Concurrent Programming in Go basics and programming

Concurrent Programming in Go
11/27/24, 12:28 PM Concurrent Programming in Go
https://zdu.binghamton.edu/cs571/slides/golang/golang.html 1/32
The Need for Concurrency
Due to physical limits, CPU clock speed has basically been frozen since the early
2000's.
Moore's law is still operational and has resulted in on-chip hardware expanding
dramatically. In particular, CPU chips contain multiple cores which allow multiple
parallel threads of execution.
To obtain better performance, programs need to exploit the multiple cores using
techniques for concurrent execution.
Concurrent: multiple threads of control can be active at the same time.
Parallel: multiple threads of control execute simultaneously. Requires multiple
cores.
Performance is not the only motivation for concurrency; sometimes a solution is
most conveniently expressed as concurrent computations.

Concurrency Models
[Taken from Kevlin Henney, Thinking Outside the Synchronization Quadrant, NDC
2017.]
Concurrency Models Continued
A major problem in concurrent programs is mutable shared state. Multiple models for
concurrency include:
Shared Mutable State
Leads to lock-based programming with synchronization primitives like mutexes
and semaphores used to protect shared data. Synchronization is a global property
of the program. Very low-level and error-prone but very popular.
Shared Immutable State
Leads to functional programming exemplified by Haskell or Clojure. Persistent
data (old versions of data always available as long as needed).
Mutable State, No Sharing
Actors model. No shared data. All communication is via passing of immutable
messages. More object-oriented than so-called object-oriented programming.
Golang.
Immutable State, No Sharing
All communication via passing messages between threads which never modify
existing data. Exemplified by Erlang, Elixir.

Go Features
Imperative language.
Only single keyword for all iteration: for.
Types in declarations can be inferred from initializers.
Short declarations.
No need for semicolons except within for statements.
Similar to C in that there is no character data type, rune alias for int32 for
Unicode code points, but string and character literals.
Go Features Continued
Garbage collection.
Error values.
Arrays with constant sizes.
Slices.
Strings are constant byte slices.
No classes, but structurally typed interfaces without needing any implements
relationship.
Concurrency without shared state with channels used for passing data.
More conventional concurrency with shared state.
Batteries included.

Go Hello World
In hello1.go:
package main
import "fmt"
func main() {
fmt.Println("hello world")
}
Log:
$ ./hello1
hello world
Yet Another Hello
In hello2.go:
package main
import (
"fmt"
"os"
)
func main() {
if len(os.Args) < 2 {
fmt.Printf("usage: %s GREETEE...n",
os.Args[0])
os.Exit(1)
}
for _, greetee := range os.Args[1:] {
fmt.Printf("hello %sn", greetee)
}
}
Log:
$ ./hello2 tom sue
hello tom
hello sue

Lissajous GIF in Browser
Use net/http and image library functions.
In main.go and lissajous.go.
Start HTTP server on local machine on port 8000:
$ ./lissajous 8000
Then visit http://localhost:8000/?f=1&p=0.1 where f is the relative frequency of wrt
and p is the phase inc for each iteration.
y x
Concurrent URL Fetch using Go Routines
In fetchall.go
Log:
$ ./fetchall https://golang.org
https://www.binghamton.edu
https://gopl.io
https://godoc.org
0.47s 23662 https://www.binghamton.edu
0.73s 62384 https://golang.org
0.77s 32284 https://godoc.org
0.85s 4154 https://gopl.io
0.85s elapsed

Go Primitive Data
Boolean type bool with literals true and false.
Declarations:
var cond bool; //initialized to false
var toBe = false; //type inferred from initializer
isDone := false; //short declaration
Integer types int, int8, int16 int32, int64. Also corresponding unsigned types
uint, uint8, ; literals in decimal, hexadecimal with 0[xX] prefix, octal with 0
or 0[oO] prefix, binary with 0[bB] prefix. Internal underscores for readability. The
following are all equivalent: 12_34, 0x4d2, 0o2322 or 02322,
0b0100_1101_0010.
A byte is an alias for a uint8 and rune an alias for int32 (representing a unicode
code point).
Declarations:
var i int //initialized to 0
var MaxInt64 uint64 = 1<<64 - 1
age := 42 //short declaration
IEEE 754 floating point types float32 and float64. Literals in decimal or
hexadecimal. Internal underscores for readability. For hexadecimal literals, the
exponent part after p or P scales the mantissa by a power-of-2. Examples: 0., .1e5,
0x2.0p5 (== 64), 0x1.Fp+0 (== 1.9375).
Complex types complex64, complex128.
…
If Statements
Traditional if-then and if-then-else statements.
No parentheses around condition; braces required around statements.
Can precede condition by a declaration.
if x := f(); x % 2 == 0 {
fmt.Print(x*3)
}
else if y := g(x); y < 0 {
fmt.Print(x - y)
}
else {
fmt.Println(x + y)
}

Switch Statements
Basically a multi-branch if. No default fall-through behavior.
Basic switch:
switch coinflip() {
case "heads":
heads++
case "tails":
tails++
default:
fmt.Print("coin landed on edge!")
}
A tagless switch does not have an operand, simply test cases:
func Signum(x int) int {
switch {
case x > 0:
return +1
case x < 0:
return -1
default:
return 0
}
}
Type Switch
Can switch on the type of an expression:
var z int
switch t := x.(type) {
case int:
z = t*2
case string:
z = len(t)*2
default:
z = 1
}

Loops
Uses single keyword for:
C style for-loop: for inits; cond; updates { ... }
No parentheses; must have braces around body.
Omit inits and updates for a while-loop.
Omit inits, cond and updates for a forever loop.
Use for-range to iterate over a range of integers or key-value pairs of a collection.
Allows break and continue statements.
Example loops.go
Arrays
Size of array must be known at compile time (no equivalent of C's variable-length
arrays VLAs).
Two arrays of different sizes are of different types.
Unlike C, always passed by value unless caller explicitly passes a pointer.
const n = 10
var arr [n]string //array of 10 "" strings
//[...] means # of elements fixed by initializer
days [...]string{
"mo", "tu", "we", "th", "fr", "sa", "su",
}

Slices
Allows using arrays with dynamic sizing.
A slice for base type T looks like:
struct {
T *arr //backing array
int len, capacity
}
Declaration looks like an array declaration, but no size specification. For example,
var bytes []byte will declare bytes as a nil byte-slice.
words := []string{"hello", "world"} will declare words as a slice into a
string array, wih len(word) == cap(words) == 2.
ints := make([]int, 3, 5) will declare ints as a slice into an int array
with len(ints) == 3 and cap(ints) == 5. The first len elements of the slice
will be initialized to 0.
Very efficient subslicing: given slice a, a[lo:hi], a[lo:], a[:hi] returns sub-
slices with inclusive lo bound and exclusive hi bound.
Slices share the underlying array entries. Example days.go:
Appending to a Slice
If sliceT []T and v1 T, , vn T, then append(sliceT, v1, ..., vn)
returns sliceT with v1, , vn appended onto its end.
If an append() requires exceeding the capacity of the original backing array, then
the backing array is reallocated.
append.go shows that appending reallocates underlying array so that backing array
changes.
copy(destSlice, srcSlice) copies srcSlice to destSlice.
…
…

Strings
A string is an immutable slice of bytes. Since byte slices are mutable, use []byte
slices or bytes.Buffer() for efficient string mutation.
May contain a '0' bytes.
Often interpreted as UTF-8 encoded sequence of Unicode code points. Go's range
loop over a string will process a character at a time (rather than a byte at a time).
More flexibility from unicode/utf8 package. Unicode replacement character
ufffd � produced for invalid utf8 encoding.
len(s) returns # of bytes (not runes) in string s.
s[i] returns i'th byte with a panic if 0 < i || i >= len(s). Cast string to a
rune array to get i'th character: []rune(s)[i].
String concatentation using + and += produce new strings.
Lexicographical string comparison using usual relational operators.
Since strings are immutable, slice operations s[i:j] (substrings) are efficient since
they share the same byte sequence.
String literals within " can contain conventional C-style escape sequences as well as
uhhhh or Uhhhhhhhh for 16-bit or 32-bit Unicode code points.
Raw string literals enclosed within ` (back-quotes) can contain any characters other
than back-quote. Carriage returns r are deleted. Useful for regexs, HTML
templates and JSON literals, etc.
Characters versus Bytes
From strbytes.go
func main() {
hellos := map[string]string{
"en": "Hello World!", //english
"es": "¡Hola Mundo!", //spanish
"hi": "नमस्ते दुनिया!", //hindi
"ja": "こんにちは、 ", //japanese
"pt": "Olá Mundo!", //portuguese
"zh": " !", //chinese
}
for code,greet := range hellos {
fmt.Printf("%s: bytes[%d] = % x; " +
"chars[%d] = %qn",
code, len(greet), greet,
utf8.RuneCountInString(greet),
[]rune(greet))
}
}

Characters versus Bytes: Log
Edited Log:
$ ./strbytes
zh: bytes[13] = e4 b8 96 e7 95 8c e6 82 a8 e5 a5 bd 21;
chars[5] = [' ' ' ' ' ' ' ' '!']
en: bytes[12] = 48 65 6c 6c 6f 20 57 6f 72 6c 64 21;
chars[12] = ['H' 'e' 'l' 'l' 'o' ' '
'W' 'o' 'r' 'l' 'd' '!']
es: bytes[13] = c2 a1 48 6f 6c 61 20 4d 75 6e 64 6f 21;
chars[12] = ['¡' 'H' 'o' 'l' 'a' ' '
'M' 'u' 'n' 'd' 'o' '!']
hi: bytes[38] = e0 a4 a8 e0 a4 ae e0 a4 b8 e0 a5 8d e0
a4 a4 e0 a5 87 20 e0 a4 a6 e0 a5 81 e0 a4
a8 e0 a4 bf e0 a4 af e0 a4 be 21;
chars[14] = ['न' 'म' 'स' '् ' 'त' 'े ' ' '
'द' 'ु ' 'न' 'ि' 'य' 'ा' '!']
ja: bytes[25] = e3 81 93 e3 82 93 e3 81 ab e3 81 a1 e3 81
af e3 80 81 20 e4 b8 96 e7 95 8c;
chars[9] = ['こ' 'ん' 'に' 'ち' 'は' '、' ' ' ' ' ' ']
pt: bytes[11] = 4f 6c c3 a1 20 4d 75 6e 64 6f 21;
chars[10] = ['O' 'l' 'á' ' ' 'M' 'u' 'n' 'd' 'o' '!']
Maps
Go supports maps from any type which implements equality to an arbitrary type.
Syntax for type is map[K]V where K is the type of the key and V is the type of
value. Examples:
var counts map[string]int
colorSpecs := make(map[string]string)
colorIndexes := map[string]int{"black": 0, "red": 1}
for-range loops over key-value pairs. Given the above colorIndexes:
for c, i := range colorIndexes {
fmt.Printf("%q: %dn", c, i)
}
// "black": 0
// "red": 1

Pointers
Variables which hold addresses.
Cannot take address of a slice element. Why?
Introduces aliasing.
No pointer arithmetic.
i, j := 33, 42
p := &i
j = *p; //j == 33
*p = 99 //i == 99
fmt.Println(i, j) //99, 33
Structures
Collection of heterogenous fields.
type Circle struct {
Label string
X, Y int
Radius int
}
Note uppercase on field names to allow access outside package.
Can refer to fields using . notation for both struct values and struct pointers.
c1 := Circle{Label: "c1", X: 1, Y: 1}
c1P := &c1
fmt.Printf("struct circle %s at (%g, %g)n",
c1.label, c1.X, c1.Y)
fmt.Printf("pointed to circle %s at (%g, %g)n",
c1P.label, c1P.X, c1P.Y)

Structure Embedding
Can embed structures anonymously:
type Point struct { X, Y int }
type Circle struct {
Point
Radius string
}
type Rect struct {
Point
Width, Height int
}
var c1 Circle //initialized with "zero" values
c1.X = 1 //ok: c1.Point.X = 1
c1.Y = 1 //ok: c1.Point.Y = 1
//cannot initialize through anon struct directly
c1 = { X: 2, Y: 2 } //error, must use intermediate
c2 = { Point{X: 2, Y: 2}, Radius: 2 } //ok
Goroutines
Simply precede function call with go keyword, to run call in a separate goroutine.
Goroutines are ike a thread, but very lightweight. Can easily have tens or hundreds
of thousands.
An example which uses a goroutine to display a spinner while performing a
computation: fib-spinner.go.

Channels
Channels are used for communicating between and synchronizing goroutines.
Like Unix pipes but Unix pipes are untyped byte streams, whereas go channels
transmit typed data.
Declare a channel having element type T using chan T and create using make():
ch := make(chan string) //ch has type chan string
A channel can be both read and written by a goroutime, but typically a goroutine
only reads a channel or only writes a channel. We can declare a readonly int
channel as <-chan int and a writeonly int channel as chan<- int.
Channel Example
First goroutine converts command-line args to int.
Subsequent goroutines compute function.
Main goroutine outputs.
Termination because each goroutine closes output channel.
1 2 3 4 go
os.Args[1:]
go
n + 1
go
2 * n
go
n * n
main
output
16 36 64 100
In fn-pipes.go

Multiplexing using select
We may need to receive from first among several channels which has data without
blocking on the others.
Use the select statement:
select {
case <- ch1:
// ch1 had discarded data
case x <- ch2:
// process x read from ch1
...
case ch3 <- y:
// ch3 was able to receive y
default:
// no channel was ready
}
select {} blocks for ever.
If multiple channels are ready, select will choose one at random, to ensure
fairness.
Countdown Timer
In countdown.go
func main() {
fmt.Println("counting down ...")
ticker := time.NewTicker(1 * time.Second)
//don't care about value
abort := make(chan struct{})
go func() { //reads garbage from stdin
os.Stdin.Read(make([]byte, 1))
abort <- struct{}{}
}() //called IIFE in JS
for count := 5; count > 0; count-- {
select {
case <- ticker.C: //ticker.C is chan
fmt.Printf("%d... ", count)
case <- abort:
fmt.Println("ABORTED!")
return
}
}
fmt.Println("LAUNCH!")
}
Log:
$ ./countdown
counting down ...
5... 4... 3... 2... 1... LAUNCH!
$ ./countdown
counting down ...
5... 4... # typed input to console
ABORTED!
$

Not Covered
Interfaces.
Buffered channels.
Goroutine cancellation.
Concurrency with shared data: mutexes.
References
Alan A. A. Donovan, Brian W. Kernighan, The Go Programming Language, 1st Edition,
Addison-Wesley, 2016.

Concurrent Programming in Go basics and programming

Recommended

More Related Content

Similar to Concurrent Programming in Go basics and programming (20)

Recently uploaded (20)

Concurrent Programming in Go basics and programming