added `TFunctor` #1815

kailuowang · 2017-08-10T21:41:05Z

added instances for

WriterT
Kleisli
OptionT
EitherT
IdT
EitherK
Tuple2K
Free
FreeT
Func
Nested
OneAnd

Fixes #1812, #1713 and #1492. also sort of a continuation of #1620
Any instances I miss?
~~Question, do we need an instance for Free, not sure if there is a use case in the wild.~~

Update: Added instance for Free, EitherK, TupleK, Func, Nested and OneAnd

iravid · 2017-08-11T06:41:37Z

Not sure about the laws but this could work for StateT as well, no? Get Outlook for iOS<https://aka.ms/o0ukef>

…

________________________________ From: Kai(luo) Wang <notifications@github.com> Sent: Friday, August 11, 2017 12:41:08 AM To: typelevel/cats Cc: Subscribed Subject: [typelevel/cats] added `TFunctor` (#1815) added instances for WriterT, Kleisli, OptionT, EitherT and FreeT Fixes #1812<#1812>

________________________________ You can view, comment on, or merge this pull request online at: #1815 Commit Summary * added TFunctor File Changes * A core/src/main/scala/cats/TFunctor.scala<https://github.com/typelevel/cats/pull/1815/files#diff-0> (31) * M core/src/main/scala/cats/data/EitherT.scala<https://github.com/typelevel/cats/pull/1815/files#diff-1> (4) * M core/src/main/scala/cats/data/IdT.scala<https://github.com/typelevel/cats/pull/1815/files#diff-2> (4) * M core/src/main/scala/cats/data/Kleisli.scala<https://github.com/typelevel/cats/pull/1815/files#diff-3> (5) * M core/src/main/scala/cats/data/OptionT.scala<https://github.com/typelevel/cats/pull/1815/files#diff-4> (5) * M core/src/main/scala/cats/data/WriterT.scala<https://github.com/typelevel/cats/pull/1815/files#diff-5> (5) * M core/src/main/scala/cats/syntax/all.scala<https://github.com/typelevel/cats/pull/1815/files#diff-6> (1) * M core/src/main/scala/cats/syntax/package.scala<https://github.com/typelevel/cats/pull/1815/files#diff-7> (1) * A core/src/main/scala/cats/syntax/tfunctor.scala<https://github.com/typelevel/cats/pull/1815/files#diff-8> (4) * M free/src/main/scala/cats/free/FreeT.scala<https://github.com/typelevel/cats/pull/1815/files#diff-9> (6) * M free/src/test/scala/cats/free/FreeTTests.scala<https://github.com/typelevel/cats/pull/1815/files#diff-10> (12) * A laws/src/main/scala/cats/laws/TFunctorLaws.scala<https://github.com/typelevel/cats/pull/1815/files#diff-11> (24) * M laws/src/main/scala/cats/laws/discipline/Arbitrary.scala<https://github.com/typelevel/cats/pull/1815/files#diff-12> (11) * A laws/src/main/scala/cats/laws/discipline/TFunctorTests.scala<https://github.com/typelevel/cats/pull/1815/files#diff-13> (37) * M tests/src/test/scala/cats/tests/EitherTests.scala<https://github.com/typelevel/cats/pull/1815/files#diff-14> (4) * M tests/src/test/scala/cats/tests/IdTTests.scala<https://github.com/typelevel/cats/pull/1815/files#diff-15> (3) * M tests/src/test/scala/cats/tests/KleisliTests.scala<https://github.com/typelevel/cats/pull/1815/files#diff-16> (5) * M tests/src/test/scala/cats/tests/OptionTTests.scala<https://github.com/typelevel/cats/pull/1815/files#diff-17> (5) * M tests/src/test/scala/cats/tests/WriterTTests.scala<https://github.com/typelevel/cats/pull/1815/files#diff-18> (3) Patch Links: * https://github.com/typelevel/cats/pull/1815.patch * https://github.com/typelevel/cats/pull/1815.diff — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#1815>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AARSxEqQZHPYP58wp27n6ak5a0UGJBlcks5sW3jzgaJpZM4O0A8c>.

kailuowang · 2017-08-11T13:36:50Z

@iravid StateT's inner data has a nested F[_], i.e. F[S => F[(S, A)]], I don't know how to transform that to a G[S => G[(S, A)]]using just F~>G, any ideas?

added instances for WriterT, Kleisli, OptionT, EitherT and FreeT

iravid · 2017-08-11T14:19:05Z

Ah, you're right - I missed the part where StateT.transformF would need to impose more constraints on F. Oh well :-)

codecov-io · 2017-08-11T14:23:56Z

Codecov Report

Merging #1815 into master will increase coverage by 0.05%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #1815      +/-   ##
==========================================
+ Coverage   94.88%   94.94%   +0.05%     
==========================================
  Files         241      244       +3     
  Lines        4148     4193      +45     
  Branches      102      104       +2     
==========================================
+ Hits         3936     3981      +45     
  Misses        212      212

Impacted Files	Coverage Δ
...rc/main/scala/cats/laws/discipline/Arbitrary.scala	`92.85% <100%> (+0.66%)`	⬆️
core/src/main/scala/cats/data/Kleisli.scala	`100% <100%> (ø)`	⬆️
free/src/main/scala/cats/free/FreeT.scala	`91.89% <100%> (+0.22%)`	⬆️
core/src/main/scala/cats/data/EitherT.scala	`98.3% <100%> (+0.02%)`	⬆️
core/src/main/scala/cats/data/OneAnd.scala	`98.41% <100%> (+0.05%)`	⬆️
core/src/main/scala/cats/TFunctor.scala	`100% <100%> (ø)`
laws/src/main/scala/cats/laws/TFunctorLaws.scala	`100% <100%> (ø)`
...ain/scala/cats/laws/discipline/TFunctorTests.scala	`100% <100%> (ø)`
core/src/main/scala/cats/data/IdT.scala	`97.29% <100%> (+0.15%)`	⬆️
core/src/main/scala/cats/data/Tuple2K.scala	`100% <100%> (ø)`	⬆️
... and 9 more

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 22e7175...3f0f79d. Read the comment docs.

peterneyens · 2017-08-12T10:11:14Z

core/src/main/scala/cats/data/Func.scala

@@ -31,6 +31,11 @@ object Func extends FuncInstances {
 }

 private[data] abstract class FuncInstances extends FuncInstances0 {
+
+  implicit def catsDataTFunctorFoFunc[A]: TFunctor[Func[?[_], A, ?]] = new TFunctor[Func[?[_], A, ?]] {


Missing "r" in "foFunc", similarly in "foNested".

edmundnoble · 2017-08-12T21:02:38Z

I don't think we need to provide this type class to have a uniform API. We can just have a uniform API. Type classes are abstractions; unless there's a useful way to abstract over this, I think we should not include it. This is the standard for all other type classes in cats if I'm not mistaken.

kailuowang · 2017-08-13T18:30:31Z

@edmundnoble
The law tests provided by the type class is more thorough than the existing test coverage (e.g. FreeT and Kleisili) - also plus many of the instances I added doesn't have a mapNT equivalent yet, not to mention the tests, so I would need to duplicate the same tests code for 12 instances.

I think the type class might be able to provide reusable code other than liftNT . I am thinking compose. I will try implement that and maybe try find more after vacation.

johnynek · 2017-08-20T23:53:09Z

core/src/main/scala/cats/TFunctor.scala

+ * as `OptionT`, `EitherT`
+ *
+ */
+@typeclass trait TFunctor[T[_[_], _]] {


do we need T[_[_], _] or would T[_[_]] do? I'm not sure what the second parameter is buying us. Maybe I'm missing it. Can't we just ignore the inner value type? Also, wouldn't we have more flexibility only acting on T[_[_]]?

Agreed here. If we had T[_[_]] I could come up with a lot more usecases. Maybe call it FunctorK then. We'd have to wrap the transformers in Forall though, which may mean our unified transform syntax fails to be found implicitly.

The main reason I chose T[_[_], _] here, is that given any Functor F[_], T[F, _] is also a Functor. So T[_[_], _] can be deemed as a functor between the F[_] Functor and the T[F, _] Functor. i.e. a endofunctor in the category of endofunctors, thus more like a higher kinded Functor.
If you use T[_[_]], then given any Functor F[_], T[F] isn't a Functor, it's a kind * type, not a type constructor. So T[_[_]] is a functor between F[_] Functor and kind * type T[F]s. That makes it less a high kinded Functor. However, you can work around with the parametricity in instance definitions, e.g. for OptionT,

def functorKInstance[A]: FunctorK[OptionT[?[_], A]]

With parametric A, we can have a FunctorK instance forall As. In some sense, we may choose to see this set of FunctorK instances as a single truely higher kinded functor.

Practical use wise, I can't think of any downside of using T[_[_]]. The only downside is its awkward encoding of the higher kinded Functor - it relies on instance definition. It may have more usecases, one that immediately jumps out is FunctionK. So maybe practically it's more useful.

Now, the fact that we aren't sure about this thing makes me think that maybe we shouldn't place this in cats-core. I already have FunctorK[T[_]] defined in mainecoon, I could add all these instances there. Or we can create a new cats-extra module for this type of less commonly used type class.

With the example of FunctionK, my first thought is that that’s what’s defined incorrectly. We really want FunctionK[F[_], G[_], A] and Forall[FunctionK[F, G, ?]] for a natural transformation. Then you do have a FunctorK[FunctionK[F, ?[_], ?]]. But that might just be crazy talk.

I'm not sure if "higher-kinded X" is a really rigid term. In my mind, higher-kinded functor could equally well mean a functor from (endofunctors in Scal) to Scal, or a functor from endofunctors in Scal to other endofunctors in Scal. This is my main problem with the -K naming scheme; it gets even worse when type classes have multiple type parameters.

As well, the encoding proposed above does not reify that the A can vary with the instance remaining valid. The consumer of the instance needs a Forall[Lambda[A =>FunctorK[OptionT[?[_], A]]]] to have that fact in hand, and that makes it much more unwieldy, though a type alias could help.

To my mind if we're going to do a cats-extras project we'll want this there.

I agree with @kailuowang and @edmundnoble, I don't think it's all that useful in cats-core. I'd like to see conformity in Monad Transformers for defining a mapK method, but it doesn't need a type class in core IMO.

I'm not sure if "higher-kinded X" is a really rigid term.

Absolutely. I don’t think it helps at all. But if Functor is really “endofunctor in the category of Scal”, I’d expect FunctorK to mean the same thing as K in other instances, which basically adds “endofunctors” to the description – i.e., “endofunctor in the category of endofunctors in Scal”.

I’d also be happy with Functor being renamed to Endofunctor and having EndofunctorK and something like LowerFunctor (for a functor from the category of endofunctors to the category of Skal).

Also agreed that none of this should be in cats-core. I would like to find some time to work with -Ykind-polymorphism and to define these things as specializations of kind-polymorphic endofunctors, etc.

This sort of thing has been started in a few different libraries already – khats (which is too homophonically-named to be useful), Griffins (which never really got off the ground and mostly lives inside a Matryoshka branch), and I think one other that I have forgotten 😕

johnynek · 2017-08-21T06:53:55Z

+1 for FunctorK. Seems more consistent with cats naming styles.

kailuowang · 2018-01-04T20:52:47Z

Closing this for now since 1.0 already released

kailuowang added the in progress label Aug 10, 2017

kailuowang added 2 commits August 11, 2017 09:37

added TFunctor

6e843e3

added instances for WriterT, Kleisli, OptionT, EitherT and FreeT

fix format

c5f2874

kailuowang force-pushed the TFunctor branch from 5ccd905 to c5f2874 Compare August 11, 2017 13:49

added TFunctor instance for Free

bfb0d8a

added doctest example for liftNT

d9d4f5a

kailuowang added this to the 1.0.0-RC1 milestone Aug 11, 2017

kailuowang requested review from edmundnoble and johnynek August 11, 2017 15:28

kailuowang added 2 commits August 11, 2017 11:48

added instances for Tuple2K and EitherK

9621b9b

added instances for Nested, Func and OneAnd

1d8f33a

peterneyens reviewed Aug 12, 2017

View reviewed changes

fix typo

3f0f79d

johnynek reviewed Aug 20, 2017

View reviewed changes

kailuowang removed this from the 1.0.0-RC1 milestone Aug 29, 2017

kailuowang mentioned this pull request Aug 29, 2017

Add MFunctor type class for mapping monad morphisms over monad transformers #1492

Closed

kailuowang mentioned this pull request Oct 20, 2017

A unified mapK #1982

Closed

kailuowang closed this Jan 4, 2018

kailuowang removed the in progress label Jan 4, 2018

kailuowang mentioned this pull request Aug 16, 2018

Provide instances for cats core types typelevel/cats-tagless#9

Closed

kailuowang mentioned this pull request Sep 12, 2018

Add Kleisli#mapK equivalent returning FunctionK #2450

Merged

LukaJCB mentioned this pull request Jan 12, 2019

Add FunctorK trait #2697

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

added `TFunctor` #1815

added `TFunctor` #1815

kailuowang commented Aug 10, 2017 •

edited

Loading

iravid commented Aug 11, 2017 via email

kailuowang commented Aug 11, 2017

iravid commented Aug 11, 2017 •

edited

Loading

codecov-io commented Aug 11, 2017 •

edited

Loading

peterneyens Aug 12, 2017

edmundnoble commented Aug 12, 2017

kailuowang commented Aug 13, 2017

johnynek Aug 20, 2017

edmundnoble Aug 21, 2017

kailuowang Aug 28, 2017

sellout Aug 28, 2017

edmundnoble Aug 29, 2017

LukaJCB Aug 29, 2017

sellout Aug 29, 2017

johnynek commented Aug 21, 2017

kailuowang commented Jan 4, 2018

added TFunctor #1815

added TFunctor #1815

Conversation

kailuowang commented Aug 10, 2017 • edited Loading

iravid commented Aug 11, 2017 via email

kailuowang commented Aug 11, 2017

iravid commented Aug 11, 2017 • edited Loading

codecov-io commented Aug 11, 2017 • edited Loading

Codecov Report

Choose a reason for hiding this comment

edmundnoble commented Aug 12, 2017

kailuowang commented Aug 13, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

johnynek commented Aug 21, 2017

kailuowang commented Jan 4, 2018

added `TFunctor` #1815

added `TFunctor` #1815

kailuowang commented Aug 10, 2017 •

edited

Loading

iravid commented Aug 11, 2017 •

edited

Loading

codecov-io commented Aug 11, 2017 •

edited

Loading