scala213-functional-programming-functor-monoid-monad-workshop

references

preface

goals of this workshop
- introduction to scala features:
  - for-comprehension
  - higher kinded types
- developing basic intuitions concerning standard functional structures
  - monoid
  - functor and applicative functor
  - monad
- show fundamental differences between above constructs
- apply knowledge to real life
  - functional approach to validation
  - functional approach to IO
workshops order
1. ForComprehensionWorkshop
2. MonoidWorkshop
3. FunctorWorkshop
4. ApplicativeWorkshop
5. PersonValidatorWorkshop
6. MonadWorkshop
7. EchoWorkshop

introduction

whenever we create an abstraction like Functor we should
- consider abstract methods it should have
- and laws we expect to hold for the implementations
  - of course Scala won’t enforce any of these laws
laws are important for two reasons
1. help an interface form a new semantic level
  - algebra may be reasoned about independently of the instances
  - example
    - we could proof that Monoid[(A,B)] constructed from Monoid[A] and Monoid[B] is actually a monoid
      - we don’t need to know anything about A and B to conclude this
2. we often rely on laws when writing various combinators
  - example
    - unzip List[(A, B)] into List[A], List[B]
      - same length
      - corresponding elements in the same order
    - algebraic reasoning can potentially save us a lot of work, since we don’t have to write separate tests for these properties
      - we could write a generic unzip function that works for any functor

monoids

monoid consists of the following:

trait

trait Monoid[A] {
  def combine(a1: A, a2: A): A

  def zero: A
}

laws (semigroup with an identity element)

associativity

combine(combine(x,y), z) == combine(x, combine(y,z)) for any choice of x: A, y: A, z: A

identity

exists zero: A, that combine(x, zero) == x and combine(zero, x) == x for any x: A

is a type A and an implementation of Monoid[A] that satisfies the laws

example
- string monoid
```
val stringMonoid = new Monoid[String] {
    def combine(a1: String, a2: String) = a1 + a2
    val zero = ""
}
```
- various Monoid instances don’t have much to do with each other
  - monoid is a type, together with the monoid operations and a set of laws
  - you may build some intuition by considering the various concrete instances
    - but nothing guarantees all monoids you encounter will match your intuition
- we can say that
  - type A forms a monoid
  - type A is monoidal
  - less precisely: type A is a monoid or even type A is monoidal
  - but not: type A has monoid
    - analogy: the page you’re reading forms a rectangle, is rectangular or it is a rectangle but not: it has a rectangle
  - in any case, the Monoid[A] instance is simply evidence of this fact
folding context
```
def foldRight[B](z: B)(f: (A, B) => B): B
```
and we want to fold into the same type so
```
def foldRight[A](z: A)(f: (A, A) => A): A
```
monoid fit these argument types like a glove
```
foldRight(monoid.zero)(monoid.combine)
```
foldLeft and foldRight gives the same results when folding with monoid (associativity)

parallelism

below three operations give the same result

combine(a, combine(b, combine(c, d)))
combine(combine(combine(a, b), c), d)
combine(combine(a, b), combine(c, d)) // allows for parallelism

proof (associativity)

combine(combine(combine(a, b), c), d) = combine(combine(x, y), z), where x = combine(a, b), y = c, z = d
combine(combine(x, y), z) = combine(x, combine(y, z))
combine(x, combine(y, z)) = combine(combine(a, b), combine(c, z))
so: combine(combine(combine(a, b), c), d) = combine(combine(a, b), combine(c, d))

additional properties
- monoids compose
  - example: A, B monoids -> (A, B) is also a monoid
- monoid homomorphism
  - when `A, B monoids => A.combine(f(x), f(y)) == f(B.combine(x, y))”
- monoid isomorphism
  - homomorphism in both directions
  - example: String and List[Char] monoids with concatenation
- commutative combine
  - when combine(x, y) == combine(y, x))
semigroup
- semigroup for a type A has an associative combine operation that returns an A given two input A values
```
trait Semigroup[A] {
  def combine(a1: A, a2: A): A
}
```
- example: NonEmptyList[A]
  - there is no instance Monoid[NonEmptyList[A]] - you cannot specify zero element

functors

definition
- we say that a type constructor F is a functor, and the Functor[F] instance constitutes proof that F is in fact a functor
```
trait Functor[F[_]] {
  def map[A, B](fa: F[A])(f: A => B): F[B]
}
```
must obey two laws:
- fa.map(f).map(g) = fa.map(f.andThen(g))
- fa.map(x => x) = fa
functors compose: https://github.com/mtumilowicz/scala212-cats-category-theory-composing-functors
for more formal reasoning, please refer: https://github.com/mtumilowicz/java11-category-theory-optional-is-not-functor

applicative functors

definition
```
trait Applicative[F[_]] {

  def unit[A](a: => A): F[A]

  def map2[A, B, C](fa: F[A], fb: F[B])(f: (A, B) => C): F[C]
}
```
- could be formulated using unit and apply (therefore called Applicative), rather than unit and map2
  - def apply[A,B](fab: F[A => B])(fa: F[A]): F[B]
applicative laws
- left and right identity
```
map(v)(id) == v
map(map(v)(g))(f) == map(v)(f compose g)
```
  in other words
```
map2(unit(()), fa)((_,a) => a) == fa
map2(fa, unit(()))((a,_) => a) == fa     
```
- associativity (in terms of product)
  - def product[A,B](fa: F[A], fb: F[B]): F[(A,B)] = map2(fa, fb)((_,_))
  - def assoc[A,B,C](p: (A,(B,C))): ((A,B), C) = p match { case (a, (b, c)) => ((a,b), c) }
  - product(product(fa,fb),fc) == map(product(fa, product(fb,fc)))(assoc)
- naturality law
  - fa.map2(fb)((a, b) => (f(a), g(b))) == fa.map(f).product(fb.map(g))
all applicatives are functors
advantages
- preferable to implement combinators using as few assumptions as possible
  - it’s better to assume that a data type can provide map2 than flatMap
  - otherwise we’d have to write a new traverse every time we encountered a type that’s Applicative but not a Monad
validation context
- only reporting the first error means the user would have to repeatedly submit the form and fix one error at a time
- consider what happens in a sequence of flatMap calls like the following
```
validName(field1) flatMap (f1 =>
    validBirthdate(field2) flatMap (f2 =>
        validPhone(field3) map (f3 => ValidInput(f1, f2, f3))
```
  - if validName fails with an error, then validBirthdate and validPhone won’t even run
- now think of doing the same thing with map3
```
map3(validName(field1), validBirthdate(field2), validPhone(field3))(ValidInput(_,_,_))
```
  - no dependency implied between the three expressions
additional properties
- large number of the useful combinators on Monad can be defined using only unit and map2
  - example
```
def map2[B, C](fb: F[B])(f: (A, B) => C): F[C] =
    fa.flatMap(a => fb.map(b => f(a, b)))
```
- applicative functors compose
  - F[_] and G[_] are applicative functors => F[G[_]] is applicative functor

monads

definition

trait Monad[F[_]] {

  def unit[A](a: => A): F[A]

  def flatMap[A, B](ma: F[A])(f: A => F[B]): F[B]

}

monad laws

given compose function

def compose[A,B,C](f: A => F[B], g: B => F[C]): A => F[C] =
    a => flatMap(f(a))(g)

digression
- functions like that: A => F[B] are called Kleisli arrows

left identity and right identity

compose(f, unit) == f // m.flatMap(unit) == m
compose(unit, f) == f // unit(a).flatMap(func) == func(a)

// example for option
option.flatMap(Some(_)) == option

Some(value).flatMap(f) == f(value)

similar to zero element in monoids

associative law for monads

compose(compose(f, g), h) == compose(f, compose(g, h))

// example for option
option.flatMap(f).flatMap(g) == option.flatMap(f(_).flatMap(g))

similar to associative law for monoids

monad is an implementation of one of the minimal sets of primitive combinators satisfying the monad laws

combinator sets

unit and flatMap

def unit[A](a: => A): F[A]
def flatMap[A, B](ma: F[A])(f: A => F[B]): F[B]

unit and compose

def unit[A](a: => A): F[A]
def compose[A,B,C](f: A => F[B], g: B => F[C]): A => F[C]

unit, map and join

def unit[A](a: => A): F[A]
def map[A, B](fa: F[A])(f: A => B): F[B]
def join[A](mma: F[F[A]]): F[A]

each monad brings its own set of additional primitive operations that are specific to it
- example: Option, Either, List
vs interfaces
- interfaces ~ relatively complete API for an abstract data type
  - merely abstracting over the specific representation
  - example: List - LinkedList, ArrayList
- monad ~ many vastly different data types can satisfy the Monad interface and laws
  - like Monoid, is an abstract, purely algebraic interface
chain of flatMap calls is like an imperative program with statements that assign to variables
- monad specifies what occurs at statement boundaries
- example
  - Option monad - may return None at some point and effectively terminate the processing
- example: identity monad
```
case class Id[A](value: A) {
  def map[B](f: A => B): Id[B] = Id(f(value))

  def flatMap[B](f: A => Id[B]): Id[B] = f(value)
}
```
```
val id = Id("Hello, ")
  .flatMap(a => Id("monad!")
  .flatMap(b => Id(a + b)))

// id = Id(Hello, monad!)
```
  - simply variable substitution
```
for {
    a <- Id("Hello, ")
    b <- Id("monad!")
} yield a + b
```
    vs
```
val a = "Hello, "
val b = "monad!"
val c = a + b
```
    - variables a and b get bound to "Hello, " and "monad!”
  - monads provide a context for introducing and binding variables, and performing variable substitution
unlike Functors and Applicatives, not all Monads compose
- to show that "monads do not compose", it is sufficient to find a counterexample, namely two monads f and g such that f g is not a monad
- proof: https://carlo-hamalainen.net/2014/01/02/applicatives-compose-monads-do-not/
- problem
  - to implement join for nested monads F and G, write something of a type like:
```
F[G[F[G[A]]]] => F[G[A]]
```
  - that can’t be written generally
  - but if G also happens to have a Traverse instance
```
F[F[G[G[A]]]] => F[G[A]] // use sequence to turn G[F[_]] into F[G[_]], then use join on F then on G
```
- problem is often addressed with a custom-written version specifically constructed for composition
  - example: monad transformers like OptionT
all monads are applicatives
- not all applicatives are monads
- difference between monads and applicatives
  - applicative constructs context-free computations, while Monad allows for context sensitivity
  - join and flatMap can’t be implemented with just map2 and unit
    - def join[A](f: F[F[A]]): F[A] // removes a layer of F
    - unit function only lets us add an F layer
    - map2 lets us apply a function within F but does no flattening of layers
- Option applicative versus the Option monad
  - combine the results from two (independent) lookups: map2
```
val F: Applicative[Option] = ...
val departments: Map[String,String] = ...
val salaries: Map[String,Double] = ...
val o: Option[String] = F.map2(departments.get("Alice"), salaries.get("Alice")) {
    (dept, salary) => s"Alice in $dept makes $salary per year"
}
```
  - result of one lookup to affect next lookup: flatMap
```
val idsByName: Map[String,Int]
val departments: Map[Int,String] = ...

val o: Option[String] = idsByName.get("Bob")
    .flatMap { id => departments.get(id) }
}
```
each monad brings its own set of additional primitive operations that are specific to it
- example: State[S, A]
```
def get[S]: State[S, S]
def set[S](s: => S): State[S, Unit]
```
monad is a monoid in a category of endofunctors
- endofunctor - functor that maps a category to itself
- let's look for minimal set of combinators for a monad: unit, map and join
  - monoid = unit + join
  - endofunctor = map
- Monoid[M] - objects: Scala types, arrows: Scala functions
- Monad[F] - objects: Scala functors, arrows: natural transformations
- overview
  
  zero/unit op/join
  
  Monoid[M] 1 => M M² => M
  
  Monad[F] 1 ~> F F² ~> F
  - ~> is natural transformation
  - M² is (M, M)
  - F² is F[F[A]]
  - 1 is
    - for Monoid: Unit type
    - for Monad: identity functor

IO context

sealed trait IO[A] {
  self => //  lets us refer to this object inside closures
  def unsafeRunSync: A

  def map[B](f: A => B): IO[B] =
    new IO[B] {
      def run: B = f(self.run)
    }

  def flatMap[B](f: A => IO[B]): IO[B] =
    new IO[B] {
      def run: B = f(self.run).run
    }
}

in short: IO is a magic thing that says this function depends on something other than its arguments
clearly separates pure code from impure code, forcing us to be honest about where interactions with the outside world are occurring
- referentially transparent description of a computation with effects

IO computations are ordinary values

we can store them in lists, pass them to functions, create them dynamically, and so on

example

def PrintLine(msg: String): IO = new IO {
  def unsafeRun = println(msg)
}

def contest(p1: Player, p2: Player): IO = // pure function
  PrintLine(winnerMsg(winner(p1, p2)))

given IO[A] will overflow the runtime call stack and throw a StackOverflowError
- solution: encapsulate operations into dedicated case classes and use tail recursion for evaluation
  - io.flatMap(f).flatMap(g) = io.flatMap(v => f(v) flatMap g)
  - approach called trampoline

appendix

higher kinded type

represent an ability to abstract over type constructors
suppose we have same implementations, different type constructors
```
def tuple[A, B](as: List[A], bs: List[B]): List[(A, B)] =
  as.flatMap{a =>
    bs.map((a, _))}
    
def tuple[A, B](as: Option[A], bs: Option[B]): Option[(A, B)] =
  as.flatMap{a =>
    bs.map((a, _))}
    
def tuple[E, A, B](as: Either[E, A], bs: Either[E, B]): Either[E, (A, B)] =
  as.flatMap{a =>
    bs.map((a, _))}
```
- in programming, when we encounter such great sameness—not merely similar code, but identical code—we would like the opportunity to parameterize: extract the parts that are different to arguments, and recycle the common code for all situations
- we have a way to pass in implementations; that’s just higher-order functions
- we need "type constructor as argument"
```
def tuplef[F[_], A, B](fa: F[A], fb: F[B]): F[(A, B)] = ???      
```
  - F[_] means that F may not be a simple type, like Int or String, but instead a one-argument type constructor, like List or Option
higher-kinded types are sometimes used in libraries
- example: standard Scala collections
Scala doesn’t allow us to use underscore syntax to simply say State[Int, _] to create an anonymous type constructor like we create anonymous functions
- we have to go through type projections
  - ({type L[a] = Map[K, a]})#L
    - declares an anonymous type
    - then access its L member with the # syntax
    - type constructor declared inline like this is often called a type lambda in Scala
- removed in Scala 3

for comprehension

each line in the expression using the <- is translated to a flatMap call, except
- the last line (yield) - it is translated to a concluding map call
- x <- c if cond is translated to c.filter(x => cond)

flatMap / map

for {
  bound <- list
  out <- f(bound)
} yield out

// is equivalent to (could be desugared with IntelliJ)

list.flatMap { bound =>
  f(bound).map { out =>
    out
  }
}

flatMap / map / filter

for {
  sl <- l
  el <- sl if el > 0
} yield el.toString.length

// is equivalent to

l.flatMap(sl => sl.filter(el => el > 0).map(el => el.toString.length))

Name		Name	Last commit message	Last commit date
Latest commit History 92 Commits
project		project
src		src
.gitignore		.gitignore
.travis.yml		.travis.yml
LICENSE		LICENSE
README.md		README.md
build.sbt		build.sbt

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

scala213-functional-programming-functor-monoid-monad-workshop

preface

introduction

monoids

functors

applicative functors

monads

IO context

appendix

higher kinded type

for comprehension

About

Releases

Packages

Languages

	`zero`/`unit`	`op`/`join`
`Monoid[M]`	`1 => M`	`M² => M`
`Monad[F]`	`1 ~> F`	`F² ~> F`

License

mtumilowicz/scala213-functional-programming-functor-monoid-monad-workshop

Folders and files

Latest commit

History

Repository files navigation

scala213-functional-programming-functor-monoid-monad-workshop

preface

introduction

monoids

functors

applicative functors

monads

IO context

appendix

higher kinded type

for comprehension

About

Topics

Resources

License

Stars

Watchers

Forks

Releases

Packages 0

Languages

Packages