Scala @ TomTom

1. Scala @ TomTomEric [email protected] May 2010

2. What’s He Building In There?Not just PNDsTraffic ServicesLocation ServicesAddress & Local SearchTons of Data -> Tons of AnalysisDeep Commitment to Server-Side Java2

5. 5Company Confidential21,345 pages

6. 40 reams of paper

7. 5 laptops

8. 8 keyboards

9. 1 back operation

10. 30,000 strands of hair6

12. Tutorial on Good Lisp Programming Style8“Most algorithms can be characterized as: Searching (some find find-if mismatch)

13. Sorting (sort merge remove-duplicates)

14. Filtering (remove remove-if mapcan)

15. Mapping (map mapcarmapc)

16. Combining (reduce mapcan)

17. Counting (count count-if)These functions abstract common control patterns. Code that uses them is: Concise

18. Self-documenting

19. Easy to understand

20. Often reusable

21. Usually efficient” From Peter Norvig’sTutorial on Good Lisp Programming Style, 1993 (http://www.norvig.com/luv-slides.ps)

22. But...“Lisp is good for:Exploratory programmingRapid prototypingMinimizing time-to-marketSingle-programmer (or single-digit team) projectsSource-to-source or data-to-data transformation” (ibid.)9

23. I LikeStatic typesRefactoring ToolsLazy SundaysPaying Some Kid To Cut The GrassSpreadsheetsNot Flying The Plane10

24. What is this Scala thing?11

25. Scala Is...An escaped research language from EPFL targeting JVM and CLRObject-oriented and functionalStatically typedLisp, ML, Haskell, Erlang, etc.DSL-friendly12

26. Bi-Scalar Disorder13

27. Functional Programming?Functions as first class objectsImmutabilityClosuresBinding free variables to enclosing lexical scope Higher-order Functionsfunctions as input and/or outputA different set of idioms...14

28. Keep Backfor (int i = 1; i <= 256; i++) { if (array[i-1] == ...15

29. LOLCrash16

30. IdiomsReduce cognitive overheadReduce bugsMake intention clearMini-Patterns17

31. Java Idioms++C++No more 0xDEADBEEFLeads to lots of loops and copies, if you’re doing it rightHard programs get complex doing common thingsNested loops begin to look Harmful...18

32. 19

33. For Comprehensionsfor(inti=0; i<100; i++) { ... }for (i <- 0 until 100) { ... /* do something with i */ }(0.until(100)).foreach(i => /* something with i */)20

34. For ComprehensionsLists algorithmic “sweet spot”Syntactic Sugar for:foreachmapfilterflatMap21

35. The Easy Onesfor (i <- 1 to 6) yield i * 2(1 to 6).map(_ * 2) (2,4,6,8,10,12)for (i <- 1 to 6 if i % 2 == 0) yield i(1 to 6).filter(_ % 2 == 0) (2,4,6)for (i <- 1 to 6) { println(i + “ “) }(1 to 6).foreach { i => print(i + “ “) } 1 2 3 4 5 622

36. A Harder One...List<Integer> array = new ArrayList<Integer>();for (i = 1; i <= 3; i++) { for (j = i; j <= 3; j++) {array.add(j); }}System.out.println(array);[1, 2, 3, 2, 3, 3]for (i <- 1 to 3; j <- i to 3) yield j(1, 2, 3, 2, 3, 3)(1 to 3).flatMap(i => (i to 3).map(j => j)) 23

37. flatMapSubtle thing...“Applies the given function f to each element, then concatenates the results”Turns a list of lists into a listList(List(1,2,3), List(4,5,6)).flatMap(x => x)List(1, 2, 3, 4, 5, 6)List(“tom”, “tom”).flatMap(_.capitalize).mkStringTomTom“Special sauce” for nested looping constructs(Equivalent to Haskell’s monadic “bind”)24

38. IteratorIteratorpackage org.hyperic.sigar.util; public static class IteratorIteratorimplements java.util.Iterator { private java.util.ArrayListiterators; public IteratorIterator() {} public void add(java.util.Iteratoriterator) {} public booleanhasNext() {} public java.lang.Object next() {} public void remove() {} }25

39. flatMapflatMapdef foo(arg: Iterable[Int]) { ... Do something with arg}val some: Iterable[Int] = getSomeIterablefoo(some)val more: Iterable[Int] = getMorefoo(List(some, more).flatMap(x => x))26

40. Real-Life Example“mntstamarg”Each term has aliasesNeed all permutations27

41. Java Versionstatic List<String> aliasReplace(String place) { String[] bits = nospaces.split(place); List<String> result = new ArrayList<String>(); List<StringBuilder> allAliases = new ArrayList<StringBuilder>(); for (String bit : bits) { String[] ales = aliases.get(bit); if (ales != null) { if (allAliases.size() == 0) { allAliases.add(new StringBuilder(bit)); for (String alias : ales) { allAliases.add(new StringBuilder(alias)); } } else { List<StringBuilder> clones = new ArrayList<StringBuilder>(); for (StringBuilder a : allAliases) { clones.add(new StringBuilder(a).append(" ").append( bit)); for (String alias : ales) { clones.add(new StringBuilder(a).append(" ").append( alias)); } } allAliases = clones; } } else { if (allAliases.size() == 0) { allAliases.add(new StringBuilder(bit)); } else { for (StringBuilder b : allAliases) { b.append(" ").append(bit); } } } } for (StringBuildermunge: allAliases) { result.add(munge.toString()); } return result; }28

42. 29

43. Scala Versiondef alias(query: String, aliases: String => List[String]): List[String] = { def recurse(prefix: List[String], remainder: List[String]): List[List[String]] = { remainder match { case Nil => prefix :: Nilcase head :: tail => aliases(head).flatMap(term => recurse(term :: prefix, tail)) }recurse(Nil,query.split(“\\s”).toList.reverse).map( _.mkString(“ “))}30

44. 31

45. Yeah yeahyeah but“Perl Whitespace Law” “Each line of perl should be surrounded by whitespace equivalent to what it would take to achieve the same functionality in a normal programming language.” -- Don HopkinsIf it compiles, it nearly works. Really.Visual Plane-Oriented ProgrammingI ♥ Idioms“But in Java, each little part is so very simple...”32

46. 33

47. Discovering Fire34

48. “Weak developers will move heaven and earth to do the wrong thing. You can’t limit the damage they do by locking up the sharp tools. They’ll just swing the blunt tools harder.” – Glenn Vandenburg35

49. @TomTomTestingMiddleware “Smart Content Switch”We needed it quickly...B2B/B2G Traffic Query EngineClustering AlgorithmDSL Templating EngineContent CleanupNext-Generation Geocoder36

50. How To Sneak It InStart with TestingScalaCheckIS AWESOME.Specs, ScalaTest testing DSLsStart with something low riskOk, well, we didn’t do thatPrepare for steep learning curve...followed by a productivity hockey stick37

51. Another Example8000 lines of broken Java -> 400 lines of broken Scala ->hyp.take(1).flatMap(_.dropDistricts) match { case Nil => hyp case head => {hyp.tail.foldLeft(head) {case (run: List[Hypothesis], h: Hypothesis) => { run.flatMap(_.merge(h)) match { case Nil => run case newRun => newRun.removeDuplicates} } } }}*Includes suggested improvements by Martin Odersky, I hope38

52. TestingFunctional architect wrote a bunch of test cases, like:Requirement R.17:D1 -> N1, L1 -> N1, N1, D2 -> L2, L3, N3, D3 should cluster as (L1,N1,D1), (L2,D2), (L3), (N3), (D3)Vim-macro’d into:it should “satisfy R.17” in { cluster(D1 -> N1,L1 -> N1,N1,D2 -> L2,L3,N3,D3) should equal {groups(group(L1,N1,D1),group(L2,D2),group(L3),group(N3),group(D3)))}39

53. ScalaCheckThe Dog’s BollocksSteep Learning Curveobject BinarySpecification extends Properties(“bin”) { specify(“toBinary”,Prop.forAllNoShrink( for{n <- Gen.choose(0d, 0.99999999d) m <- Gen.choose(20,31) } yield Pair(n, m)) { p => Math.abs(toDecimal(toBinary(p._1, p._2)) – p._1 < 1e-10 })}40

54. So little time...TraitsCase classesPattern matchingStructural typesSelf typesXMLOptionCovariance/Contravariance@specialized41

55. This slide left intentionally blank.

Scala @ TomTom

More Related Content

What's hot (20)

Viewers also liked (7)

Similar to Scala @ TomTom (20)

Recently uploaded (20)

Scala @ TomTom