56 views

Skip to first unread message

Apr 19, 2017, 3:16:59 AM4/19/17

to scala-user

In the following code snippet, I define a Function1 Functor:

```
implicit def Function1Functor[R]: Functor[({type l[a]=(R) => a})#l] = new Functor[({type l[a]=(R) => a})#l] {
def fmap[A, B](r: R => A, f: A => B) = r andThen f
}
```

1. But why is that signature ({type l[a]=(R) => a})#l like that? how to interpret that?

2. I could now randomly pick any type and qualify that as a Functor. For example.,

case class SomeRandomType[A] (param: A) // this does not know about the fact that it is going to be a Functor soon

val someRandomTypeFunctor = new Functor[SomeRandomType] {

def map[A, B](fa: SomeRandomType[A])(f: A => B): SomeRandomType[B] =

SomeRandomType(f(fa.param))

}

So this to me is a Typeclass definition. Is that correct?

Apr 19, 2017, 4:59:38 AM4/19/17

to scala-user

W dniu środa, 19 kwietnia 2017 09:16:59 UTC+2 użytkownik Joe San napisał:

In the following code snippet, I define a Function1 Functor:`implicit def Function1Functor[R]: Functor[({type l[a]=(R) => a})#l] = new Functor[({type l[a]=(R) => a})#l] { def fmap[A, B](r: R => A, f: A => B) = r andThen f }`

1. But why is that signature ({type l[a]=(R) => a})#l like that? how to interpret that?

This is called a type lambda. You can think about it as an anonymous function in the domain of types. It is typically used in Scala to partially apply a type constructor, ie. convert a type constructor that takes more parameters (Function1[-A,+R] in your case) to a type constructor that takes less parameters (Function1Functor[A] respectively). Google it up, there are some really good articles about it out there.

2. I could now randomly pick any type and qualify that as a Functor. For example.,

Well, for any type constructor F[_] an instance of Functor[F] can be summoned, using Yoneda lemma. But this is advanced category-theoretical wizardry, on the far boundary of my expertise. There are also materials and talks out there.

Google up "free monads", because deriving Functors for arbitrary F[_] is an essential step in this technique.

case class SomeRandomType[A] (param: A) // this does not know about the fact that it is going to be a Functor soonval someRandomTypeFunctor = new Functor[SomeRandomType] {def map[A, B](fa: SomeRandomType[A])(f: A => B): SomeRandomType[B] =SomeRandomType(f(fa.param))}So this to me is a Typeclass definition. Is that correct?

It appears to be, yes. Unfortunately I don't know if it's consistent, or the correct alignment of types there is a mere coincidence ;)

cheers,

Rafał

Apr 19, 2017, 5:08:17 AM4/19/17

to scala-user

Op woensdag 19 april 2017 09:16:59 UTC+2 schreef Joe San:

In the following code snippet, I define a Function1 Functor:`implicit def Function1Functor[R]: Functor[({type l[a]=(R) => a})#l] = new Functor[({type l[a]=(R) => a})#l] { def fmap[A, B](r: R => A, f: A => B) = r andThen f }`

1. But why is that signature ({type l[a]=(R) => a})#l like that? how to interpret that?

Functor is defined like this (more or less): trait Functor[F[_]]. This means that it accepts a type constructor **of arity 1**.

Function1 is defined like this: trait Function1[-A,+B]. So Function1 is a type constructor **of arity 2**.

If you write a Functor for List (arity 1) you can just say Functor[List] and you're done. But Functor[Function1] is impossible.

So to create a Functor for a Function1 you have to supply 1 type argument to Function1 and leave 1 parameter open, or in other words **partially apply** the Function1 type constructor.

Scala doesn't have special syntax for partially applying type constructors, so you have to do it with ({ type Lambda[x] = Function1[SomeType,x] })#Lambda.

How this works is that you create a structural type that contains a type alias called Lambda which is a type constructor of arity 1. Then you use the # operator on that structural type to extract the Lambda type constructor; this is called type projection.

Apr 19, 2017, 5:13:49 AM4/19/17

to scala-user

There is a compiler plugin (https://github.com/non/kind-projector) which adds special syntax for this purpose.

It enables you to write Functor[R => ?].

Op woensdag 19 april 2017 11:08:17 UTC+2 schreef Jasper-M:

Op woensdag 19 april 2017 11:08:17 UTC+2 schreef Jasper-M:

Apr 28, 2017, 9:22:53 PM4/28/17

to scala...@googlegroups.com

On 4/19/17 3:16 AM, Joe San wrote:

1. But why is that signature ({type l[a]=(R) => a})#l like that? how to interpret that?

Given

type l[a]=(R) => a

then

Or, alternatively, you can think of it as meaning

where

l

Or, alternatively, you can think of it as meaning

l

where

type l[a]=(R) => a

Jasper has described the mechanism by which this is implemented.

2. I could now randomly pick any type and qualify that as a Functor.

Not *any* type, but lots of them, sure. Including the one
below.

For example.,case class SomeRandomType[A] (param: A) // this does not know about the fact that it is going to be a Functor soon

val someRandomTypeFunctor = new Functor[SomeRandomType] {def map[A, B](fa: SomeRandomType[A])(f: A => B): SomeRandomType[B] =SomeRandomType(f(fa.param))}So this to me is a Typeclass definition. Is that correct?

Yep.

--

Stephen Compall

--

Stephen Compall

Reply all

Reply to author

Forward

0 new messages

Search

Clear search

Close search

Google apps

Main menu