The Power of Types, KotlinConf 2019

From Assembler, over Fortran and C to modern Kotlin, we came a long way and improved the way we can express our thoughts in code.

One thing that happened during this journey is that languages allow us to use types for our problem domain, independent of the underlying computer architecture. Types became a powerful tool.

Types improve readability which probably is the most important aspect of programming! But types also prevent you from making mistakes at compile time. This is why Kotlin translated the null problem into the type system.

We can use these powers in our daily lives. We can write better code by avoiding “primitive obsession”. I’ll show you how!

Transcript

1. Copenhagen
   Denmark
   THE POWER OF TYPES
   DANNY PREUSSLER
   @PreusslerBerlin
   

   View Slide

2. View Slide

3. Do you enjoy
   debugging?
   

   View Slide

4. In an ideal world,
   in the right language,
   buggy code is impossible.
   The compiler simply refuses
   to build
   

   View Slide

5. In an ideal world,
   in the right language,
   buggy code is impossible.
   The compiler simply refuses
   to build
   

   View Slide

6. In an ideal world,
   in the right language,
   buggy code is impossible.
   The compiler simply refuses
   to build
   

   View Slide

7. In an ideal world,
   in the right language,
   buggy code is impossible.
   The compiler simply refuses
   to build
   

   View Slide

8. No language can prevent
   us from writing bugs!
   

   View Slide

9. But we can avoid
   as many as possible,
   as early as possible
   

   View Slide

10. TYPES
    

    View Slide

11. Well-typed expressions
    do not go wrong
    A theory of type polymorphism in programming.
    Journal of Computer and System Sciences, Robert Milner, 1978
    

    View Slide

12. WHAT IS A TYPE?
    

    View Slide

13. abstraction mechanisms
    over data
    Several Types of Types in Programming Languages
    Simone Martini, Universit`a di Bologna
    

    View Slide

14. A type is a concise, formal
    description of the behavior of
    a program fragment
    Type systems for programming languages,
    Didier Rémy, INRIA Paris
    

    View Slide

15. THE HISTORY OF TYPES
    

    View Slide

16. 1950 1960 1970 1980 1990 2000 2010 2020
    

    View Slide

17. Types are
    pretty much absent
    1950 1960 1970 1980 1990 2000 2010 2020
    

    View Slide

18. 1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    

    View Slide

19. “Two types of constants are permissible: fixed
    points (restricted to integers) and floating points”
    FORTRAN manual, 1956
    But at same time, “mode” is used instead
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    

    View Slide

20. 1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Types
    were born
    

    View Slide

21. Integers are of type integer.
    All other numbers are of type real.
    Algol 60
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    

    View Slide

22. the results of any program [..]
    should be comprehensible
    without knowing anything about the machine
    or its storage layout.
    Tony Hoare 1960
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    

    View Slide

23. I realised that [types] were essential
    not only for determining memory requirements,
    but also for avoiding machine-dependent error
    Tony Hoare 1960
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    

    View Slide

24. 1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    1966
    Algol W
    

    View Slide

25. One missing piece
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    

    View Slide

26. the type system
    (including user-defined types)
    [should] guarantee that only the prescribed
    operations on a type could operate on its values
    James H. Morris. Types are not sets, New York, NY, USA, 1973. ACM
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    

    View Slide

27. 1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    Pascal
    

    View Slide

28. Types became
    a central feature of
    programming languages
    as we understand them today
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    Pascal
    

    View Slide

29. The dawn of
    Object Orientation
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    Pascal
    C++
    

    View Slide

30. Type structure is
    a syntactic discipline
    for enforcing levels of
    abstraction
    Types, abstraction and parametric polymorphism, John C. Reynolds, 1983
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    Pascal
    C++
    

    View Slide

31. The age of
    the internet languages
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    Pascal
    C++
    Java
    

    View Slide

32. Object-oriented languages
    had made static typing so
    painful that many preferred to
    give up and just wing it.
    Ben Lynn, https://benlynn.blogspot.com/
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    Pascal
    C++
    Java
    

    View Slide

33. Dynamic
    Typed
    languages
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    Pascal
    C++
    Java
    JS
    Ruby
    

    View Slide

34. It is important to note
    that this is different
    than being typeless
    Dynamically Typed Languages,
    Laurence Tratt, Bournemouth University
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    Pascal
    C++
    Java
    JS
    Ruby
    

    View Slide

35. the chief technical difference
    between them being when
    types are enforced
    Dynamically Typed Languages,
    Laurence Tratt, Bournemouth University
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    Pascal
    C++
    Java
    JS
    Ruby
    

    View Slide

36. After you ship, it costs you
    about 10,000 times as much
    to fix a software bug
    Bill Joy, Java-One, 1999
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    Pascal
    C++
    Java
    JS
    Ruby
    Java Generics
    

    View Slide

37. Functional
    Programming
    arises from the grave
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    Pascal
    C++
    Java
    JS
    Ruby
    Scala
    Java Generics
    

    View Slide

38. Functional programming languages
    are traditionally typed
    (Scheme and Erlang are exceptions)
    Type systems for programming languages,
    Didier Rémy, INRIA Paris
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    Pascal
    C++
    Java
    JS
    Ruby
    Scala
    Java Generics
    

    View Slide

39. Anders Hejlsberg, 2012
    you can write large programs in
    JavaScript.
    You just can’t maintain them.
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    Pascal
    C++
    Java
    JS
    Ruby
    Scala
    Typescript
    Java Generics
    

    View Slide

40. The typed v.s. untyped flavor [..]
    should not be confused
    with [..] whether types [..]
    are explicit or implicit.
    Type systems for programming languages, Didier Rémy, INRIA Paris
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    Pascal
    C++
    Java
    JS
    Ruby
    Scala
    Typescript
    Java Generics
    

    View Slide

41. type reconstruction
    starts spreading
    1950 1960 1970 1980 1990 2000 2010 2020
    Fortran
    Algol 60
    Algol W
    Pascal
    C++
    Java
    JS
    Ruby
    Scala
    Kotlin
    Typescript
    Java Generics
    

    View Slide

42. KOTLIN AND TYPES
    

    View Slide

43. The type system is
    Kotlin’s super power
    Nullability,
    1st class functions,
    Nothing
    

    View Slide

44. If Kotlin uses
    the power of Types
    So can you!
    

    View Slide

45. PRIMITIVE OBSESSION
    

    View Slide

46. Usage of primitives
    instead of small objects for simple tasks
    

    View Slide

47. Usage of constants for coding information
    

    View Slide

48. View Slide

49. View Slide

50. View Slide

51. View Slide

52. View Slide

53. View Slide

54. View Slide

55. Don’t put in name
    what can be in type
    

    View Slide

56. Encoding the type [..]
    into the name [..]
    is brain damaged
    the compiler knows the types
    Linus Torvalds about Hungarian notation
    

    View Slide

57. View Slide

58. View Slide

59. View Slide

60. View Slide

61. View Slide

62. actually...
    

    View Slide

63. actually...
    

    View Slide

64. View Slide

65. Don’t put in name what could be in type!
    

    View Slide

66. Express your Domain knowledge
    

    View Slide

67. Sample:
    At Soundcloud a URN
    uniquely identifies an entity:
    a user, track playlist…
    soundcloud:tracks:1234
    

    View Slide

68. View Slide

69. Anytime you find yourself writing code
    of the form
    if the object is of type T1, then
    do something, but if it’s of type
    T2, then do something else,
    slap yourself.
    Scott Meyers, Effective C++
    

    View Slide

70. “type” properties/functions are a smell
    

    View Slide

71. Depending properties might be hint for hidden type
    

    View Slide

72. Factory methods are hint for different types
    

    View Slide

73. View Slide

74. View Slide

75. View Slide

76. View Slide

77. Booleans are a smell: 2 things in 1 place
    

    View Slide

78. View Slide

79. Don’t let the creep in
    

    View Slide

80. View Slide

81. View Slide

82. View Slide

83. View Slide

84. View Slide

85. View Slide

86. View Slide

87. COSTS
    

    View Slide

88. ALLOCATION COSTS
    

    View Slide

89. On JVM,
    objects allocated in Eden space
    

    View Slide

90. To avoid thread synchronization
    Eden is split into
    Thread Local Allocation Buffer
    

    View Slide

91. Allocating space in TLAB
    is often just a pointer bump
    

    View Slide

92. The compiler might even
    inline the whole class:
    Allocation elimination
    with escape analysis
    Urban performance legends, revisited, Brian Goetz 2005
    

    View Slide

93. Allocations on JVM
    are very cheap
    

    View Slide

94. A word on Android
    ART is optimized for allocations
    

    View Slide

95. Allocations are OK
    Types are OK
    Yes, even enums
    Modern Android development:
    Android Jetpack, Kotlin, and more (Google I/O 2018)
    

    View Slide

96. MEMORY COSTS
    

    View Slide

97. Memory
    A plain Object takes 8 bytes
    On Sun's JDK 1.3.1 for Windows
    

    View Slide

98. Memory
    Any wrapper will be
    8 bytes larger
    than wrapped primitive
    On Sun's JDK 1.3.1 for Windows
    

    View Slide

99. But careful
    

    View Slide

100. Memory
     An empty String
     takes 40 bytes
     On Sun's JDK 1.3.1 for Windows
     

     View Slide

101. Memory
     Boolean:
     header has 12 bytes
     + 1 for the boolean
     + 3 for the granularity
     On HotSpot JVM 1.7.0_51
     

     View Slide

102. Memory
     its "size" isn't something
     that's precisely defined
     https://docs.oracle.com/javase/tutorial/java/nutsandbolts/datatypes.html
     

     View Slide

103. Kotlin compiler translates to int
     

     View Slide

104. Kotlin compiler translates to java.lang.Integer
     

     View Slide

105. Kotlin compiler can use primitive int inside Age
     

     View Slide

106. Kotlin compiler translates to int
     

     View Slide

107. Kotlin keeps Age with int
     

     View Slide

108. GARBAGE COSTS
     

     View Slide

109. the performance of the garbage
     collector is not determined by the
     number of dead objects but rather by
     the number of live ones
     Analyzing Java Memory, Java enterprise performance book
     

     View Slide

110. Short lived objects are cheap
     

     View Slide

111. SUM UP
     

     View Slide

112. ● Types improve readability (documentation)
     ● Types help compile time correctness
     (safety, machine checked documentation)
     ● Types improve abstraction (modularity)
     -> Types are enabler
     Sum up
     

     View Slide

113. Watch out for “hidden type” smells:
     ● Types information in names
     ● Type attributes
     ● Booleans
     ● Primitives
     Sum up
     

     View Slide

114. My goal was to ensure that all
     use of references should be
     absolutely safe,
     with checking performed
     automatically by the compiler
     Tony Hoare about Algol W
     

     View Slide

115. But I couldn't resist the
     temptation to put in a null
     reference, simply because it
     was so easy to implement.
     Tony Hoare about Algol W
     

     View Slide

116. call it my
     billion-dollar mistake
     Tony Hoare about Algol W
     

     View Slide

117. Readability
     Let the compiler
     do the work!
     

     View Slide

118. TYPE
     ALL THE THINGS!
     

     View Slide

119. ● Several Types of Types in Programming Languages
     Simone Martini, Universit`a di Bologna https://arxiv.org/pdf/1510.03726
     ● History and Philosophy of Computing
     Third International Conference, HaPoC 2015, Pisa, Italy
     ● Type systems for programming languages
     Didier Rémy, INRIA Paris, http://gallium.inria.fr/~remy/mpri/cours1.pdf
     ● A theory of type polymorphism in programming.
     Journal of Computer and System Sciences, Robert Milner, 1978
     https://homepages.inf.ed.ac.uk/wadler/papers/papers-we-love/milner-type-polymorphism.pdf
     ● Dynamically Typed Languages
     Laurence Tratt,Bournemouth University, 2009
     https://tratt.net/laurie/research/pubs/html/tratt__dynamically_typed_languages/
     Sources
     History of types
     

     View Slide

120. ● Type Wars, The Clean Code Blog, Uncle Bob
     http://blog.cleancoder.com/uncle-bob/2016/05/01/TypeWars.html
     ● From Primitive Obsession to Domain Modelling, Mark Seemann
     https://blog.ploeh.dk/2015/01/19/from-primitive-obsession-to-domain-modelling/
     ● Phantom types in Kotlin, Danny Preussler
     https://proandroiddev.com/phantom-types-in-kotlin-afd3f59fde10?
     ● Primitive Data types, Oracle
     https://docs.oracle.com/javase/tutorial/java/nutsandbolts/datatypes.html
     Sources
     Types
     

     View Slide

121. ● The cost of object creation in Java, including garbage collection
     https://www.bettercodebytes.com/the-cost-of-object-creation-in-java-including-garbage-collection/
     ● Thread-Local Allocation Buffers in JVM
     https://alidg.me/blog/2019/6/21/tlab-jvm
     ● Urban performance legends, revisited
     https://www.ibm.com/developerworks/java/library/j-jtp09275/index.html#artrelatedtopics
     ● E-book: Java enterprise performance
     https://www.dynatrace.com/resources/ebooks/javabook/
     ● Do you know your data size?
     https://www.javaworld.com/article/2077496/java-tip-130--do-you-know-your-data-size-.html
     ● The impact of Garbage Collection on Application Performance
     https://www.dynatrace.com/resources/ebooks/javabook/impact-of-garbage-collection-on-performance/
     Sources
     Costs
     

     View Slide

122. #KotlinConf
     THANK YOU
     AND
     REMEMBER
     TO VOTE
     Danny Preussler @ PreusslerBerlin
     

     View Slide