,340,20))
Returns: A Vector of colors of length n, of the type specified in seed.
source

30.8 Colors.protanopic
Colors.protanopic — Method.
protanopic(c)
protanopic(c, p)
Convert a color to simulate protanopic color deficiency (lack of the long-
wavelength photopigment). c is the input color; the optional argument p is the
fraction of photopigment loss, in the range 0 (no loss) to 1 (complete loss).
source
30.9. COLORS.TRITANOPIC 359

30.9 Colors.tritanopic
Colors.tritanopic — Method.

tritanopic(c)
tritanopic(c, p)

Convert a color to simulate tritanopic color deficiency (lack of the short-

wavelength photogiment). See protanopic for more detail about the arguments.
source

30.10 Colors.weighted color mean

Colors.weighted color mean — Method.

weighted_color_mean(w1, c1, c2)

Returns the color w1*c1 + (1-w1)*c2 that is the weighted mean of c1 and
c2, where c1 has a weight 0 w1 1.
source

30.11 Colors.whitebalance
Colors.whitebalance — Method.

whitebalance(c, src_white, ref_white)

Whitebalance a color.
Input a source (adopted) and destination (reference) white. E.g., if you have
a photo taken under florencent lighting that you then want to appear correct
under regular sunlight, you might do something like whitebalance(c, WP F2,
WP D65).
Args:

• c: An observed color.

• src white: Adopted or source white corresponding to c

• ref white: Reference or destination white.

Returns: A whitebalanced color.

source
Chapter 31

Interact

31.1 Interact.button
Interact.button — Method.

button(label; value=nothing, signal)

Create a push button. Optionally specify the label, the value emitted
when then button is clicked, and/or the (Reactive.jl) signal coupled to this
button.
source

31.2 Interact.checkbox
Interact.checkbox — Method.

checkbox(value=false; label="", signal)

Provide a checkbox with the specified starting (boolean) value. Optional

provide a label for this widget and/or the (Reactive.jl) signal coupled to this
widget.
source

31.3 Interact.dropdown
Interact.dropdown — Method.

dropdown(choices; label="", value, typ, icons, tooltips, signal)

Create a “dropdown” widget. choices can be a vector of options. Option-

ally specify the starting value (defaults to the first choice), the typ of elements

360
31.4. INTERACT.RADIOBUTTONS 361

in choices, supply custom icons, provide tooltips, and/or specify the (Re-
active.jl) signal coupled to this widget.
Examples

a = dropdown(["one", "two", "three"])

To link a callback to the dropdown, use

f = dropdown(["turn red"=>colorize_red, "turn green"=>colorize_green])

map(g->g(image), signal(f))

source

31.4 Interact.radiobuttons
Interact.radiobuttons — Method.
radiobuttons: see the help for dropdown
source

31.5 Interact.selection
Interact.selection — Method.
selection: see the help for dropdown
source

31.6 Interact.selection slider

Interact.selection slider — Method.
selection slider: see the help for dropdown If the slider has numeric (<:Real)
values, and its signal is updated, it will update to the nearest value from the
range/choices provided. To disable this behaviour, so that the widget state will
only update if an exact match for signal value is found in the range/choice, use
syncnearest=false.
source

31.7 Interact.set!
Interact.set! — Method.
set!(w::Widget, fld::Symbol, val)
Set the value of a widget property and update all displayed instances of the
widget. If val is a Signal, then updates to that signal will be reflected in
widget instances/views.
If fld is :value, val is also push!ed to signal(w)
source
362 CHAPTER 31. INTERACT

31.8 Interact.slider
Interact.slider — Method.

slider(range; value, signal, label="", readout=true, continuous_update=true)

Create a slider widget with the specified range. Optionally specify the
starting value (defaults to the median of range), provide the (Reactive.jl)
signal coupled to this slider, and/or specify a string label for the widget.
source

31.9 Interact.textarea
Interact.textarea — Method.

textarea(value=""; label="", signal)

Creates an extended text-entry area. Optionally provide a label and/or the

(Reactive.jl) signal associated with this widget. The signal updates when you
type.
source

31.10 Interact.textbox
Interact.textbox — Method.

textbox(value=""; label="", typ=typeof(value), range=nothing, signal)

Create a box for entering text. value is the starting value; if you don’t want
to provide an initial value, you can constrain the type with typ. Optionally pro-
vide a label, specify the allowed range (e.g., -10.0:10.0) for numeric entries,
and/or provide the (Reactive.jl) signal coupled to this text box.
source

31.11 Interact.togglebutton
Interact.togglebutton — Method.

togglebutton(label=""; value=false, signal)

Create a toggle button. Optionally specify the label, the initial state
(value=false is off, value=true is on), and/or provide the (Reactive.jl) signal
coupled to this button.
source
31.12. INTERACT.TOGGLEBUTTONS 363

31.12 Interact.togglebuttons
Interact.togglebuttons — Method.
togglebuttons: see the help for dropdown
source

31.13 Interact.update!
Interact.update! — Method.
Update output widgets
source

31.14 Interact.vselection slider

Interact.vselection slider — Method.
vselection slider(args...; kwargs...)
Shorthand for selection slider(args...; orientation="vertical", kwargs...)
source

31.15 Interact.vslider
Interact.vslider — Method.
vslider(args...; kwargs...)
Shorthand for slider(args...; orientation="vertical", kwargs...)
source
Chapter 32

FFTW

32.1 Base.DFT.FFTW.dct
Base.DFT.FFTW.dct — Function.

dct(A [, dims])

Performs a multidimensional type-II discrete cosine transform (DCT) of the

array A, using the unitary normalization of the DCT. The optional dims argu-
ment specifies an iterable subset of dimensions (e.g. an integer, range, tuple, or
array) to transform along. Most efficient if the size of A along the transformed
dimensions is a product of small primes; see nextprod. See also plan dct for
even greater efficiency.
source

32.2 Base.DFT.FFTW.dct!
Base.DFT.FFTW.dct! — Function.

dct!(A [, dims])

Same as dct!, except that it operates in-place on A, which must be an array

of real or complex floating-point values.
source

32.3 Base.DFT.FFTW.idct
Base.DFT.FFTW.idct — Function.

idct(A [, dims])

364
32.4. BASE.DFT.FFTW.IDCT! 365

Computes the multidimensional inverse discrete cosine transform (DCT) of

the array A (technically, a type-III DCT with the unitary normalization). The
optional dims argument specifies an iterable subset of dimensions (e.g. an inte-
ger, range, tuple, or array) to transform along. Most efficient if the size of A
along the transformed dimensions is a product of small primes; see nextprod.
See also plan idct for even greater efficiency.
source

32.4 Base.DFT.FFTW.idct!
Base.DFT.FFTW.idct! — Function.
idct!(A [, dims])
Same as idct!, but operates in-place on A.
source

32.5 Base.DFT.FFTW.plan dct

Base.DFT.FFTW.plan dct — Function.
plan_dct(A [, dims [, flags [, timelimit]]])
Pre-plan an optimized discrete cosine transform (DCT), similar to plan fft
except producing a function that computes dct. The first two arguments have
the same meaning as for dct.
source

32.6 Base.DFT.FFTW.plan dct!

Base.DFT.FFTW.plan dct! — Function.
plan_dct!(A [, dims [, flags [, timelimit]]])
Same as plan dct, but operates in-place on A.
source

32.7 Base.DFT.FFTW.plan idct

Base.DFT.FFTW.plan idct — Function.
plan_idct(A [, dims [, flags [, timelimit]]])
Pre-plan an optimized inverse discrete cosine transform (DCT), similar to
plan fft except producing a function that computes idct. The first two argu-
ments have the same meaning as for idct.
source
366 CHAPTER 32. FFTW

32.8 Base.DFT.FFTW.plan idct!

Base.DFT.FFTW.plan idct! — Function.

plan_idct!(A [, dims [, flags [, timelimit]]])

Same as plan idct, but operates in-place on A.

source

32.9 Base.DFT.FFTW.plan r2r

Base.DFT.FFTW.plan r2r — Function.

plan_r2r(A, kind [, dims [, flags [, timelimit]]])

Pre-plan an optimized r2r transform, similar to plan fft except that the
transforms (and the first three arguments) correspond to r2r and r2r!, respec-
tively.
source

32.10 Base.DFT.FFTW.plan r2r!

Base.DFT.FFTW.plan r2r! — Function.

plan_r2r!(A, kind [, dims [, flags [, timelimit]]])

Similar to plan fft, but corresponds to r2r!.

source

32.11 Base.DFT.FFTW.r2r
Base.DFT.FFTW.r2r — Function.

r2r(A, kind [, dims])

Performs a multidimensional real-input/real-output (r2r) transform of type

kind of the array A, as defined in the FFTW manual. kind specifies either
a discrete cosine transform of various types (FFTW.REDFT00, FFTW.REDFT01,
FFTW.REDFT10, or FFTW.REDFT11), a discrete sine transform of various types
(FFTW.RODFT00, FFTW.RODFT01, FFTW.RODFT10, or FFTW.RODFT11), a real-input
DFT with halfcomplex-format output (FFTW.R2HC and its inverse FFTW.HC2R),
or a discrete Hartley transform (FFTW.DHT). The kind argument may be an array
or tuple in order to specify different transform types along the different dimen-
sions of A; kind[end] is used for any unspecified dimensions. See the FFTW
manual for precise definitions of these transform types, at http://www.fftw.org/doc.
32.12. BASE.DFT.FFTW.R2R! 367

The optional dims argument specifies an iterable subset of dimensions (e.g. an

integer, range, tuple, or array) to transform along. kind[i] is then the trans-
form type for dims[i], with kind[end] being used for i > length(kind).
See also plan r2r to pre-plan optimized r2r transforms.
source

32.12 Base.DFT.FFTW.r2r!
Base.DFT.FFTW.r2r! — Function.

r2r!(A, kind [, dims])

Same as r2r, but operates in-place on A, which must be an array of real or

complex floating-point numbers.
source
Chapter 33

WebSockets

33.1 Base.close
Base.close — Method.
close(ws::WebSocket)
close(ws::WebSocket, statusnumber = n)
close(ws::WebSocket, statusnumber = n, freereason = "my reason")
Send an OPCODE CLOSE frame, and wait for the same response or until
a reasonable amount of time, 10.0 s, has passed. Data received while closing is
dropped. Status number n according to RFC 6455 7.4.1 can be included, see
WebSockets.codeDesc
source

33.2 Base.read
Base.read — Method.
read(ws::WebSocket)
Typical use: msg = String(read(ws)) Read one non-control message from a
WebSocket. Any control messages that are read will be handled by the han-
dle control frame function. Only the data (contents/body/payload) of the mes-
sage will be returned as a Vector{UInt8}.
This function will not return until a full non-control message has been read.
source

33.3 Base.write
Base.write — Method.
Write binary data; will be sent as one frame.
source

368
33.4. BASE.WRITE 369

33.4 Base.write
Base.write — Method.
Write text data; will be sent as one frame.
source

33.5 WebSockets.origin
WebSockets.origin — Function.
origin(request) => String Convenience function for checking which server
/ port address the client claims its code was downloaded from. The resource
path can be found with target(req). E.g.
[] function gatekeeper(req, ws) orig = WebSockets.origin(req) if startswith(orig,
”http://localhost”) —— startswith(orig, ”http://127.0.0.1”) handlewebsocket(ws)
end end end
source

33.6 WebSockets.readguarded
WebSockets.readguarded — Method.
readguarded(websocket) => (Vector, Bool)
Return (data::Vector, true) or (Vector{UInt8}(), false)
The peer can potentially disconnect at any time, but no matter the cause
you will usually just want to exit your websocket handling function when you
can’t write to it.
E.g.
[] while true data, success = readguarded(websocket) !success break
println(String(data)) end
source

33.7 WebSockets.send ping

WebSockets.send ping — Method.
Send a ping message, optionally with data.
source

33.8 WebSockets.send pong

WebSockets.send pong — Method.
Send a pong message, optionally with data.
source
370 CHAPTER 33. WEBSOCKETS

33.9 WebSockets.subprotocol
WebSockets.subprotocol — Function.
subprotocol(request) => String
Convenience function for reading upgrade request subprotocol. Acceptable
subprotocols need to be predefined using addsubproto(myprotocol). No other
subprotocols will pass the handshake. E.g.
[] WebSockets.addsubproto(”instructions”) WebSockets.addsubproto(”relayb ackend”)f unctiongateke

Then, in browser javascript (or equivalent with Julia WebSockets.open( , ))

[] function load(){ var wsuri = document.URL.replace(”http:”, ”ws:”); ws1
= new WebSocket(wsuri, ”instructions”); ws2 = new WebSocket(wsuri, ”relayb ackend”);ws1.onmessage=

source

33.10 WebSockets.target
WebSockets.target — Function.
target(request) => String
Convenience function for reading upgrade request target. E.g.
[] function gatekeeper(req, ws) if target(req) == ”/gamepad” @spawnat 2
gamepad(ws) elseif target(req) == ”/console” @spawnat 3 console(ws) ... end
end
Then, in browser javascript (or equivalent with Julia WebSockets.open( , ))
[] function load(){ var wsuri = document.URL.replace(”http:”, ”ws:”); ws1
= new WebSocket(wsuri + ”/gamepad”); ws2 = new WebSocket(wsuri +
”/console”); ws3 = new WebSocket(wsuri + ”/graphics”); ws4 = new Web-
Socket(wsuri + ”/audiochat”); ws1.onmessage = function(e){vibrate(e.data)}
} // load
source

33.11 WebSockets.writeguarded
WebSockets.writeguarded — Method.
writeguarded(websocket, message) => Bool
Return true if write is successful, false if not. The peer can potentially
disconnect at any time, but no matter the cause you will usually just want to
exit your websocket handling function when you can’t write to it.
source
Chapter 34

AxisArrays

34.1 AxisArrays.axes
AxisArrays.axes — Method.

axes(A::AxisArray) -> (Axis...)

axes(A::AxisArray, ax::Axis) -> Axis
axes(A::AxisArray, dim::Int) -> Axis

Returns the tuple of axis vectors for an AxisArray. If an specific Axis is

specified, then only that axis vector is returned. Note that when extracting a
single axis vector, axes(A, Axis{1})) is type-stable and will perform better
than axes(A)[1].
For an AbstractArray without Axis information, axes returns the default
axes, i.e., those that would be produced by AxisArray(A).
source

34.2 AxisArrays.axisdim
AxisArrays.axisdim — Method.

axisdim(::AxisArray, ::Axis) -> Int

axisdim(::AxisArray, ::Type{Axis}) -> Int

Given an AxisArray and an Axis, return the integer dimension of the Axis
within the array.
source

34.3 AxisArrays.axisnames
AxisArrays.axisnames — Method.

371
372 CHAPTER 34. AXISARRAYS

axisnames(A::AxisArray) -> (Symbol...)

axisnames(::Type{AxisArray{...}}) -> (Symbol...)
axisnames(ax::Axis...) -> (Symbol...)
axisnames(::Type{Axis{...}}...) -> (Symbol...)

Returns the axis names of an AxisArray or list of Axises as a tuple of Sym-

bols.
source

34.4 AxisArrays.axisvalues
AxisArrays.axisvalues — Method.

axisvalues(A::AxisArray) -> (AbstractVector...)

axisvalues(ax::Axis...) -> (AbstractVector...)

Returns the axis values of an AxisArray or list of Axises as a tuple of vectors.

source

34.5 AxisArrays.collapse
AxisArrays.collapse — Method.

collapse(::Type{Val{N}}, As::AxisArray...) -> AxisArray

collapse(::Type{Val{N}}, labels::Tuple, As::AxisArray...) -> AxisArray
collapse(::Type{Val{N}}, ::Type{NewArrayType}, As::AxisArray...) -> AxisArray
collapse(::Type{Val{N}}, ::Type{NewArrayType}, labels::Tuple, As::AxisArray...) -> Axis

Collapses AxisArrays with N equal leading axes into a single AxisArray. All
additional axes in any of the arrays are collapsed into a single additional axis
of type Axis{:collapsed, CategoricalVector{Tuple}}.
Arguments

• ::Type{Val{N}}: the greatest common dimension to share between all

input arrays. The remaining axes are collapsed. All N axes must be
common to each input array, at the same dimension. Values from 0 up to
the minimum number of dimensions across all input arrays are allowed.

• labels::Tuple: (optional) an index for each array in As used as the

leading element in the index tuples in the :collapsed axis. Defaults to
1:length(As).

• ::Type{NewArrayType<:AbstractArray{ , N+1}}: (optional) the desired

underlying array type for the returned AxisArray.

• As::AxisArray...: AxisArrays to be collapsed together.

34.5. AXISARRAYS.COLLAPSE 373

Examples

julia> price_data = AxisArray(rand(10), Axis{:time}(Date(2016,01,01):Day(1):Date(2016,01,10)))

1-dimensional AxisArray{Float64,1,...} with axes:
:time, 2016-01-01:1 day:2016-01-10
And data, a 10-element Array{Float64,1}:
0.885014
0.418562
0.609344
0.72221
0.43656
0.840304
0.455337
0.65954
0.393801
0.260207

julia> size_data = AxisArray(rand(10,2), Axis{:time}(Date(2016,01,01):Day(1):Date(2016,01,10)), A

2-dimensional AxisArray{Float64,2,...} with axes:
:time, 2016-01-01:1 day:2016-01-10
:measure, Symbol[:area, :volume]
And data, a 102 Array{Float64,2}:
0.159434 0.456992
0.344521 0.374623
0.522077 0.313256
0.994697 0.320953
0.95104 0.900526
0.921854 0.729311
0.000922581 0.148822
0.449128 0.761714
0.650277 0.135061
0.688773 0.513845

julia> collapsed = collapse(Val{1}, (:price, :size), price_data, size_data)

2-dimensional AxisArray{Float64,2,...} with axes:
:time, 2016-01-01:1 day:2016-01-10
:collapsed, Tuple{Symbol,Vararg{Symbol,N} where N}[(:price,), (:size, :area), (:size, :volume
And data, a 103 Array{Float64,2}:
0.885014 0.159434 0.456992
0.418562 0.344521 0.374623
0.609344 0.522077 0.313256
0.72221 0.994697 0.320953
0.43656 0.95104 0.900526
0.840304 0.921854 0.729311
0.455337 0.000922581 0.148822
0.65954 0.449128 0.761714
374 CHAPTER 34. AXISARRAYS

0.393801 0.650277 0.135061

0.260207 0.688773 0.513845

julia> collapsed[Axis{:collapsed}(:size)] == size_data

true

source
Chapter 35

BusinessDays

35.1 BusinessDays.advancebdays
BusinessDays.advancebdays — Method.
advancebdays(calendar, dt, bdays_count)
Increments given date dt by bdays count. Decrements it if bdays count is
negative. bdays count can be a Int, Vector{Int} or a UnitRange.
Computation starts by next Business Day if dt is not a Business Day.
source

35.2 BusinessDays.bdays
BusinessDays.bdays — Method.
bdays(calendar, dt0, dt1) :: Dates.Day
Counts the number of Business Days between dt0 and dt1. Returns in-
stances of Dates.Day.
Computation is always based on next Business Day if given dates are not
Business Days.
source

35.3 BusinessDays.bdayscount
BusinessDays.bdayscount — Method.
bdayscount(calendar, dt0, dt1) :: Int
Counts the number of Business Days between dt0 and dt1. Returns Int.
Computation is always based on next Business Day if given dates are not
Business Days.
source

375
376 CHAPTER 35. BUSINESSDAYS

35.4 BusinessDays.firstbdayofmonth
BusinessDays.firstbdayofmonth — Method.

firstbdayofmonth(calendar, dt)
firstbdayofmonth(calendar, yy, mm)

Returns the first business day of month.

source

35.5 BusinessDays.isbday
BusinessDays.isbday — Method.

isbday(calendar, dt)

Returns false for weekends or holidays. Returns true otherwise.

source

35.6 BusinessDays.isholiday
BusinessDays.isholiday — Method.

isholiday(calendar, dt)

Checks if dt is a holiday based on a given calendar of holidays.

calendar can be an instance of HolidayCalendar, a Symbol or an AbstractString.
Returns boolean values.
source

35.7 BusinessDays.isweekday
BusinessDays.isweekday — Method.

isweekday(dt)

Returns true for Monday to Friday. Returns false otherwise.

source

35.8 BusinessDays.isweekend
BusinessDays.isweekend — Method.

isweekend(dt)

Returns true for Saturdays or Sundays. Returns false otherwise.

source
35.9. BUSINESSDAYS.LASTBDAYOFMONTH 377

35.9 BusinessDays.lastbdayofmonth
BusinessDays.lastbdayofmonth — Method.

lastbdayofmonth(calendar, dt)
lastbdayofmonth(calendar, yy, mm)

Returns the last business day of month.

source

35.10 BusinessDays.listbdays
BusinessDays.listbdays — Method.

listbdays(calendar, dt0::Date, dt1::Date) Vector{Date}

Returns the list of business days between dt0 and dt1.

source

35.11 BusinessDays.listholidays
BusinessDays.listholidays — Method.

listholidays(calendar, dt0::Date, dt1::Date) Vector{Date}

Returns the list of holidays between dt0 and dt1.

source

35.12 BusinessDays.tobday
BusinessDays.tobday — Method.

tobday(calendar, dt; [forward=true])

Adjusts dt to next Business Day if it’s not a Business Day. If isbday(dt),

returns dt.
source
Chapter 36

QuadGK

36.1 Base.quadgk
Base.quadgk — Method.

quadgk(f, a,b,c...; rtol=sqrt(eps), atol=0, maxevals=10^7, order=7, norm=vecnorm)

Numerically integrate the function f(x) from a to b, and optionally over

additional intervals b to c and so on. Keyword options include a relative error
tolerance rtol (defaults to sqrt(eps) in the precision of the endpoints), an
absolute error tolerance atol (defaults to 0), a maximum number of function
evaluations maxevals (defaults to 10^7), and the order of the integration rule
(defaults to 7).
Returns a pair (I,E) of the estimated integral I and an estimated up-
per bound on the absolute error E. If maxevals is not exceeded then E <=
max(atol, rtol*norm(I)) will hold. (Note that it is useful to specify a posi-
tive atol in cases where norm(I) may be zero.)
The endpoints a et cetera can also be complex (in which case the integral is
performed over straight-line segments in the complex plane). If the endpoints
are BigFloat, then the integration will be performed in BigFloat precision as
well.
!!! note It is advisable to increase the integration order in rough proportion
to the precision, for smooth integrands.
More generally, the precision is set by the precision of the integration end-
points (promoted to floating-point types).
The integrand f(x) can return any numeric scalar, vector, or matrix type,
or in fact any type supporting +, -, multiplication by real values, and a norm
(i.e., any normed vector space). Alternatively, a different norm can be specified
by passing a norm-like function as the norm keyword argument (which defaults
to vecnorm).
!!! note Only one-dimensional integrals are provided by this function. For
multi-dimensional integration (cubature), there are many different algorithms

378
36.2. QUADGK.GAUSS 379

(often much better than simple nested 1d integrals) and the optimal choice
tends to be very problem-dependent. See the Julia external-package listing for
available algorithms for multidimensional integration or other specialized tasks
(such as integrals of highly oscillatory or singular functions).
The algorithm is an adaptive Gauss-Kronrod integration technique: the in-
tegral in each interval is estimated using a Kronrod rule (2*order+1 points) and
the error is estimated using an embedded Gauss rule (order points). The inter-
val with the largest error is then subdivided into two intervals and the process
is repeated until the desired error tolerance is achieved.
These quadrature rules work best for smooth functions within each interval,
so if your function has a known discontinuity or other singularity, it is best to
subdivide your interval to put the singularity at an endpoint. For example, if f
has a discontinuity at x=0.7 and you want to integrate from 0 to 1, you should
use quadgk(f, 0,0.7,1) to subdivide the interval at the point of discontinuity.
The integrand is never evaluated exactly at the endpoints of the intervals, so it
is possible to integrate functions that diverge at the endpoints as long as the
singularity is integrable (for example, a log(x) or 1/sqrt(x) singularity).
For real-valued endpoints, the starting and/or ending points may be infinite.
(A coordinate transformation is performed internally to map the infinite interval
to a finite one.)
source

36.2 QuadGK.gauss
QuadGK.gauss — Method.

gauss([T,] N)

Return a pair (x, w) of N quadrature points x[i] and weights w[i] to

integrate functions on the interval (-1, 1), i.e. sum(w .* f.(x)) approximates
the integral. Uses the method described in Trefethen & Bau, Numerical Linear
Algebra, to find the N-point Gaussian quadrature in O(N) operations.
T is an optional parameter specifying the floating-point type, defaulting to
Float64. Arbitrary precision (BigFloat) is also supported.
source

36.3 QuadGK.kronrod
QuadGK.kronrod — Method.

kronrod([T,] n)

Compute 2n+1 Kronrod points x and weights w based on the description

in Laurie (1997), appendix A, simplified for a=0, for integrating on [-1,1].
Since the rule is symmetric, this only returns the n+1 points with x <= 0. The
380 CHAPTER 36. QUADGK

function Also computes the embedded n-point Gauss quadrature weights gw

(again for x <= 0), corresponding to the points x[2:2:end]. Returns (x,w,wg)
in O(n) operations.
T is an optional parameter specifying the floating-point type, defaulting to
Float64. Arbitrary precision (BigFloat) is also supported.
Given these points and weights, the estimated integral I and error E can be
computed for an integrand f(x) as follows:

x, w, wg = kronrod(n)
fx = f(x[end]) # f(0)
x = x[1:end-1] # the x < 0 Kronrod points
fx = f.(x) .+ f.((-).(x)) # f(x < 0) + f(x > 0)
I = sum(fx .* w[1:end-1]) + fx * w[end]
if isodd(n)
E = abs(sum(fx[2:2:end] .* wg[1:end-1]) + fx*wg[end] - I)
else
E = abs(sum(fx[2:2:end] .* wg[1:end])- I)
end

source
Chapter 37

Dagger

37.1 Dagger.alignfirst
Dagger.alignfirst — Method.

alignfirst(a)

Make a subdomain a standalone domain. For example,

alignfirst(ArrayDomain(11:25, 21:100))
# => ArrayDomain((1:15), (1:80))

source

37.2 Dagger.cached stage

Dagger.cached stage — Method.
A memoized version of stage. It is important that the tasks generated for
the same DArray have the same identity, for example:

A = rand(Blocks(100,100), Float64, 1000, 1000)

compute(A+A’)

must not result in computation of A twice.

source

37.3 Dagger.compute
Dagger.compute — Method.
A DArray object may contain a thunk in it, in which case we first turn it
into a Thunk object and then compute it.
source

381
382 CHAPTER 37. DAGGER

37.4 Dagger.compute
Dagger.compute — Method.
Compute a Thunk - creates the DAG, assigns ranks to nodes for tie breaking
and runs the scheduler.
source

37.5 Dagger.domain
Dagger.domain — Function.

domain(x::T)

Returns metadata about x. This metadata will be in the domain field of a

Chunk object when an object of type T is created as the result of evaluating a
Thunk.
source

37.6 Dagger.domain
Dagger.domain — Method.
The domain of an array is a ArrayDomain
source

37.7 Dagger.domain
Dagger.domain — Method.
If no domain method is defined on an object, then we use the UnitDomain
on it. A UnitDomain is indivisible.
source

37.8 Dagger.load
Dagger.load — Method.

load(ctx, file_path)

Load an Union{Chunk, Thunk} from a file.

source
37.9. DAGGER.LOAD 383

37.9 Dagger.load
Dagger.load — Method.

load(ctx, ::Type{Chunk}, fpath, io)

Load a Chunk object from a file, the file path is required for creating a
FileReader object
source

37.10 Dagger.save
Dagger.save — Method.

save(io::IO, val)

Save a value into the IO buffer. In the case of arrays and sparse matrices,
this will save it in a memory-mappable way.
load(io::IO, t::Type, domain) will load the object given its domain
source

37.11 Dagger.save
Dagger.save — Method.

save(ctx, chunk::Union{Chunk, Thunk}, file_path::AbsractString)

Save a chunk to a file at file path.

source

37.12 Dagger.save
Dagger.save — Method.
special case distmem writing - write to disk on the process with the chunk.
source
Chapter 38

NNlib

38.1 NNlib.elu
NNlib.elu — Function.

elu(x, = 1) =
x > 0 ? x : * (exp(x) - 1)

Exponential Linear Unit activation function. See Fast and Accurate Deep
Network Learning by Exponential Linear Units. You can also specify the coef-
ficient explicitly, e.g. elu(x, 1).
source

38.2 NNlib.leakyrelu
NNlib.leakyrelu — Function.

leakyrelu(x) = max(0.01x, x)

Leaky Rectified Linear Unit activation function. You can also specify the
coefficient explicitly, e.g. leakyrelu(x, 0.01).
source

38.3 NNlib.logsoftmax
NNlib.logsoftmax — Function.

logsoftmax(xs) = log.(exp.(xs) ./ sum(exp.(xs)))

logsoftmax computes the log of softmax(xs) and it is more numerically stable

than softmax function in computing the cross entropy loss.
source

384
38.4. NNLIB.LOG 385

38.4 NNlib.log
NNlib.log — Method.

log(x)

Return log((x)) which is computed in a numerically stable way.

julia> log(0.)
-0.6931471805599453
julia> log.([-100, -10, 100.])
3-element Array{Float64,1}:
-100.0
-10.0
-0.0

source

38.5 NNlib.relu
NNlib.relu — Method.

relu(x) = max(0, x)

Rectified Linear Unit activation function.

source

38.6 NNlib.selu
NNlib.selu — Method.

selu(x) = * (x 0 ? x : * (exp(x) - 1))

1.0507
1.6733

Scaled exponential linear units. See Self-Normalizing Neural Networks.

source

38.7 NNlib.softmax
NNlib.softmax — Function.

softmax(xs) = exp.(xs) ./ sum(exp.(xs))

386 CHAPTER 38. NNLIB

Softmax takes log-probabilities (any real vector) and returns a probability

distribution that sums to 1.
If given a matrix it will treat it as a batch of vectors, with each column
independent.
julia> softmax([1,2,3.])
3-element Array{Float64,1}:
0.0900306
0.244728
0.665241
source

38.8 NNlib.softplus
NNlib.softplus — Method.
softplus(x) = log(exp(x) + 1)
See Deep Sparse Rectifier Neural Networks.
source

38.9 NNlib.softsign
NNlib.softsign — Method.
softsign(x) = x / (1 + |x|)
See Quadratic Polynomials Learn Better Image Features.
source

38.10 NNlib.swish
NNlib.swish — Method.
swish(x) = x * (x)
Self-gated actvation function. See Swish: a Self-Gated Activation Function.
source

38.11 NNlib.
NNlib. — Method.
(x) = 1 / (1 + exp(-x))
Classic sigmoid activation function.
source
Chapter 39

ShowItLikeYouBuildIt

39.1 ShowItLikeYouBuildIt.showarg
ShowItLikeYouBuildIt.showarg — Method.

showarg(stream::IO, x)

Show x as if it were an argument to a function. This function is used in the

printing of “type summaries” in terms of sequences of function calls on objects.
The fallback definition is to print x as ::$(typeof(x)), representing argu-
ment x in terms of its type. However, you can specialize this function for specific
types to customize printing.
Example
A SubArray created as view(a, :, 3, 2:5), where a is a 3-dimensional
Float64 array, has type

SubArray{Float64,2,Array{Float64,3},Tuple{Colon,Int64,UnitRange{Int64}},false}

and this type will be printed in the summary. To change the printing of this
object to

view(::Array{Float64,3}, Colon(), 3, 2:5)

you could define

function ShowItLikeYouBuildIt.showarg(io::IO, v::SubArray)

print(io, "view(")
showarg(io, parent(v))
print(io, ", ", join(v.indexes, ", "))
print(io, ’)’)
end

387
388 CHAPTER 39. SHOWITLIKEYOUBUILDIT

Note that we’re calling showarg recursively for the parent array type. Print-
ing the parent as ::Array{Float64,3} is the fallback behavior, assuming no
specialized method for Array has been defined. More generally, this would
display as

view(<a>, Colon(), 3, 2:5)

where <a> is the output of showarg for a.

This printing might be activated any time v is a field in some other container,
or if you specialize Base.summary for SubArray to call summary build.
See also: summary build.
source

39.2 ShowItLikeYouBuildIt.summary build

ShowItLikeYouBuildIt.summary build — Function.

summary_build(A::AbstractArray, [cthresh])

Return a string representing A in terms of the sequence of function calls that

might be used to create A, along with information about A’s size or indices and
element type. This function should never be called directly, but instead used to
specialize Base.summary for specific AbstractArray subtypes. For example, if
you want to change the summary of SubArray, you might define

Base.summary(v::SubArray) = summary_build(v)

This function goes hand-in-hand with showarg. If you have defined a showarg
method for SubArray as in the documentation for showarg, then the summary
of a SubArray might look like this:

34 view(::Array{Float64,3}, :, 3, 2:5) with element type Float64

instead of this:

34 SubArray{Float64,2,Array{Float64,3},Tuple{Colon,Int64,UnitRange{Int64}},false}

The optional argument cthresh is a “complexity threshold”; objects with

type descriptions that are less complex than the specified threshold will be
printed using the traditional type-based summary. The default value is n+1,
where n is the number of parameters in typeof(A). The complexity is calculated
with type complexity. You can choose a cthresh of 0 if you want to ensure
that your showarg version is always used.
See also: showarg, type complexity.
source
39.3. SHOWITLIKEYOUBUILDIT.TYPE COMPLEXITY 389

39.3 ShowItLikeYouBuildIt.type complexity

ShowItLikeYouBuildIt.type complexity — Method.

type_complexity(T::Type) -> c

Return an integer c representing a measure of the “complexity” of type T.

For unnested types, c = n+1 where n is the number of parameters in type T.
However, type complexity calls itself recursively on the parameters, so if the
parameters have their own parameters then the complexity of the type will be
(potentially much) higher than this.
Examples

# Array{Float64,2}
julia> a = rand(3,5);

julia> type_complexity(typeof(a))
3

julia> length(typeof(a).parameters)+1
3

# Create an object that has type:

# SubArray{Int64,2,Base.ReshapedArray{Int64,2,UnitRange{Int64},Tuple{}},Tuple{StepRange{Int64,
julia> r = reshape(1:9, 3, 3);

julia> v = view(r, 1:2:3, :);

julia> type_complexity(typeof(v))
15

julia> length(typeof(v).parameters)+1
6

The second example indicates that the total complexity of v’s type is con-
siderably higher than the complexity of just its “outer” SubArray type.
source
Chapter 40

CoordinateTransformations

40.1 CoordinateTransformations.cameramap
CoordinateTransformations.cameramap — Method.

cameramap()
cameramap(scale)
cameramap(scale, offset)

Create a transformation that takes points in real space (e.g. 3D) and projects
them through a perspective transformation onto the focal plane of an ideal
(pinhole) camera with the given properties.
The scale sets the scale of the screen. For a standard digital camera, this
would be scale = focal length / pixel size. Non-square pixels are sup-
ported by providing a pair of scales in a tuple, scale = (scale x, scale y).
Positive scales represent a camera looking in the +z axis with a virtual screen
in front of the camera (the x,y coordinates are not inverted compared to 3D
space). Note that points behind the camera (with negative z component) will
be projected (and inverted) onto the image coordinates and it is up to the user
to cull such points as necessary.
The offset = (offset x, offset y) is used to define the origin in the
imaging plane. For instance, you may wish to have the point (0,0) represent the
top-left corner of your imaging sensor. This measurement is in the units after
applying scale (e.g. pixels).
(see also PerspectiveMap)
source

40.2 CoordinateTransformations.compose
CoordinateTransformations.compose — Method.

compose(trans1, trans2)

390
40.3. COORDINATETRANSFORMATIONS.RECENTER 391

trans1 trans2

Take two transformations and create a new transformation that is equivalent

to successively applying trans2 to the coordinate, and then trans1. By default
will create a ComposedTransformation, however this method can be overloaded
for efficiency (e.g. two affine transformations naturally compose to a single affine
transformation).
source

40.3 CoordinateTransformations.recenter
CoordinateTransformations.recenter — Method.

recenter(trans::Union{AbstractMatrix,Transformation}, origin::AbstractVector) -> ctrans

Return a new transformation ctrans such that point origin serves as the
origin-of-coordinates for trans. Translation by origin occurs both before and
after applying trans, so that if trans is linear we have

ctrans(origin) == origin

As a consequence, recenter only makes sense if the output space of trans

is isomorphic with the input space.
For example, if trans is a rotation matrix, then ctrans rotates space around
origin.
source

40.4 CoordinateTransformations.transform deriv

CoordinateTransformations.transform deriv — Method.

transform_deriv(trans::Transformation, x)

A matrix describing how differentials on the parameters of x flow through

to the output of transformation trans.
source

40.5 CoordinateTransformations.transform deriv params

CoordinateTransformations.transform deriv params — Method.

transform_deriv_params(trans::AbstractTransformation, x)

A matrix describing how differentials on the parameters of trans flow through

to the output of transformation trans given input x.
source
Chapter 41

Graphics

41.1 Graphics.aspect ratio

Graphics.aspect ratio — Method.
aspect_ratio(bb::BoundingBox) -> r
Compute the ratio r of the height and width of bb.
source

41.2 Graphics.deform
Graphics.deform — Method.
deform(bb::BoundingBox, l, r, t, b) -> bbnew
Add l (left), r (right), t (top), and b (bottom) to the edges of a Bounding-
Box.
source

41.3 Graphics.inner canvas

Graphics.inner canvas — Method.
inner_canvas(c::GraphicsContext, device::BoundingBox, user::BoundingBox)
inner_canvas(c::GraphicsContext, x, y, w, h, l, r, t, b)
Create a rectangular drawing area inside device (represented in device-
coordinates), giving it user-coordinates user. Any drawing that occurs outside
this box is clipped.
x, y, w, and h are an alternative parametrization of device, and l, r, t, b
parametrize user.
See also: set coordinates.
source

392
41.4. GRAPHICS.ISINSIDE 393

41.4 Graphics.isinside
Graphics.isinside — Method.

isinside(bb::BoundingBox, p::Point) -> tf::Bool

isinside(bb::BoundingBox, x, y) -> tf::Bool

Determine whether the point lies within bb.

source

41.5 Graphics.rotate
Graphics.rotate — Method.

rotate(bb::BoundingBox, angle, o) -> bbnew

Rotate bb around o by angle, returning the BoundingBox that encloses the

vertices of the rotated box.
source

41.6 Graphics.rotate
Graphics.rotate — Method.

rotate(p::Vec2, angle::Real, o::Vec2) -> pnew

Rotate p around o by angle.

source

41.7 Graphics.set coordinates

Graphics.set coordinates — Method.

set_coordinates(c::GraphicsContext, device::BoundingBox, user::BoundingBox)

set_coordinates(c::GraphicsContext, user::BoundingBox)

Set the device->user coordinate transformation of c so that device, ex-

pressed in “device coordinates” (pixels), is equivalent to user as expressed in
“user coordinates”. If device is omitted, it defaults to the full span of c,
BoundingBox(0, width(c), 0, height(c)).
See also get matrix, set matrix.
source
394 CHAPTER 41. GRAPHICS

41.8 Graphics.shift
Graphics.shift — Method.

shift(bb::BoundingBox, x, y) -> bbnew

Shift center by (x,y), keeping width & height fixed.

source
Chapter 42

MacroTools

42.1 MacroTools.inexpr
MacroTools.inexpr — Method.

inexpr(expr, x)

Simple expression match; will return true if the expression x can be found
inside expr.

inexpr(:(2+2), 2) == true

source

42.2 MacroTools.isdef
MacroTools.isdef — Method.
Test for function definition expressions.
source

42.3 MacroTools.isexpr
MacroTools.isexpr — Method.

isexpr(x, ts...)

Convenient way to test the type of a Julia expression. Expression heads and
types are supported, so for example you can call

isexpr(expr, String, :string)

to pick up on all string-like expressions.

source

395
396 CHAPTER 42. MACROTOOLS

42.4 MacroTools.namify
MacroTools.namify — Method.
An easy way to get pull the (function/type) name out of expressions like
foo{T} or Bar{T} <: Vector{T}.
source

42.5 MacroTools.prettify
MacroTools.prettify — Method.

prettify(ex)

Makes generated code generaly nicer to look at.

source

42.6 MacroTools.rmlines
MacroTools.rmlines — Method.

rmlines(x)

Remove the line nodes from a block or array of expressions.

Compare quote end vs rmlines(quote end)
source

42.7 MacroTools.splitarg
MacroTools.splitarg — Method.

splitarg(arg)

Match function arguments (whether from a definition or a function call) such

as x::Int=2 and return (arg name, arg type, is splat, default). arg name
and default are nothing when they are absent. For example:
[] ¿ map(splitarg, (:(f(a=2, x::Int=nothing, y, args...))).args[2:end]) 4-element
Array{Tuple{Symbol,Symbol,Bool,Any},1}: (:a, :Any, false, 2) (:x, :Int, false,
:nothing) (:y, :Any, false, nothing) (:args, :Any, true, nothing)
source
42.8. MACROTOOLS.SPLITDEF 397

42.8 MacroTools.splitdef
MacroTools.splitdef — Method.

splitdef(fdef)

Match any function definition

[] function name{params}(args; kwargs)::rtype where {whereparams} body
end
and return Dict(:name=>..., :args=>..., etc.). The definition can be
rebuilt by calling MacroTools.combinedef(dict), or explicitly with

rtype = get(dict, :rtype, :Any)

all_params = [get(dict, :params, [])..., get(dict, :whereparams, [])...]
:(function $(dict[:name]){$(all_params...)}($(dict[:args]...);
$(dict[:kwargs]...))::$rtype
$(dict[:body])
end)

source

42.9 MacroTools.unblock
MacroTools.unblock — Method.

unblock(expr)

Remove outer begin blocks from an expression, if the block is redundant

(i.e. contains only a single expression).
source
Chapter 43

ImageMorphology

43.1 ImageMorphology.bothat
ImageMorphology.bothat — Function.
imgbh = bothat(img, [region]) performs bottom hat of an image, which
is defined as its morphological closing minus the original image. region allows
you to control the dimensions over which this operation is performed.
source

43.2 ImageMorphology.closing
ImageMorphology.closing — Function.
imgc = closing(img, [region]) performs the closing morphology oper-
ation, equivalent to erode(dilate(img)). region allows you to control the
dimensions over which this operation is performed.
source

43.3 ImageMorphology.dilate
ImageMorphology.dilate — Function.

imgd = dilate(img, [region])

perform a max-filter over nearest-neighbors. The default is 8-connectivity

in 2d, 27-connectivity in 3d, etc. You can specify the list of dimensions that
you want to include in the connectivity, e.g., region = [1,2] would exclude
the third dimension from filtering.
source

398
43.4. IMAGEMORPHOLOGY.ERODE 399

43.4 ImageMorphology.erode
ImageMorphology.erode — Function.
imge = erode(img, [region])
perform a min-filter over nearest-neighbors. The default is 8-connectivity in
2d, 27-connectivity in 3d, etc. You can specify the list of dimensions that you
want to include in the connectivity, e.g., region = [1,2] would exclude the
third dimension from filtering.
source

43.5 ImageMorphology.morphogradient
ImageMorphology.morphogradient — Function.
imgmg = morphogradient(img, [region]) returns morphological gradient
of the image, which is the difference between the dilation and the erosion of
a given image. region allows you to control the dimensions over which this
operation is performed.
source

43.6 ImageMorphology.morpholaplace
ImageMorphology.morpholaplace — Function.
imgml = morpholaplace(img, [region]) performs Morphological Laplacian
of an image, which is defined as the arithmetic difference between the internal
and the external gradient. region allows you to control the dimensions over
which this operation is performed.
source

43.7 ImageMorphology.opening
ImageMorphology.opening — Function.
imgo = opening(img, [region]) performs the opening morphology oper-
ation, equivalent to dilate(erode(img)). region allows you to control the
dimensions over which this operation is performed.
source

43.8 ImageMorphology.tophat
ImageMorphology.tophat — Function.
imgth = tophat(img, [region]) performs top hat of an image, which is
defined as the image minus its morphological opening. region allows you to
control the dimensions over which this operation is performed.
source
Chapter 44

NLSolversBase

44.1 Base.LinAlg.gradient
Base.LinAlg.gradient — Method.
Evaluates the gradient value at x
This does not update obj.DF or obj.x df.
source

44.2 Base.LinAlg.gradient
Base.LinAlg.gradient — Method.
Get the ith element of the most recently evaluated gradient of obj.
source

44.3 Base.LinAlg.gradient
Base.LinAlg.gradient — Method.
Get the most recently evaluated gradient of obj.
source

44.4 NLSolversBase.gradient!
NLSolversBase.gradient! — Method.
Evaluates the gradient value at x.
Stores the value in obj.DF.
source

400
44.5. NLSOLVERSBASE.HESSIAN 401

44.5 NLSolversBase.hessian
NLSolversBase.hessian — Method.
Get the most recently evaluated Hessian of obj
source

44.6 NLSolversBase.jacobian
NLSolversBase.jacobian — Method.
Get the most recently evaluated Jacobian of obj.
source

44.7 NLSolversBase.value!!
NLSolversBase.value!! — Method.
Force (re-)evaluation of the objective value at x.
Returns f(x) and stores the value in obj.F
source

44.8 NLSolversBase.value!
NLSolversBase.value! — Method.
Evaluates the objective value at x.
Returns f(x) and stores the value in obj.F
source

44.9 NLSolversBase.value
NLSolversBase.value — Method.
Evaluates the objective value at x.
Returns f(x), but does not store the value in obj.F
source

44.10 NLSolversBase.value
NLSolversBase.value — Method.
Get the most recently evaluated objective value of obj.
source
Chapter 45

Contour

45.1 Contour.contour
Contour.contour — Method.
contour(x, y, z, level::Number) Trace a single contour level, indicated
by the argument level.
You’ll usually call lines on the output of contour, and then iterate over
the result.
source

45.2 Contour.contours
Contour.contours — Method.
contours(x,y,z,levels) Trace the contour levels indicated by the levels
argument.
source

45.3 Contour.contours
Contour.contours — Method.
contours returns a set of isolines.
You’ll usually call levels on the output of contours.
source

45.4 Contour.contours
Contour.contours — Method.
contours(x,y,z,Nlevels::Integer) Trace Nlevels contour levels at heights
chosen by the library (using the contourlevels function).
source

402
45.5. CONTOUR.CONTOURS 403

45.5 Contour.contours
Contour.contours — Method.
contours(x,y,z) Trace 10 automatically chosen contour levels.
source

45.6 Contour.coordinates
Contour.coordinates — Method.
coordinates(c) Returns the coordinates of the vertices of the contour line
as a tuple of lists.
source

45.7 Contour.level
Contour.level — Method.
level(c) Indicates the z-value at which the contour level c was traced.
source

45.8 Contour.levels
Contour.levels — Method.
Turns the output of contours into an iterable with each of the traced contour
levels. Each of the objects support level and coordinates.
source

45.9 Contour.lines
Contour.lines — Method.
lines(c) Extracts an iterable collection of isolines from a contour level. Use
coordinates to get the coordinates of a line.
source
Chapter 46

Compose

46.1 Compose.circle
Compose.circle — Function.

circle(x, y, r)

Define a circle with its center at (x,y) and a radius of r.

source

46.2 Compose.circle
Compose.circle — Function.

circle(xs, ys, rs)

Arguments can be passed in arrays in order to perform multiple drawing

operations.
source

46.3 Compose.circle
Compose.circle — Method.

circle()

Define a circle in the center of the current context with a diameter equal to
the width of the context.
source

404
46.4. COMPOSE.POLYGON 405

46.4 Compose.polygon
Compose.polygon — Method.
polygon(points)
Define a polygon. points is an array of (x,y) tuples that specify the corners
of the polygon.
source

46.5 Compose.rectangle
Compose.rectangle — Function.
rectangle(x0, y0, width, height)
Define a rectangle of size widthxheight with its top left corner at the point
(x, y).
source

46.6 Compose.rectangle
Compose.rectangle — Function.
rectangle(x0s, y0s, widths, heights)
Arguments can be passed in arrays in order to perform multiple drawing
operations at once.
source

46.7 Compose.rectangle
Compose.rectangle — Method.
rectangle()
Define a rectangle that fills the current context completely.
source

46.8 Compose.text
Compose.text — Function.
text(xs, ys, values [,haligns::HAlignment [,valigns::VAlignment [,rots::Rotation]]])
Arguments can be passed in arrays in order to perform multiple drawing
operations at once.
source
406 CHAPTER 46. COMPOSE

46.9 Compose.text
Compose.text — Function.

text(x, y, value [,halign::HAlignment [,valign::VAlignment [,rot::Rotation]]])

Draw the text value at the position (x,y) relative to the current context.
The default alignment of the text is hleft vbottom. The vertical and hor-
izontal alignment is specified by passing hleft, hcenter or hright and vtop,
vcenter or vbottom as values for halgin and valgin respectively.
source
Chapter 47

CoupledFields

47.1 CoupledFields.CVfn
CoupledFields.CVfn — Method.

CVfn{T<:Matrix{Float64}}(parm::T, X::T, Y::T, modelfn::Function, kerneltype::DataType; verbose::B

Cross-validation function
source

47.2 CoupledFields.Rsq adj

CoupledFields.Rsq adj — Method.

Rsq_adj{T<:Array{Float64}}(Tx::T, Ty::T, df::Int):

Cross-validation metric
source

47.3 CoupledFields.bf
CoupledFields.bf — Method.

bf(x::Vector{Float64}, df::Int):

Compute a piecewise linear basis matrix for the vector x.

source

407
408 CHAPTER 47. COUPLEDFIELDS

47.4 CoupledFields.cca
CoupledFields.cca — Method.

cca{T<:Matrix{Float64}}(v::Array{Float64}, X::T,Y::T):

Regularized Canonical Correlation Analysis using SVD.

source

47.5 CoupledFields.gKCCA
CoupledFields.gKCCA — Method.

gKCCA(par::Array{Float64}, X::Matrix{Float64}, Y::Matrix{Float64}, kpars::KernelParamet

Compute the projection matrices and components for gKCCA.

source

47.6 CoupledFields.gradvecfield
CoupledFields.gradvecfield — Method.

gradvecfield{N<:Float64, T<:Matrix{Float64}}(par::Array{N}, X::T, Y::T, kpars::KernelPa

Compute the gradient vector or gradient matrix at each instance of the X

and Y fields, by making use of a kernel feature space.
source

47.7 CoupledFields.whiten
CoupledFields.whiten — Method.

whiten(x::Matrix{Float64}, d::Float64; lat=nothing): Whiten matrix

d (0-1) Percentage variance of components to retain.

lat Latitudinal area-weighting.
source
Chapter 48

AxisAlgorithms

48.1 AxisAlgorithms.A ldiv B md!

AxisAlgorithms.A ldiv B md! — Method.
A ldiv B md!(dest, F, src, dim) solves a tridiagonal system along di-
mension dim of src, storing the result in dest. Currently, F must be an LU-
factorized tridiagonal matrix. If desired, you may safely use the same array for
both src and dest, so that this becomes an in-place algorithm.
source

48.2 AxisAlgorithms.A ldiv B md

AxisAlgorithms.A ldiv B md — Method.
A ldiv B md(F, src, dim) solves F for slices b of src along dimension dim,
storing the result along the same dimension of the output. Currently, F must
be an LU-factorized tridiagonal matrix or a Woodbury matrix.
source

48.3 AxisAlgorithms.A mul B md!

AxisAlgorithms.A mul B md! — Method.
A mul B md!(dest, M, src, dim) computes M*x for slices x of src along
dimension dim, storing the result in dest. M must be an AbstractMatrix. This
uses an in-place naive algorithm.
source

48.4 AxisAlgorithms.A mul B md

AxisAlgorithms.A mul B md — Method.

409
410 CHAPTER 48. AXISALGORITHMS

A mul B md(M, src, dim) computes M*x for slices x of src along dimension
dim, storing the resulting vector along the same dimension of the output. M
must be an AbstractMatrix. This uses an in-place naive algorithm.
source

48.5 AxisAlgorithms.A mul B perm!

AxisAlgorithms.A mul B perm! — Method.
A mul B perm!(dest, M, src, dim) computes M*x for slices x of src along
dimension dim, storing the result in dest. M must be an AbstractMatrix. This
uses permutedims to make dimension dim into the first dimension, performs a
standard matrix multiplication, and restores the original dimension ordering.
In many cases, this algorithm exhibits the best cache behavior.
source

48.6 AxisAlgorithms.A mul B perm

AxisAlgorithms.A mul B perm — Method.
A mul B perm(M, src, dim) computes M*x for slices x of src along dimen-
sion dim, storing the resulting vector along the same dimension of the output.
M must be an AbstractMatrix. This uses permutedims to make dimension dim
into the first dimension, performs a standard matrix multiplication, and restores
the original dimension ordering. In many cases, this algorithm exhibits the best
cache behavior.
source
Chapter 49

Libz

49.1 Libz.ZlibDeflateInputStream
Libz.ZlibDeflateInputStream — Method.

ZlibDeflateInputStream(input[; <keyword arguments>])

Construct a zlib deflate input stream to compress gzip/zlib data.

Arguments

• input: a byte vector, IO object, or BufferedInputStream containing data

to compress.
• bufsize::Integer=8192: input and output buffer size.
• gzip::Bool=true: if true, data is gzip compressed; if false, zlib com-
pressed.
• level::Integer=6: compression level in 1-9.
• mem level::Integer=8: memory to use for compression in 1-9.
• strategy=Z DEFAULT STRATEGY: compression strategy; see zlib documen-
tation.

source

49.2 Libz.ZlibDeflateOutputStream
Libz.ZlibDeflateOutputStream — Method.

ZlibDeflateOutputStream(output[; <keyword arguments>])

Construct a zlib deflate output stream to compress gzip/zlib data.

Arguments

411
412 CHAPTER 49. LIBZ

• output: a byte vector, IO object, or BufferedInputStream to which com-

pressed data should be written.
• bufsize::Integer=8192: input and output buffer size.
• gzip::Bool=true: if true, data is gzip compressed; if false, zlib com-
pressed.
• level::Integer=6: compression level in 1-9.
• mem level::Integer=8: memory to use for compression in 1-9.
• strategy=Z DEFAULT STRATEGY: compression strategy; see zlib documen-
tation.

source

49.3 Libz.ZlibInflateInputStream
Libz.ZlibInflateInputStream — Method.

ZlibInflateInputStream(input[; <keyword arguments>])

Construct a zlib inflate input stream to decompress gzip/zlib data.

Arguments

• input: a byte vector, IO object, or BufferedInputStream containing com-

pressed data to inflate.
• bufsize::Integer=8192: input and output buffer size.
• gzip::Bool=true: if true, data is gzip compressed; if false, zlib com-
pressed.
• reset on end::Bool=true: on stream end, try to find the start of another
stream.

source

49.4 Libz.ZlibInflateOutputStream
Libz.ZlibInflateOutputStream — Method.

ZlibInflateOutputStream(output[; <keyword arguments>])

Construct a zlib inflate output stream to decompress gzip/zlib data.

Arguments

• output: a byte vector, IO object, or BufferedInputStream to which de-

compressed data should be written.
49.5. LIBZ.ADLER32 413

• bufsize::Integer=8192: input and output buffer size.

• gzip::Bool=true: if true, data is gzip compressed; if false, zlib com-
pressed.

source

49.5 Libz.adler32
Libz.adler32 — Function.

adler32(data)

Compute the Adler-32 checksum over the data input. data can be BufferedInputStream
or Vector{UInt8}.
source

49.6 Libz.crc32
Libz.crc32 — Function.

crc32(data)

Compute the CRC-32 checksum over the data input. data can be BufferedInputStream
or Vector{UInt8}.
source
Chapter 50

ImageMetadata

50.1 ImageAxes.data
ImageAxes.data — Method.
data(img::ImageMeta) -> array
Extract the data from img, omitting the properties dictionary. array shares
storage with img, so changes to one affect the other.
See also: properties.
source

50.2 ImageMetadata.copyproperties
ImageMetadata.copyproperties — Method.
copyproperties(img::ImageMeta, data) -> imgnew
Create a new “image,” copying the properties dictionary of img but using the
data of the AbstractArray data. Note that changing the properties of imgnew
does not affect the properties of img.
See also: shareproperties.
source

50.3 ImageMetadata.properties
ImageMetadata.properties — Method.
properties(imgmeta) -> props
Extract the properties dictionary props for imgmeta. props shares storage
with img, so changes to one affect the other.
See also: data.
source

414
50.4. IMAGEMETADATA.SHAREPROPERTIES 415

50.4 ImageMetadata.shareproperties
ImageMetadata.shareproperties — Method.

shareproperties(img::ImageMeta, data) -> imgnew

Create a new “image,” reusing the properties dictionary of img but using the
data of the AbstractArray data. The two images have synchronized properties;
modifying one also affects the other.
See also: copyproperties.
source

50.5 ImageMetadata.spatialproperties
ImageMetadata.spatialproperties — Method.

spatialproperties(img)

Return a vector of strings, containing the names of properties that have been
declared “spatial” and hence should be permuted when calling permutedims.
Declare such properties like this:

img["spatialproperties"] = ["spacedirections"]

source
Chapter 51

LegacyStrings

51.1 LegacyStrings.utf16
LegacyStrings.utf16 — Function.

utf16(::Union{Ptr{UInt16},Ptr{Int16}} [, length])

Create a string from the address of a NUL-terminated UTF-16 string. A

copy is made; the pointer can be safely freed. If length is specified, the string
does not have to be NUL-terminated.
source

51.2 LegacyStrings.utf16
LegacyStrings.utf16 — Method.

utf16(s)

Create a UTF-16 string from a byte array, array of UInt16, or any other
string type. (Data must be valid UTF-16. Conversions of byte arrays check for
a byte-order marker in the first two bytes, and do not include it in the resulting
string.)
Note that the resulting UTF16String data is terminated by the NUL code-
point (16-bit zero), which is not treated as a character in the string (so that
it is mostly invisible in Julia); this allows the string to be passed directly to
external functions requiring NUL-terminated data. This NUL is appended au-
tomatically by the utf16(s) conversion function. If you have a UInt16 array A
that is already NUL-terminated valid UTF-16 data, then you can instead use
UTF16String(A) to construct the string without making a copy of the data and
treating the NUL as a terminator rather than as part of the string.
source

416
51.3. LEGACYSTRINGS.UTF32 417

51.3 LegacyStrings.utf32
LegacyStrings.utf32 — Function.

utf32(::Union{Ptr{Char},Ptr{UInt32},Ptr{Int32}} [, length])

Create a string from the address of a NUL-terminated UTF-32 string. A

copy is made; the pointer can be safely freed. If length is specified, the string
does not have to be NUL-terminated.
source

51.4 LegacyStrings.utf32
LegacyStrings.utf32 — Method.

utf32(s)

Create a UTF-32 string from a byte array, array of Char or UInt32, or any
other string type. (Conversions of byte arrays check for a byte-order marker in
the first four bytes, and do not include it in the resulting string.)
Note that the resulting UTF32String data is terminated by the NUL code-
point (32-bit zero), which is not treated as a character in the string (so that
it is mostly invisible in Julia); this allows the string to be passed directly to
external functions requiring NUL-terminated data. This NUL is appended au-
tomatically by the utf32(s) conversion function. If you have a Char or UInt32
array A that is already NUL-terminated UTF-32 data, then you can instead use
UTF32String(A) to construct the string without making a copy of the data and
treating the NUL as a terminator rather than as part of the string.
source
Chapter 52

BufferedStreams

52.1 BufferedStreams.anchor!
BufferedStreams.anchor! — Method.
Set the buffer’s anchor to its current position.
source

52.2 BufferedStreams.fillbuffer!
BufferedStreams.fillbuffer! — Method.
Refill the buffer, optionally moving and retaining part of the data.
source

52.3 BufferedStreams.isanchored
BufferedStreams.isanchored — Method.
Return true if the stream is anchored.
source

52.4 BufferedStreams.peek
BufferedStreams.peek — Method.
Return the next byte from the input stream without advancing the position.
source

52.5 BufferedStreams.peekbytes!
BufferedStreams.peekbytes! — Function.
Fills buffer with bytes from stream’s buffer without advancing the position.

418
52.6. BUFFEREDSTREAMS.TAKEANCHORED! 419

Unless the buffer is empty, we do not re-fill it. Therefore the number of bytes
read is limited to the minimum of nb and the remaining bytes in the buffer.
source

52.6 BufferedStreams.takeanchored!
BufferedStreams.takeanchored! — Method.
Copy and return a byte array from the anchor up to, but not including the
current position, also removing the anchor.
source

52.7 BufferedStreams.upanchor!
BufferedStreams.upanchor! — Method.
Remove and return a buffer’s anchor.
source
Chapter 53

MbedTLS

53.1 MbedTLS.decrypt
MbedTLS.decrypt — Function.
decrypt(cipher, key, msg, [iv]) -> Vector{UInt8}
Decrypt a message using the given cipher. The cipher can be specified as

• a generic cipher (like CIPHER AES)

• a specific cipher (like CIPHER AES 256 CBC)

• a Cipher object

key is the symmetric key used for cryptography, given as either a String or a
Vector{UInt8}. It must be the right length for the chosen cipher; for example,
CIPHER AES 256 CBC requires a 32-byte (256-bit) key.
msg is the message to be encoded. It should either be convertible to a String
or be a Vector{UInt8}.
iv is the initialization vector, whose size must match the block size of the
cipher (eg, 16 bytes for AES) and correspond to the iv used by the encryptor.
By default, it will be set to all zeros.
source

53.2 MbedTLS.digest
MbedTLS.digest — Function.
digest(kind::MDKind, msg::Vector{UInt8}, [key::Vector{UInt8}]) ->
Vector{UInt8}
Perform a digest of the given type on the given message (a byte array),
return a byte array with the digest.
If an optional key is given, perform an HMAC digest.
source

420
53.3. MBEDTLS.DIGEST! 421

53.3 MbedTLS.digest!
MbedTLS.digest! — Function.
digest!(kind::MDKind, msg::Vector{UInt8}, [key::Vector{UInt8}, ],
buffer::Vector{UInt8})
In-place version of digest that stores the digest to buffer.
It is the user’s responsibility to ensure that buffer is long enough to contain
the digest. get size(kind::MDKind) returns the appropriate size.
source

53.4 MbedTLS.encrypt
MbedTLS.encrypt — Function.
encrypt(cipher, key, msg, [iv]) -> Vector{UInt8}
Encrypt a message using the given cipher. The cipher can be specified as

• a generic cipher (like CIPHER AES)

• a specific cipher (like CIPHER AES 256 CBC)

• a Cipher object

key is the symmetric key used for cryptography, given as either a String or a
Vector{UInt8}. It must be the right length for the chosen cipher; for example,
CIPHER AES 256 CBC requires a 32-byte (256-bit) key.
msg is the message to be encoded. It should either be convertible to a String
or be a Vector{UInt8}.
iv is the initialization vector, whose size must match the block size of the
cipher (eg, 16 bytes for AES). By default, it will be set to all zeros, which is
not secure. For security reasons, it should be set to a different value for each
encryption operation.
source
Chapter 54

DataValues

54.1 DataValues.dropna!
DataValues.dropna! — Method.

dropna!(X::DataValueVector)

Remove missing entries of X in-place and return a Vector view of the un-
wrapped DataValue entries.
source

54.2 DataValues.dropna!
DataValues.dropna! — Method.

dropna!(X::AbstractVector)

Remove missing entries of X in-place and return a Vector view of the un-
wrapped DataValue entries. If no missing values are present, this is a no-op
and X is returned.
source

54.3 DataValues.dropna
DataValues.dropna — Method.

dropna(X::AbstractVector)

Return a vector containing only the non-missing entries of X, unwrapping

DataValue entries. A copy is always returned, even when X does not contain
any missing values.
source

422
54.4. DATAVALUES.PADNA! 423

54.4 DataValues.padna!
DataValues.padna! — Method.

padna!(X::DataValueVector, front::Integer, back::Integer)

Insert front null entries at the beginning of X and add back null entries at
the end of X. Returns X.
source

54.5 DataValues.padna
DataValues.padna — Method.

padna(X::DataValueVector, front::Integer, back::Integer)

return a copy of X with front null entries inserted at the beginning of the
copy and back null entries inserted at the end.
source
Chapter 55

NearestNeighbors

55.1 NearestNeighbors.injectdata
NearestNeighbors.injectdata — Method.

injectdata(datafreetree, data) -> tree

Returns the KDTree/BallTree wrapped by datafreetree, set up to use

data for the points data.
source

55.2 NearestNeighbors.inrange
NearestNeighbors.inrange — Method.

inrange(tree::NNTree, points, radius [, sortres=false]) -> indices

Find all the points in the tree which is closer than radius to points. If
sortres = true the resulting indices are sorted.
source

55.3 NearestNeighbors.knn
NearestNeighbors.knn — Method.

knn(tree::NNTree, points, k [, sortres=false]) -> indices, distances

Performs a lookup of the k nearest neigbours to the points from the data
in the tree. If sortres = true the result is sorted such that the results are in
the order of increasing distance to the point. skip is an optional predicate to
determine if a point that would be returned should be skipped.
source

424
Chapter 56

IJulia

56.1 IJulia.installkernel
IJulia.installkernel — Method.

installkernel(name, options...; specname=replace(lowercase(name), " ", "-")

Install a new Julia kernel, where the given options are passed to the julia
executable, and the user-visible kernel name is given by name followed by the
Julia version.
Internally, the Jupyter name for the kernel (for the jupyter kernelspec
command is given by the optional keyword specname (which defaults to name,
converted to lowercase with spaces replaced by hyphens), followed by the Julia
version number.
Both the kernelspec command (a Cmd object) and the new kernel name are
returned by installkernel. For example:

kernelspec, kernelname = installkernel("Julia O3", "-O3")

creates a new Julia kernel in which julia is launched with the -O3 optimiza-
tion flag. The returned kernelspec command will be something like jupyter
kernelspec (perhaps with different path), and kernelname will be something
like julia-O3-0.6 (in Julia 0.6). You could uninstall the kernel by running e.g.

run(‘$kernelspec remove -f $kernelname‘)

source

56.2 IJulia.notebook
IJulia.notebook — Method.

notebook(; dir=homedir(), detached=false)

425
426 CHAPTER 56. IJULIA

The notebook() function launches the Jupyter notebook, and is equivalent

to running jupyter notebook at the operating-system command-line. The ad-
vantage of launching the notebook from Julia is that, depending on how Jupyter
was installed, the user may not know where to find the jupyter executable.
By default, the notebook server is launched in the user’s home directory, but
this location can be changed by passing the desired path in the dir keyword
argument. e.g. notebook(dir=pwd()) to use the current directory.
By default, notebook() does not return; you must hit ctrl-c or quit Ju-
lia to interrupt it, which halts Jupyter. So, you must leave the Julia termi-
nal open for as long as you want to run Jupyter. Alternatively, if you run
notebook(detached=true), the jupyter notebook will launch in the back-
ground, and will continue running even after you quit Julia. (The only way to
stop Jupyter will then be to kill it in your operating system’s process manager.)
source
Chapter 57

HDF5

57.1 HDF5.get datasets

HDF5.get datasets — Method.

get_datasets(file::HDF5File) -> datasets::Vector{HDF5Dataset}

Get all the datasets in an hdf5 file without loading the data.
source

57.2 HDF5.h5open
HDF5.h5open — Function.

h5open(filename::AbstractString, mode::AbstractString="r"; swmr=false)

Open or create an HDF5 file where mode is one of:

• “r” read only

• “r+” read and write

• “cw” read and write, create file if not existing, do not truncate

• “w” read and write, create a new file (destroys any existing contents)

Pass swmr=true to enable (Single Writer Multiple Reader) SWMR write

access for “w” and “r+”, or SWMR read access for “r”.
source

427
428 CHAPTER 57. HDF5

57.3 HDF5.h5open
HDF5.h5open — Method.

function h5open(f::Function, args...; swmr=false)

Apply the function f to the result of h5open(args...;kwargs...) and close

the resulting HDF5File upon completion. For example with a do block:

h5open("foo.h5","w") do h5
h5["foo"]=[1,2,3]
end

source

57.4 HDF5.ishdf5
HDF5.ishdf5 — Method.

ishdf5(name::AbstractString)

Returns true if name is a path to a valid hdf5 file, false otherwise.

source

57.5 HDF5.set dims!

HDF5.set dims! — Method.

set_dims!(dset::HDF5Dataset, new_dims::Dims)

Change the current dimensions of a dataset to new dims, limited by max dims
= get dims(dset)[2]. Reduction is possible and leads to loss of truncated
data.
source
Chapter 58

CUDAapi

58.1 CUDAapi.find binary

CUDAapi.find binary — Method.

find_binary(names; locations=String[])

Similar to find library, performs an exhaustive search for a binary in

various subdirectories of locations, and finally PATH.
source

58.2 CUDAapi.find libdevice

CUDAapi.find libdevice — Method.

find_libdevice(targets::Vector{VersionNumber}, toolkit_dirs::Vector{String})

Look for the CUDA device library supporting targets in any of the CUDA
toolkit directories toolkit dirs. On CUDA >= 9.0, a single library unified
library is discovered and returned as a string. On older toolkits, individual
libraries for each of the targets are returned as a vector of strings.
source

58.3 CUDAapi.find library

CUDAapi.find library — Method.

find_library(names; locations=String[], versions=VersionNumber[], word_size=Sys.WORD_SIZE)

Wrapper for Libdl.find library, performing a more exhaustive search:

• variants of the library name (including version numbers, platform-specific

tags, etc);

429
430 CHAPTER 58. CUDAAPI

• various subdirectories of the locations list, and finally system library

directories.

Returns the full path to the library.

source

58.4 CUDAapi.find toolkit

CUDAapi.find toolkit — Method.

find_toolkit()::Vector{String}

Look for directories where (parts of) the CUDA toolkit might be installed.
This returns a (possibly empty) list of paths that can be used as an argument
to other discovery functions.
The behavior of this function can be overridden by defining the CUDA PATH,
CUDA HOME or CUDA ROOT environment variables, which should point to the root
of the CUDA toolkit.
source
Chapter 59

AutoGrad

59.1 AutoGrad.getval
AutoGrad.getval — Method.

getval(x)

Unbox x if it is a boxed value (Rec), otherwise return x.

source

59.2 AutoGrad.grad
AutoGrad.grad — Function.

grad(fun, argnum=1)

Take a function fun(X...)->Y and return another function gfun(X...)->dXi

which computes its gradient with respect to positional argument number argnum.
The function fun should be scalar-valued. The returned function gfun takes the
same arguments as fun, but returns the gradient instead. The gradient has the
same type and size as the target argument which can be a Number, Array,
Tuple, or Dict.
source

59.3 AutoGrad.gradcheck
AutoGrad.gradcheck — Method.

gradcheck(f, w, x...; kwargs...)

431
432 CHAPTER 59. AUTOGRAD

Numerically check the gradient of f(w,x...;o...) with respect to its first

argument w and return a boolean result.
The argument w can be a Number, Array, Tuple or Dict which in turn can
contain other Arrays etc. Only the largest 10 entries in each numerical gradient
array are checked by default. If the output of f is not a number, gradcheck
constructs and checks a scalar function by taking its dot product with a random
vector.
Keywords

• gcheck=10: number of largest entries from each numeric array in gradient

dw=(grad(f))(w,x...;o...) compared to their numerical estimates.
• verbose=false: print detailed messages if true.
• kwargs=[]: keyword arguments to be passed to f.

• delta=atol=rtol=cbrt(eps(w)): tolerance parameters. See isapprox

for their meaning.

source

59.4 AutoGrad.gradloss
AutoGrad.gradloss — Function.

gradloss(fun, argnum=1)

Another version of grad where the generated function returns a (gradi-

ent,value) pair.
source
Chapter 60

ComputationalResources

60.1 ComputationalResources.addresource
ComputationalResources.addresource — Method.

addresource(T)

Add T to the list of available resources. For example, addresource(OpenCLLibs)

would indicate that you have a GPU and the OpenCL libraries installed.
source

60.2 ComputationalResources.haveresource
ComputationalResources.haveresource — Method.

haveresource(T)

Returns true if T is an available resource. For example, haveresource(OpenCLLibs)

tests whether the OpenCLLibs have been added as an available resource. This
function is typically used inside a module’s init function.
Example:
[] The i nitf unctionf orM yP ackage: f unctioni nit ) ...otherinitializationcode,possiblysettingtheLOADP AT Hif haveresource(OpenCLLibs)@evalusin
(

source

60.3 ComputationalResources.rmresource
ComputationalResources.rmresource — Method.

rmresource(T)

433
434 CHAPTER 60. COMPUTATIONALRESOURCES

Remove T from the list of available resources. For example, rmresource(OpenCLLibs)

would indicate that any future package loads should avoid loading their special-
izations for OpenCL.
source
Chapter 61

MicroLogging

61.1 MicroLogging.configure logging

MicroLogging.configure logging — Method.

configure_logging(args...; kwargs...)

Call configure logging with the current logger, and update cached log
filtering information.
source

61.2 MicroLogging.current logger

MicroLogging.current logger — Method.

current_logger()

Return the logger for the current task, or the global logger if none is is
attached to the task.
source

61.3 MicroLogging.disable logging

MicroLogging.disable logging — Method.

disable_logging(level)

Disable all log messages at log levels equal to or less than level. This is a
global setting, intended to make debug logging extremely cheap when disabled.
source

435
436 CHAPTER 61. MICROLOGGING

61.4 MicroLogging.global logger

MicroLogging.global logger — Method.

global_logger()

Return the global logger, used to receive messages when no specific logger
exists for the current task.

global_logger(logger)

Set the global logger to logger.

source

61.5 MicroLogging.with logger

MicroLogging.with logger — Method.

with_logger(function, logger)

Execute function, directing all log messages to logger.

Example
[] function test(x) @info ”x = x”end
withl ogger(logger)dotest(1)test([1,2])end
source
Chapter 62

Clustering

62.1 Clustering.dbscan
Clustering.dbscan — Method.

dbscan(points, radius ; leafsize = 20, min_neighbors = 1, min_cluster_size = 1) -> clusters

Cluster points using the DBSCAN (density-based spatial clustering of ap-

plications with noise) algorithm.
Arguments

• points: matrix of points

• radius::Real: query radius

Keyword Arguments

• leafsize::Int: number of points binned in each leaf node in the KDTree

• min neighbors::Int: minimum number of neighbors to be a core point

• min cluster size::Int: minimum number of points to be a valid cluster

Output

• Vector{DbscanCluster}: an array of clusters with the id, size core indices

and boundary indices

Example:
[] points = randn(3, 10000) clusters = dbscan(points, 0.05, minn eighbors =
3, minc lusters ize =20)clusterswithlessthan20pointswillbediscarded
source

437
438 CHAPTER 62. CLUSTERING

62.2 Clustering.mcl
Clustering.mcl — Method.

mcl(adj::Matrix; [keyword arguments])::MCLResult

Identify clusters in the weighted graph using Markov Clustering Algorithm

(MCL).
Arguments

• adj::Matrix{Float64}: adjacency matrix that defines the weighted graph

to cluster

• add loops::Bool: whether edges of weight 1.0 from the node to itself
should be appended to the graph (enabled by default)
• expansion::Number: MCL expansion constant (2)
• inflation::Number: MCL inflation constant (2.0)

• save final matrix::Bool: save final equilibrium state in the result, oth-
erwise leave it empty; disabled by default, could be useful if MCL doesn’t
converge
• max iter::Integer: max number of MCL iterations

• tol::Number: MCL adjacency matrix convergence threshold

• prune tol::Number: pruning threshold
• display::Symbol: :none for no output or :verbose for diagnostic mes-
sages

See original MCL implementation.

Ref: Stijn van Dongen, “Graph clustering by flow simulation”, 2001
source
Chapter 63

JuliaWebAPI

63.1 JuliaWebAPI.apicall
JuliaWebAPI.apicall — Method.
Calls a remote api cmd with args... and data.... The response is format-
ted as specified by the formatter specified in conn.
source

63.2 JuliaWebAPI.fnresponse
JuliaWebAPI.fnresponse — Method.
extract and return the response data as a direct function call would have
returned but throw error if the call was not successful.
source

63.3 JuliaWebAPI.httpresponse
JuliaWebAPI.httpresponse — Method.
construct an HTTP Response object from the API response
source

63.4 JuliaWebAPI.process
JuliaWebAPI.process — Method.
start processing as a server
source

439
440 CHAPTER 63. JULIAWEBAPI

63.5 JuliaWebAPI.register
JuliaWebAPI.register — Method.
Register a function as API call. TODO: validate method belongs to module?
source
Chapter 64

DecisionTree

64.1 DecisionTree.apply adaboost stumps proba

DecisionTree.apply adaboost stumps proba — Method.

apply_adaboost_stumps_proba(stumps::Ensemble, coeffs, features, labels::Vector)

computes P(L=label|X) for each row in features. It returns a N row x

n labels matrix of probabilities, each row summing up to 1.
col labels is a vector containing the distinct labels (eg. [“versicolor”, “vir-
ginica”, “setosa”]). It specifies the column ordering of the output matrix.
source

64.2 DecisionTree.apply forest proba

DecisionTree.apply forest proba — Method.

apply_forest_proba(forest::Ensemble, features, col_labels::Vector)

computes P(L=label|X) for each row in features. It returns a N row x

n labels matrix of probabilities, each row summing up to 1.
col labels is a vector containing the distinct labels (eg. [“versicolor”, “vir-
ginica”, “setosa”]). It specifies the column ordering of the output matrix.
source

64.3 DecisionTree.apply tree proba

DecisionTree.apply tree proba — Method.

apply_tree_proba(::Node, features, col_labels::Vector)

441
442 CHAPTER 64. DECISIONTREE

computes P(L=label|X) for each row in features. It returns a N row x

n labels matrix of probabilities, each row summing up to 1.
col labels is a vector containing the distinct labels (eg. [“versicolor”, “vir-
ginica”, “setosa”]). It specifies the column ordering of the output matrix.
source
Chapter 65

Parameters

65.1 Parameters.reconstruct
Parameters.reconstruct — Method.
Make a new instance of a type with the same values as the input type except
for the fields given in the AbstractDict second argument or as keywords. Works
for types, Dicts, and NamedTuples.
Note: this is not very performant. Check Setfield.jl for a faster & nicer
implementation.

julia> struct A
a
b
end

julia> a = A(3,4)
A(3, 4)

julia> b = reconstruct(a, b=99)

A(3, 99)

source

65.2 Parameters.type2dict
Parameters.type2dict — Method.
Transforms a type-instance into a dictionary.

julia> type T
a
b
end

443
444 CHAPTER 65. PARAMETERS

julia> type2dict(T(4,5))
Dict{Symbol,Any} with 2 entries:
:a => 4
:b => 5

source

65.3 Parameters.with kw
Parameters.with kw — Function.
This function is called by the @with kw macro and does the syntax transfor-
mation from:
[] @withk wstructM M {R}r :: R =1000.a :: Rend
into
[] struct MM{R} r::R a::R MM{R}(r,a) where {R} = new(r,a) MM{R}(;r=1000.,
a=error(”no default for a”)) where {R} = MM{R}(r,a) inner kw, type-paras
are required when calling end MM(r::R,a::R) where {R} = MM{R}(r,a) de-
fault outer positional constructor MM(;r=1000,a=error(”no default for a”)) =
MM(r,a) outer kw, so no type-paras are needed when calling MM(m::MM;
kws...) = reconstruct(mm,kws) MM(m::MM, di::Union{AbstractDict, Tuple{Symbol,Any}})
= reconstruct(mm, di) macro unpackM M (varname)esc(quoter = varname.ra = varname.aend)endmac

source
Chapter 66

Blosc

66.1 Blosc.compress
Blosc.compress — Function.

compress(data; level=5, shuffle=true, itemsize)

Return a Vector{UInt8} of the Blosc-compressed data, where data is an

array or a string.
The level keyword indicates the compression level (between 0=no compres-
sion and 9=max), shuffle indicates whether to use Blosc’s shuffling precondi-
tioner, and the shuffling preconditioner is optimized for arrays of binary items
of size (in bytes) itemsize (defaults to sizeof(eltype(data)) for arrays and
the size of the code units for strings).
source

66.2 Blosc.compress!
Blosc.compress! — Function.

compress!(dest::Vector{UInt8}, src; kws...)

Like compress(src; kws...), but writes to a pre-allocated array dest of

bytes. The return value is the size in bytes of the data written to dest, or 0 if
the buffer was too small.
source

66.3 Blosc.decompress!
Blosc.decompress! — Method.

decompress!(dest::Vector{T}, src::Vector{UInt8})

445
446 CHAPTER 66. BLOSC

Like decompress, but uses a pre-allocated destination buffer dest, which is

resized as needed to store the decompressed data from src.
source

66.4 Blosc.decompress
Blosc.decompress — Method.

decompress(T::Type, src::Vector{UInt8})

Return the compressed buffer src as an array of element type T.

source
Chapter 67

Missings

67.1 Missings.allowmissing
Missings.allowmissing — Method.

allowmissing(x::AbstractArray)

Return an array equal to x allowing for missing values, i.e. with an element
type equal to Union{eltype(x), Missing}.
When possible, the result will share memory with x (as with convert).
See also: disallowmissing
source

67.2 Missings.disallowmissing
Missings.disallowmissing — Method.

disallowmissing(x::AbstractArray)

Return an array equal to x not allowing for missing values, i.e. with an
element type equal to Missings.T(eltype(x)).
When possible, the result will share memory with x (as with convert). If x
contains missing values, a MethodError is thrown.
See also: allowmissing
source

67.3 Missings.levels
Missings.levels — Method.

levels(x)

447
448 CHAPTER 67. MISSINGS

Return a vector of unique values which occur or could occur in collection x,

omitting missing even if present. Values are returned in the preferred order for
the collection, with the result of sort as a default.
Contrary to unique, this function may return values which do not actually
occur in the data, and does not preserve their order of appearance in x.
source
Chapter 68

ProgressMeter

68.1 ProgressMeter.cancel
ProgressMeter.cancel — Function.
cancel(prog, [msg], [color=:red]) cancels the progress display before
all tasks were completed. Optionally you can specify the message printed and
its color.
See also finish!.
source

68.2 ProgressMeter.finish!
ProgressMeter.finish! — Method.
finish!(prog) indicates that all tasks have been completed.
See also cancel.
source

68.3 ProgressMeter.next!
ProgressMeter.next! — Method.
next!(prog, [color]) reports that one unit of progress has been made.
Depending on the time interval since the last update, this may or may not
result in a change to the display.
You may optionally change the color of the display. See also update!.
source

68.4 ProgressMeter.update!
ProgressMeter.update! — Method.

449
450 CHAPTER 68. PROGRESSMETER

update!(prog, counter, [color]) sets the progress counter to counter,

relative to the n units of progress specified when prog was initialized. Depending
on the time interval since the last update, this may or may not result in a change
to the display.
If prog is a ProgressThresh, update!(prog, val, [color]) specifies the
current value.
You may optionally change the color of the display. See also next!.
source
Chapter 69

HttpServer

69.1 Base.run
Base.run — Method.
run starts server
Functionality:

• Accepts incoming connections and instantiates each Client.

• Manages the client.id pool.
• Spawns a new Task for each connection.
• Blocks forever.

Method accepts following keyword arguments:

• host - binding address

• port - binding port
• ssl - SSL configuration. Use this argument to enable HTTPS support.

Can be either an MbedTLS.SSLConfig object that already has associated

certificates, or a tuple of an MbedTLS.CRT (certificate) and MbedTLS.PKContext
(private key)). In the latter case, a configuration with reasonable defaults will
be used.

• socket - named pipe/domain socket path. Use this argument to enable

Unix socket support.

It’s available only on Unix. Network options are ignored.

Compatibility:

• for backward compatibility use run(server::Server, port::Integer)

451
452 CHAPTER 69. HTTPSERVER

Example:

server = Server() do req, res

"Hello world"
end

# start server on localhost

run(server, host=IPv4(127,0,0,1), port=8000)
# or
run(server, 8000)

source

69.2 Base.write
Base.write — Method.
Converts a Response to an HTTP response string
source

69.3 HttpServer.setcookie!
HttpServer.setcookie! — Function.
Sets a cookie with the given name, value, and attributes on the given re-
sponse object.
source
Chapter 70

MappedArrays

70.1 MappedArrays.mappedarray
MappedArrays.mappedarray — Method.

mappedarray(f, A)

creates a view of the array A that applies f to every element of A. The view
is read-only (you can get values but not set them).
source

70.2 MappedArrays.mappedarray
MappedArrays.mappedarray — Method.

mappedarray((f, finv), A)

creates a view of the array A that applies f to every element of A. The inverse
function, finv, allows one to also set values of the view and, correspondingly,
the values in A.
source

70.3 MappedArrays.of eltype

MappedArrays.of eltype — Method.

of_eltype(T, A)
of_eltype(val::T, A)

creates a view of A that lazily-converts the element type to T.

source

453
Chapter 71

TextParse

71.1 TextParse.csvread
TextParse.csvread — Function.

csvread(file::Union{String,IO}, delim=’,’; <arguments>...)

Read CSV from file. Returns a tuple of 2 elements:

1. A tuple of columns each either a Vector, DataValueArray or PooledArray

2. column names if header exists=true, empty array otherwise

Arguments:

• file: either an IO object or file name string

• delim: the delimiter character
• spacedelim: (Bool) parse space-delimited files. delim has no effect if
true.
• quotechar: character used to quote strings, defaults to "
• escapechar: character used to escape quotechar in strings. (could be the
same as quotechar)
• pooledstrings: whether to try and create PooledArray of strings
• nrows: number of rows in the file. Defaults to 0 in which case we try to
estimate this.
• skiplines begin: skips specified number of lines at the beginning of the
file
• header exists: boolean specifying whether CSV file contains a header

454
71.1. TEXTPARSE.CSVREAD 455

• nastrings: strings that are to be considered NA. Defaults to TextParse.NA STRINGS

• colnames: manually specified column names. Could be a vector or a
dictionary from Int index (the column) to String column name.
• colparsers: Parsers to use for specified columns. This can be a vector
or a dictionary from column name / column index (Int) to a “parser”.
The simplest parser is a type such as Int, Float64. It can also be a
dateformat"...", see CustomParser if you want to plug in custom pars-
ing behavior
• type detect rows: number of rows to use to infer the initial colparsers
defaults to 20.

source
Chapter 72

LossFunctions

72.1 LearnBase.scaled
LearnBase.scaled — Method.

scaled(loss::SupervisedLoss, K)

Returns a version of loss that is uniformly scaled by K. This function dis-

patches on the type of loss in order to choose the appropriate type of scaled
loss that will be used as the decorator. For example, if typeof(loss) <:
DistanceLoss then the given loss will be boxed into a ScaledDistanceLoss.
Note: If typeof(K) <: Number, then this method will poison the type-
inference of the calling scope. This is because K will be promoted to a type
parameter. For a typestable version use the following signature: scaled(loss,
Val{K})
source

72.2 LossFunctions.weightedloss
LossFunctions.weightedloss — Method.

weightedloss(loss, weight)

Returns a weighted version of loss for which the value of the positive class
is changed to be weight times its original, and the negative class 1 - weight
times its original respectively.
Note: If typeof(weight) <: Number, then this method will poison the
type-inference of the calling scope. This is because weight will be promoted
to a type parameter. For a typestable version use the following signature:
weightedloss(loss, Val{weight})
source

456
Chapter 73

LearnBase

73.1 LearnBase.grad
LearnBase.grad — Function.
Return the gradient of the learnable parameters w.r.t. some objective
source

73.2 LearnBase.grad!
LearnBase.grad! — Function.
Do a backward pass, updating the gradients of learnable parameters and/or
inputs
source

73.3 LearnBase.inputdomain
LearnBase.inputdomain — Function.
Returns an AbstractSet representing valid input values
source

73.4 LearnBase.targetdomain
LearnBase.targetdomain — Function.
Returns an AbstractSet representing valid output/target values
source

73.5 LearnBase.transform!
LearnBase.transform! — Function.

457
458 CHAPTER 73. LEARNBASE

Do a forward pass, and return the output

source
Chapter 74

Juno

74.1 Juno.clearconsole
Juno.clearconsole — Method.

clearconsole()

Clear the console if Juno is used; does nothing otherwise.

source

74.2 Juno.input
Juno.input — Function.

input(prompt = "") -> "..."

Prompt the user to input some text, and return it. Optionally display a
prompt.
source

74.3 Juno.selector
Juno.selector — Method.

selector([xs...]) -> x

Allow the user to select one of the xs.

xs should be an iterator of strings. Currently there is no fallback in other
environments.
source

459
460 CHAPTER 74. JUNO

74.4 Juno.structure
Juno.structure — Method.

structure(x)

Display x’s underlying representation, rather than using its normal display
method.
For example, structure(:(2x+1)) displays the Expr object with its head
and args fields instead of printing the expression.
source
Chapter 75

StatsModels

75.1 StatsBase.coefnames
StatsBase.coefnames — Method.
coefnames(mf::ModelFrame)
Returns a vector of coefficient names constructed from the Terms member
and the types of the evaluation columns.
source

75.2 StatsModels.dropterm
StatsModels.dropterm — Method.
dropterm(f::Formula, trm::Symbol)
Return a copy of f without the term trm.
Examples
[] julia¿ dropterm(@formula(foo 1 + bar + baz), :bar) Formula: foo 1+
baz
julia¿ dropterm(@formula(foo 1 + bar + baz), 1) Formula: foo 0 + bar
+ baz
source

75.3 StatsModels.setcontrasts!
StatsModels.setcontrasts! — Method.
setcontrasts!(mf::ModelFrame, new_contrasts::Dict)
Modify the contrast coding system of a ModelFrame in place.
source

461
Chapter 76

InternedStrings

76.1 InternedStrings.intern
InternedStrings.intern — Method.
Substrings are interned as their parent string type
source

76.2 InternedStrings.intern
InternedStrings.intern — Method.

intern(::Type{T}, s)

Intern s as if it were type T, converting it if required. Note that this will lead
to unexpected behavour if the type of s, and T, do not have equivalent equality
and hash functions (i.e. this is not safe if hash(s) != hash(convert(T, s))).
source

76.3 InternedStrings.intern
InternedStrings.intern — Method.

intern(s::T)

Return a reference to a interned instance of s, adding it to the interning

pool if it did not already exist.
source

462
Chapter 77

HttpParser

77.1 HttpParser.http method str

HttpParser.http method str — Method.
Returns a string version of the HTTP method.
source

77.2 HttpParser.http parser execute

HttpParser.http parser execute — Method.
Run a request through a parser with specific callbacks on the settings in-
stance.
source

77.3 HttpParser.http parser init

HttpParser.http parser init — Function.
Intializes the Parser object with the correct memory.
source

77.4 HttpParser.parse url

HttpParser.parse url — Method.
Parse a URL
source

463
Chapter 78

ImageAxes

78.1 ImageAxes.istimeaxis
ImageAxes.istimeaxis — Method.

istimeaxis(ax)

Test whether the axis ax corresponds to time.

source

78.2 ImageAxes.timeaxis
ImageAxes.timeaxis — Method.

timeaxis(A)

Return the time axis, if present, of the array A, and nothing otherwise.
source

78.3 ImageAxes.timedim
ImageAxes.timedim — Method.

timedim(img) -> d::Int

Return the dimension of the array used for encoding time, or 0 if not using
an axis for this purpose.
Note: if you want to recover information about the time axis, it is generally
better to use timeaxis.
source

464
Chapter 79

Flux

79.1 Flux.GRU
Flux.GRU — Method.

GRU(in::Integer, out::Integer, = tanh)

Gated Recurrent Unit layer. Behaves like an RNN but generally exhibits a
longer memory span over sequences.
See this article for a good overview of the internals.
source

79.2 Flux.LSTM
Flux.LSTM — Method.

LSTM(in::Integer, out::Integer, = tanh)

Long Short Term Memory recurrent layer. Behaves like an RNN but gener-
ally exhibits a longer memory span over sequences.
See this article for a good overview of the internals.
source

79.3 Flux.RNN
Flux.RNN — Method.

RNN(in::Integer, out::Integer, = tanh)

The most basic recurrent layer; essentially acts as a Dense layer, but with
the output fed back into the input each time step.
source

465
Chapter 80

OnlineStats

80.1 OnlineStats.eachcol
OnlineStats.eachcol — Method.

eachcol(x::AbstractMatrix)

Create an iterator over the columns of a matrix as Vectors.

source

80.2 OnlineStats.eachrow
OnlineStats.eachrow — Method.

eachrow(x::AbstractMatrix)

Create an iterator over the rows of a matrix as Vectors.

eachrow(x::AbstractMatrix, y::Vector)

Create an iterator over the rows of a matrix/vector as Tuple{Vector, eltype(y)}s.

source

80.3 StatsBase.confint
StatsBase.confint — Function.

confint(b::Bootstrap, coverageprob = .95)

Return a confidence interval for a Bootstrap b.

source

466
Chapter 81

IntervalSets

81.1 IntervalSets.:..
IntervalSets.:.. — Method.

iv = l..r

Construct a ClosedInterval iv spanning the region from l to r.

source

81.2 IntervalSets.:
IntervalSets.: — Method.

iv = centerhalfwidth

Construct a ClosedInterval iv spanning the region from center - halfwidth

to center + halfwidth.
source

81.3 IntervalSets.width
IntervalSets.width — Method.

w = width(iv)

Calculate the width (max-min) of interval iv. Note that for integers l and
r, width(l..r) = length(l:r) - 1.
source

467
Chapter 82

Media

82.1 Media.getdisplay
Media.getdisplay — Method.

getdisplay(T)

Find out what output device T will display on.

source

82.2 Media.media
Media.media — Method.
media(T) gives the media type of the type T. The default is Textual.

media(Gadfly.Plot) == Media.Plot

source

82.3 Media.setdisplay
Media.setdisplay — Method.

setdisplay([input], T, output)

Display T objects using output when produced by input.

T is an object type or media type, e.g. Gadfly.Plot or Media.Graphical.

display(Editor(), Image, Console())

source

468
Chapter 83

Rotations

83.1 Rotations.isrotation
Rotations.isrotation — Method.

isrotation(r)
isrotation(r, tol)

Check whether r is a 33 rotation matrix, where r * r is within tol of the

identity matrix (using the Frobenius norm). (tol defaults to 1000 * eps(eltype(r)).)
source

83.2 Rotations.rotation between

Rotations.rotation between — Method.

rotation_between(from, to)

Compute the quaternion that rotates vector from so that it aligns with vector
to, along the geodesic (shortest path).
source

469
Chapter 84

Mustache

84.1 Mustache.render
Mustache.render — Method.
Render a set of tokens with a view, using optional io object to print or store.
Arguments

• io::IO: Optional IO object.

• tokens: Either Mustache tokens, or a string to parse into tokens

• view: A view provides a context to look up unresolved symbols demar-

cated by mustache braces. A view may be specified by a dictionary, a
module, a composite type, a vector, or keyword arguments.

source

84.2 Mustache.render from file

Mustache.render from file — Method.
Renders a template from filepath and view. If it has seen the file before
then it finds the compiled MustacheTokens in TEMPLATES rather than calling
parse a second time.
source

470
Chapter 85

TiledIteration

85.1 TiledIteration.padded tilesize

TiledIteration.padded tilesize — Method.

padded_tilesize(T::Type, kernelsize::Dims, [ncache=2]) -> tilesize::Dims

Calculate a suitable tile size to approximately maximize the amount of pro-

ductive work, given a stencil of size kernelsize. The element type of the array
is T. Optionally specify ncache, the number of such arrays that you’d like to
have fit simultaneously in L1 cache.
This favors making the first dimension larger, since the first dimension cor-
responds to individual cache lines.
Examples
julia> padded tilesize(UInt8, (3,3)) (768,18)
julia> padded tilesize(UInt8, (3,3), 4) (512,12)
julia> padded tilesize(Float64, (3,3)) (96,18)
julia> padded tilesize(Float32, (3,3,3)) (64,6,6)
source

471
Chapter 86

Distances

86.1 Distances.renyi divergence

Distances.renyi divergence — Function.

RenyiDivergence(::Real)
renyi_divergence(P, Q, ::Real)

Create a Rnyi premetric of order .

Rnyi defined a spectrum of divergence measures generalising the Kullback–
Leibler divergence (see KLDivergence). The divergence is not a semimetric
as it is not symmetric. It is parameterised by a parameter , and is equal to
Kullback–Leibler divergence at = 1:
At = 0, R0 (P |Q) = −log(sumi:pi >0 (qi ))
At = 1, R1 (P |Q) = sumi:pi >0 (pi log(pi /qi ))
At = , R(P |Q) = log(supi:pi >0 (pi /qi ))
(
Otherwise R(P |Q) = log(sumi:pi >0 ((pi )/(qi − 1)))/( − 1)
Example:

julia> x = reshape([0.1, 0.3, 0.4, 0.2], 2, 2);

julia> pairwise(RenyiDivergence(0), x, x)
22 Array{Float64,2}:
0.0 0.0
0.0 0.0

julia> pairwise(Euclidean(2), x, x)
22 Array{Float64,2}:
0.0 0.577315
0.655407 0.0

source

472
Chapter 87

VersionParsing

87.1 VersionParsing.vparse
VersionParsing.vparse — Method.

vparse(s::AbstractString)

Returns a VersionNumber representing as much as possible of the version

information in the string s. A much wider range of formats is supported than
the semver styles recognized by VersionNumber(s).
For example,

• Non-numeric prefixes are stripped along with any invalid version charac-
ters. Commas are treated as decimal points.
• Text following whitespace after the version number is ignored.

• major.minor.patch.x.y.z is supported, with x.y.z prepended to the

semver build identifier, i.e. it is parsed like major.minor.patch+x.y.z.
• Multiple +x+y build identifiers are concatenated as if they were +x.y.
• A leading 0 is prepended if needed, e.g. .x is treated as 0.x.

• When all else fails, everything except the first major.minor.patch digits
found are ignored.

source

473
Chapter 88

HttpCommon

88.1 HttpCommon.escapeHTML
HttpCommon.escapeHTML — Method.
escapeHTML(i::String)
Returns a string with special HTML characters escaped: &, <, >, “, ’
source

88.2 HttpCommon.parsequerystring
HttpCommon.parsequerystring — Method.
parsequerystring(query::String)
Convert a valid querystring to a Dict:

q = "foo=bar&baz=%3Ca%20href%3D%27https%3A%2F%2Fptop.only.wip.la%3A443%2Fhttp%2Fwww.hackershool.com%27%3Ehello%20world%2
parsequerystring(q)
# Dict{String,String} with 2 entries:
# "baz" => "<a href=’http://www.hackershool.com’>hello world!</a>"
# "foo" => "bar"

source

474
Chapter 89

StaticArrays

89.1 StaticArrays.similar type

StaticArrays.similar type — Function.

similar_type(static_array)
similar_type(static_array, T)
similar_type(array, ::Size)
similar_type(array, T, ::Size)

Returns a constructor for a statically-sized array similar to the input array

(or type) static array/array, optionally with different element type T or size
Size. If the input array is not a StaticArray then the Size is mandatory.
This differs from similar() in that the resulting array type may not be
mutable (or define setindex!()), and therefore the returned type may need to
be constructed with its data.
Note that the (optional) size must be specified as a static Size object (so
the compiler can infer the result statically).
New types should define the signature similar type{A<:MyType,T,S}(::Type{A},::Type{T},::Size{S})
if they wish to overload the default behavior.
source

475
Chapter 90

SweepOperator

90.1 SweepOperator.sweep!
SweepOperator.sweep! — Method.
Symmetric sweep operator
Symmetric sweep operator of the matrix A on element k. A is overwritten.
inv = true will perform the inverse sweep. Only the upper triangle is read and
swept.
sweep!(A, k, inv = false)
Providing a Range, rather than an Integer, sweeps on each element in the
range.
sweep!(A, first:last, inv = false)
Example:
[] x = randn(100, 10) xtx = x’x sweep!(xtx, 1) sweep!(xtx, 1, true)
source

476
Chapter 91

PaddedViews

91.1 PaddedViews.paddedviews
PaddedViews.paddedviews — Method.

Aspad = paddedviews(fillvalue, A1, A2, ....)

Pad the arrays A1, A2, . . . , to a common size or set of axes, chosen as the
span of axes enclosing all of the input arrays.
Example:
[] julia¿ a1 = reshape([1,2], 2, 1) 21 Array{Int64,2}: 1 2
julia¿ a2 = [1.0,2.0]’ 12 Array{Float64,2}: 1.0 2.0
julia¿ a1p, a2p = paddedviews(0, a1, a2);
julia¿ a1p 22 PaddedViews.PaddedView{Int64,2,Tuple{Base.OneTo{Int64},Base.OneTo{Int64}},Array{Int64,2}}:
1020
julia¿ a2p 22 PaddedViews.PaddedView{Float64,2,Tuple{Base.OneTo{Int64},Base.OneTo{Int64}},Array{Float64
1.0 2.0 0.0 0.0
source

477
Chapter 92

Pidfile

92.1 Pidfile.mkpidlock
Pidfile.mkpidlock — Function.

mkpidlock(at::String, [pid::Cint, proc::Process]; kwopts...)

Create a pidfile lock for the path “at” for the current process or the process
identified by pid or proc.
Optional keyword arguments:

• mode: file access mode (modified by the process umask). Defaults to

world-readable.
• poll interval: Specify the maximum time to between attempts (if watch file
doesn’t work)
• stale age: Delete an existing pidfile (ignoring the lock) if its mtime is older
than this. The file won’t be deleted until 25x longer than this if the pid in
the file appears that it may be valid. By default this is disabled (stale age
= 0), but a typical recommended value would be about 3-5x an estimated
normal completion time.

source

478
Chapter 93

SpecialFunctions

93.1 SpecialFunctions.cosint
SpecialFunctions.cosint — Function.

cosint(x)

Compute the cosine integral function of x, defined by Ci(x) := γ + log x +

R x cos t−1
0 t dt for real x > 0, where γ is the Euler-Mascheroni constant.
source

93.2 SpecialFunctions.sinint
SpecialFunctions.sinint — Function.

sinint(x)
Rx sin t
Compute the sine integral function of x, defined by Si(x) := 0 t dt for
real x.
source

479
Chapter 94

NamedTuples

94.1 NamedTuples.delete
NamedTuples.delete — Method.
Create a new NamedTuple with the specified element removed.
source

94.2 NamedTuples.setindex
NamedTuples.setindex — Method.
Create a new NamedTuple with the new value set on it, either overwriting
the old value or appending a new value. This copies the underlying data.
source

480
Chapter 95

OnlineStatsBase

95.1 LearnBase.value
LearnBase.value — Method.

value(o::OnlineStat)

Calculate the value of the stat from its “sufficient statistics”.

source

95.2 StatsBase.fit!
StatsBase.fit! — Method.

fit!(o::OnlineStat, data)

Update the “sufficient statistics” of a stat with more data.

source

481
Chapter 96

Loess

96.1 Loess.loess
Loess.loess — Method.

loess(xs, ys, normalize=true, span=0.75, degreee=2)

Fit a loess model.

Args: xs: A n by m matrix with n observations from m independent predic-
tors ys: A length n response vector. normalize: Normalize the scale of each
predicitor. (default true when m > 1) span: The degree of smoothing, typi-
cally in [0,1]. Smaller values result in smaller local context in fitting. degree:
Polynomial degree.
Returns: A fit LoessModel.
source

482
Chapter 97

WoodburyMatrices

97.1 WoodburyMatrices.liftFactor
WoodburyMatrices.liftFactor — Method.

liftFactor(A)

More stable version of inv(A). Returns a function which computs the inverse
on evaluation, i.e. liftFactor(A)(x) is the same as inv(A)*x.
source

483
Chapter 98

Lazy

98.1 Lazy.takeuntil
Lazy.takeuntil — Method.

takeuntil(pred, list)

Take the elements in list until the pred function return true. Notice that
the one which makes pred true is also taken. All elements will be taken if no
one satisfy the pred function.
source

484
Chapter 99

Requests

99.1 Requests.save
Requests.save — Function.
save(r::Response, path=".")
Saves the data in the response in the directory path. If the path is a direc-
tory, then the filename is automatically chosen based on the response headers.
Returns the full pathname of the saved file.
source

485
Chapter 100

LineSearches

100.1 LineSearches.Static
LineSearches.Static — Method.
Static: defines a static linesearch which returns the initial step length.
Static is intended for methods with well-scaled updates; i.e. Newton, on
well-behaved problems.
source

486
Chapter 101

MemPool

101.1 MemPool.savetodisk
MemPool.savetodisk — Method.
Allow users to specifically save something to disk. This does not free the
data from memory, nor does it affect size accounting.
source

487
Chapter 102

SimpleTraits

102.1 SimpleTraits.istrait
SimpleTraits.istrait — Method.
This function checks whether a trait is fulfilled by a specific set of types.

istrait(Tr1{Int,Float64}) => return true or false

source

488
Chapter 103

BinDeps

103.1 BinDeps.glibc version

BinDeps.glibc version — Method.

glibc_version()

For Linux-based systems, return the version of glibc in use. For non-glibc
Linux and other platforms, returns nothing.
source

489
Chapter 104

LaTeXStrings

104.1 LaTeXStrings.latexstring
LaTeXStrings.latexstring — Method.

latexstring(args...)

Similar to string(args...), but generates a LaTeXString instead of a

String.
source

490