Functional programming is one of the possible paradigm that we can use to write applications using Scala, In this talk you will learn about: * The benefits of using Functional programming approach * The type classes in Scala * And the common functional abstractions: semigroup, monoid, functor and monad The challenge that the newcomers who were writing imperative code and want to start using Scala is: to change the way of thinking and reasoning about the code in a Functional way, in this talk I will go through the details and explain functional programming paradigm and how to work with type classes in Scala.

1. Friendly Functional
Programming
Wiem Zine Elabidine
@WiemZin
フレンドリーな関数型プログラミング

2. Scary Functional
Programming
Wiem Zine Elabidine
@WiemZin
Functor
Monad
怖い関数型プログラミング

3. Friendly Functional
Programming
Wiem Zine Elabidine
@WiemZin
フレンドリーな関数型プログラミング

4. Plan
Functional
Programming
Types and Functions Type classes
今日の予定、fp概要、型と関数 型クラスの流れで説明します

5. Paradigm
Is a style and a way of writing your programs.
Each programming language follow a programming
style.
パラダイムはプログラムを書く作法

6. Paradigm
Is a style and a way of writing your programs.
Each programming language follow a programming
style.
Scala combines
object-oriented and
functional programming.
Scalaは、オブジェクト指向プログラミングと関数型プログラミングを組み合わせたものです。

7. Daniel Sophie

8. Daniel

12. Sophie

15. Declarative
Imperative
How I can do it?
Declarative Mindset
What I want to do?
Imperative Mindset
宣言型の考え方と命令型の考え方

16. Functional programming
A form of declarative programming
関数型プログラミングは宣言型プログラミングの1形態

17. immutability
03
Data & Functions Pure functions
Composition
01 02
04
Functional programming
大まかに4つに分けて見ていきます

18. Pure functions
Total Deterministic Free of side
effects
純粋関数は決定的かつ副作用を持たない全域関数

19. Referential transparency
When the function could be replaced by the result of
its evaluation without affecting the meaning of the
program.
def sum(a: Int, b: Int): Int = a + b
sum(2, 3) == 5
参照透過性は関数をその評価結果と置き換えても プログラムの振る舞いが変わらないこと

20. ● Testability
● Refactoring
● Maintainability
The benefits of
FP
FPの利点はテストのしやすさ メンテナビリティー

21. Types & Functions
Abstraction
型と関数について

22. “A data types or a simple type is an attribute of data which tells
the compiler or interpreter how the programmer intends to use
the data.” -- Wikipedia
Data Types
データ型はデータ使用の意図を表す

23. “A data types or a simple type is an attribute of data which tells
the compiler or interpreter how the programmer intends to use
the data.” -- Wikipedia
Data Types
If it compiles, it
works!
コンパイルすれば動作します！

24. “Is a type formed by combining other types.” -- Wikipedia
ADT (Algebraic Data Type)
代数的データ型は型の組み合わせ

25. Value types
Int
String
Double
Unit
Boolean ...
値型

26. Product types
Setting(true, true)
Setting(false, true)
Setting(false, true)
Setting(false, false)
case class Setting(
audioEnabled: Boolean,
videoEnabled: Boolean
)
直積型

27. Product types
Setting(true, true)
Setting(false, true)
Setting(false, true)
Setting(false, false)
2 x 2 = 4
4 possible values
case class Setting(
audioEnabled: Boolean,
videoEnabled: Boolean
)
直積型

28. Sum types
sealed trait Color
case object Purple extends Color
case object Red extends Color
case object Yellow extends Color
case object Green extends Color
?
?
?
?
val color: Color = ???
直和型

29. Sum types
sealed trait Color
case object Purple extends Color
case object Red extends Color
case object Yellow extends Color
case object Green extends Color
?
?
?
?
val color: Color = ???
4 possible values
直和型

30. case class Form(
topic: String,
bgcolor: Color
)
Geometry
ADT
?
?
?
?
val form: Form =
Form(topic = ???, color = ???)
∞

31. sealed trait Topic
case object Art extends Topic
case object History extends Topic
case object Philosophy extends Topic
case object Geography extends Topic
ADT

32. case class Form(
topic: Topic,
bgcolor: Color
)
Geometry
ADT
?
?
?
?
val form: Form =
Form(topic = ???, color = ???)
?
?
?

33. case class Form(
topic: Topic,
bgcolor: Color
)
Geometry
ADT
?
?
?
?
val form: Form =
Form(topic = ???, color = ???)
?
?
?
4 x 4 = 16
16 possible values

34. sealed trait Option[+A]
case object None extends Option[Nothing]
case class Some[A](a: A) extends Option[A]
Parameterized ADT
val choice: Option[Boolean] =
???
Some(true)
Some(false)
None
パラメータ化された代数的データ型

35. sealed trait Option[+A]
case object None extends Option[Nothing]
case class Some[A](a: A) extends Option[A]
Parameterized ADT
val choice: Option[Int] =
???
Some(1)
Some(157)
None
Some(2)
Some(3)
Some(4)
Some(5)
Some(6)
Some(7)
Some(8)
Some(158)
Some(159)
Some(845)
Some(880)
Some(8810)
Some(81945)
Some(819447)
Some(36841840)
パラメータ化された代数的データ型

36. FP is the practice of composing programs using pure functions.
Functions
は純粋関数を用いてプログラムを書くための方法

37. A function takes data as an input and returns a data as an
output
Functions
def isPositive(n: Int): Boolean = n > 0
関数はデータを受け取りデータを返す

38. A function could be defined:
● As a data type.
First class functions
type Predicate = Int => Boolean
関数そのものもデータ型

39. A function could be defined:
● As a data type.
● Using a literal syntax.
First class functions
(n: Int) => n > 0
関数リテラル構文を用いて定義できる

40. Higher order functions accept other functions as input and/or return
functions as output.
Higher-order functions
高階関数は別の関数を受け取るか 関数を戻り値として返す

41. Higher order functions accept other functions as input and/or return
functions as output.
Higher-order functions
def checkOption(option: Option[Int], p: Int => Boolean): Boolean =
option match {
case Some(v) if p(v) => true
case _ => false
}

42. Higher order functions accept other functions as input and/or return
functions as output.
Higher-order functions
def checkOption(option: Option[Int], p: Int => Boolean): Boolean =
option match {
case Some(v) if p(v) => true
case _ => false
}
val isPositive: Int => Boolean = n => n > 0
checkOption(Some(14), isPositive)

43. Higher order functions accept other functions as input and/or return
functions as output.
Higher-order functions
def checkOption(option: Option[Int], p: Int => Boolean): Boolean =
option match {
case Some(v) if p(v) => true
case _ => false
}
val isPositive: Int => Boolean = _ > 0
checkOption(Some(14), isPositive)

44. Higher order functions accept other functions as input and/or return
functions as output.
Higher-order functions
def checkOption(option: Option[Int], p: Int => Boolean): Boolean =
option match {
case Some(v) if p(v) => true
case _ => false
}
checkOption(Some(14), n => n > 0)

45. Higher order functions accept other functions as input and/or return
functions as output.
Higher-order functions
def checkOption(option: Option[Int], p: Int => Boolean): Boolean =
option match {
case Some(v) if p(v) => true
case _ => false
}
checkOption(Some(14), _ > 0)

46. Type classes
Add features to your types
型クラスは型に機能を与える

47. Type class
Is a way to define certain operations for a given type..
A type class instance for a given type provides an
implementation for these operations.
型に応じた演算を定義する方法 実装は特定の型クラスのインスタンスで定義される

48. Interface
trait Translator {
def translate(
source: String,
from: String,
to: String): String
}
翻訳インターフェイス

49. Interface
trait Translator {
def translate(
source: String,
from: String,
to: String): String
}
Feature

50. Interface
trait Translator {
def translate(
source: String,
from: String,
to: String): String
}
Language
EN JA

51. Interface
trait Translator {
def translate(
source: String,
from: Language,
to: Language): String
}
Audio
String
Document
WebPage
Discord
?

52. trait Translator[A] {
def translate(
source: A,
from: Language,
to: Language): A
}
パラメトリックなインターフェイス
Parametric Interface

53. Use Discord auto-translation bot
class DiscordTranslator extends Translator[DiscordMessage] {
override def translate(text: DiscordMessage,
from: Language, to: Language): DiscordMessage = ???
}
def program(t: DiscordTranslator) =
for {
message <- consumeMessages
(from, to) = detectLanguage(message)
newMsg = t.translate(message,from,to)
_ <- send(newMsg)
} yield ()
Discord の自動翻訳ボットの例

54. Translate a simple text
class StringTranslator extends Translator[String] {
override def translate(text: String,
from: Language, to: Language): String = ???
}
def program(t: StringTranslator) =
for {
message <- consumeMessages
(from, to) = detectLanguage(message)
newMsg = t.translate(message,from,to)
_ <- send(newMsg)
} yield ()
簡単なテキストの翻訳

55. What about Documents?
class DocumentTranslator extends Translator[Document] {
override def translate(text: Document,
from: Language, to: Language): Document = ???
}
def program(t: DocumentTranslator) =
for {
message <- consumeMessages
(from, to) = detectLanguage(message)
newMsg = t.translate(message,from,to)
_ <- send(newMsg)
} yield ()
Document ?
文書ファイルはどうしよう?

56. What about Audios?
class AudioTranslator extends Translator[Audio] {
override def translate(text: Audio,
from: Language, to: Language): Audio = ???
}
def program(t: AudioTranslator) =
for {
message <- consumeMessages
(from, to) = detectLanguage(message)
newMsg = t.translate(message,from,to)
_ <- send(newMsg)
} yield ()
Audio ?
音声はどうしよう?

57. Polymorphic program
def program[A](t: Translator[A]) =
for {
message <- consumeMessages
(from, to) = detectLanguage(message)
newMsg = t.translate(message,from,to)
_ <- send(newMsg)
} yield ()
多相的プログラム

58. Parametric Interface
trait Translator[A] {
def translate(
text: A,
from: Language,
to: Language): A
}
パラメトリックなインターフェイス

59. Type class
trait Translator[A] {
def translate(
text: A,
from: Language,
to: Language): A
}
object Translator {
def apply[A](implicit t: Translator[A]): Translator[A] = t
}
型クラス

60. Polymorphic program using context bounds
def program[A: Translator] =
for {
message <- consumeMessages
(from, to) = detectLanguage(message)
newMsg = Translator[A].translate(message,from,to)
_ <- send(newMsg)
} yield ()
型クラス

61. Polymorphic program using context bounds
def program[A: Translator] =
for {
message <- consumeMessages
(from, to) = detectLanguage(message)
newMsg = Translator[A].translate(message,from,to)
_ <- send(newMsg)
} yield ()
def program[A](implicit t: Translator[A])
構文を用いた多相プログラム

62. Type class instances
implicit val discordTranslator = new Translator[DiscordMessage] { … }
implicit val stringTranslator = new Translator[String] { … }
implicit val documentTranslator = new Translator[Document] { … }
implicit val audioTranslator = new Translator[Audio] { … }
program[DiscordMessage]
program[String]
program[Document]
program[Audio]
型クラスのインスタンス

63. Algebraic Data Structures
empty
combine
Monoid
combine
Semigroup
map
flatMap
Monad
map
Functor
代数的構造

64. Semigroup
trait Semigroup[A] {
def combine(a: A, b: A): A
}
object Semigroup {
def apply[A](implicit s: Semigroup[A]): Semigroup[A] = s
}
Semigroup (半群)

65. Semigroup
trait Semigroup[A] {
def combine(a: A, b: A): A
}
object Semigroup {
def apply[A](implicit s: Semigroup[A]): Semigroup[A] = s
}
Associativity
combine(combine(a, b), c)
==
combine(a, combine(b, c))
Laws
Semigroup則は結合律

66. Semigroup
trait Semigroup[A] {
def combine(a: A, b: A): A
}
object Semigroup {
def apply[A](implicit s: Semigroup[A]): Semigroup[A] = s
}
Associativity
(1 + 2) + 3
==
1 + (2 + 3)
Int (+)
Laws

67. Semigroup
trait Semigroup[A] {
def combine(a: A, b: A): A
}
object Semigroup {
def apply[A](implicit s: Semigroup[A]): Semigroup[A] = s
}
Associativity
(1 * 2) * 3
==
1 * (2 * 3)
Int (*)
Laws

68. Semigroup
trait Semigroup[A] {
def combine(a: A, b: A): A
}
object Semigroup {
def apply[A](implicit s: Semigroup[A]): Semigroup[A] = s
}
Associativity
Int (+, *)
String (concat)
Boolean(||, &&)
Laws

69. Semigroup
implicit val intSemigroup = new Semigroup[Int] { … }
implicit val stringSemigroup = new Semigroup[String] { … }
implicit val teamSemigroup = new Semigroup[Person] { … }
...
Associativity
Int (+, *)
String (concat)
Boolean(||, &&)
Laws
🐣
🐣
🐼
🐻🧍 🧍♀
🧍 🏻
♀

70. Monoid
trait Monoid[A] {
def empty: A
def combine(a: A, b: A): A
}
object Monoid {
def apply[A](implicit m: Monoid[A]): Monoid[A] = m
}
モノイド

71. Monoid
trait Monoid[A] {
def empty: A
def combine(a: A, b: A): A
}
object Monoid {
def apply[A](implicit m: Monoid[A]): Monoid[A] = m
}
Identity
combine(a, empty) == a
combine(empty, a ) == a
Associativity
combine(combine(a, b), c)
==
combine(a, combine(b, c)
Laws
モノイド則は単位元と結合律

72. Monoid
trait Monoid[A] {
def empty: A
def combine(a: A, b: A): A
}
object Monoid {
def apply[A](implicit m: Monoid[A]): Monoid[A] = m
}
Eg.
empty combinetype
Int 1 *
Int 0 +
String “” concat
Boolean true &&
Boolean false ||

73. Functor
trait Functor[F[_]] {
def map[A, B](f: A => B): F[A] => F[B]
}
object Functor {
def apply[F[_]](implicit f : Functor[F]) : Functor[F] = f
}
Higher kinded type
ファンクター

74. Functor
trait Functor[F[_]] {
def map[A, B](f: A => B): F[A] => F[B]
}
object Functor {
def apply[F[_]](implicit f : Functor[F]) : Functor[F] = f
}
Identity ?
composition ?
Laws
単位元? 合成律?

75. The identity function
def identity[A](a: A): A = a Identity
composition ?
Laws
恒等関数

76. The identity law
def identity[A](a: A): A = a
//REMINDER: def map[A, B](f: A => B): F[A] => F[B]
map(identity) == identity
Identity
composition ?
Laws
単位元

77. trait Function1[-B, +C] {
def apply(b: B): C
def compose[A](g: A => B): A => C = a => apply(g(a))
}
Identity
composition
Laws
The compose function
合成関数

78. trait Function1[-B, +C] {
def apply(b: B): C
def compose[A](g: A => B): A => C = a => apply(g(a))
}
f: A => B
g: B => C
h: A => C
h = a => g(f(a))
===
g compose f
f ° g
The compose function

79. trait Function1[-B, +C] {
def apply(b: B): C
def compose[A](g: A => B): A => C = a => apply(g(a))
}
//REMINDER: def map[A, B](f: A => B): F[A] => F[B]
map(g compose f) == map(g) compose map(f)
Laws
The composition law
Identity
composition

80. implicit val optionFunctor: Functor[Option] = new Functor[Option]
{
override def map[A, B](f: A => B): Option[A] => Option[B] = {
case None => None
case Some(a) => Some(f(a))
}
}
Example for a Functor instance
Functor[Option]
ファンクター・インスタンスの例

81. implicit val optionFunctor: Functor[Option] = new Functor[Option]
{
override def map[A, B](f: A => B): Option[A] => Option[B] = {
case None => None
case Some(a) => Some(f(a))
}
}
Identity law for:
Functor[Option]
f: identity
f: A => A
None => None
Some(a) => Some(a)
identity
Functor[Option] の単位元

82. implicit val optionFunctor: Functor[Option] = new Functor[Option]
{
override def map[A, B](f: A => B): Option[A] => Option[B] = {
case None => None
case Some(a) => Some(f(a))
}
}
Composition law for:
Functor[Option]
Identity Law
map(identity)(Some(1)) == Some(1)
map(identity)(None) == None
def increment(i: Int): Int
def toString(i: Int): String
map(toString compose increment)
==
map(toString) compose map(increment)
Some(41)
Functor[Option] の合成律

83. implicit val optionFunctor: Functor[Option] = new Functor[Option]
{
override def map[A, B](f: A => B): Option[A] => Option[B] = {
case None => None
case Some(a) => Some(f(a))
}
}
Composition law for:
Functor[Option]
Identity Law
map(identity)(Some(1)) == Some(1)
map(identity)(None) == None
def increment(i: Int): Int
def toString(i: Int): String
map(toString compose increment)
==
map(toString) compose map(increment)
Some(41)
Some(“42”)

84. implicit val optionFunctor: Functor[Option] = new Functor[Option]
{
override def map[A, B](f: A => B): Option[A] => Option[B] = {
case None => None
case Some(a) => Some(f(a))
}
}
Composition law for:
Functor[Option]
Identity Law
map(identity)(Some(1)) == Some(1)
map(identity)(None) == None
def increment(i: Int): Int
def toString(i: Int): String
map(toString compose increment)
==
map(toString) compose map(increment)
Some(41)
Some(“42”)
Some(“42”)

85. Monad
trait Monad[F[_]] {
def map[A, B](f: A => B): F[A] => F[B]
def flatMap[A, B](f: A => F[B]): F[A] => F[B]
}
object Monad {
def apply[F[_]](implicit m: Monad[F]) : Monad[F] = m
}
Higher kinded type
モナド

86. Monad
trait Monad[F[_]] {
def point[A](a: A): F[A]
def map[A, B](f: A => B): F[A] => F[B]
def flatMap[A, B](f: A => F[B]): F[A] => F[B]
}
object Monad {
def apply[F[_]](implicit m: Monad[F]) : Monad[F] = m
}
モナド

87. Monad
trait Monad[F[_]] {
def point[A](a: A): F[A]
def map[A, B](f: A => B): F[A] => F[B]
def flatMap[A, B](f: A => F[B]): F[A] => F[B]
}
object Monad {
def apply[F[_]](implicit m: Monad[F]) : Monad[F] = m
}
Left identity ?
Right identity
Associativity ?
Laws
左単位元 、右単位元 、結合律

88. Left identity
def increment(a: Int): Option[Int] = Some(a + 1)
Some(42).flatMap(a => increment(2))
==
increment(2)
Left identity
Right identity
Associativity ?
Laws
左単位元

89. Right identity
Some(42).flatMap(Some(_)) == Some(42)
Left identity
Right identity
Associativity ?
Laws
右単位元

90. Associativity
def increment(i: Int): Option[Int] = Some(i + 1)
def decrement(i: Int): Option[Int] = Some(i - 1)
Some(42).flatMap(increment).flatMap(decrement)
==
Some(42).flatMap(v => increment(v).flatMap(decrement))
Left identity
Right identity
Associativity
Laws
結合律

91. Monads in real life
IOモナドのコンテキストで演算する Haskeller

92. You can checkout the Scala functional programming
libraries:
● Cats
Github: https://github.com/typelevel/cats
doc: https://typelevel.org/cats
● Scalaz: https://github.com/scalaz/scalaz
● Zio Prelude: https://github.com/zio/zio-prelude
Talks by:
● John De Goes and Adam Fraser: “Reimagining
Functional Type Classes”
● Jakub Kozlowski: “Fantastic Monads and where to
find them”
Resources
次へのリソース

93. FP is
AWESOME
FP って最高

94. CREDITS: This presentation template was created by Slidesgo, including icons by
Flaticon, and infographics & images by Freepik
THANKS!
Please keep this slide for attribution
@WiemZin
ご清聴ありがとうございました