Cofree comonad #1446

edmundnoble · 2016-10-28T21:33:43Z

Basic Cofree Comonad functionality, including a couple of instances. I tried to implement a few other things (mapBranching(F ~> G)(Cofree[F, A]): Cofree[G, A], decompile(F ~> W)(W[A]): Cofree[F, A]) but I couldn't make them stack-safe. This implementation is inspired by the one in Scalaz.

codecov-io · 2016-10-28T22:18:13Z

Current coverage is 92.02% (diff: 100%)

Merging #1446 into master will increase coverage by 0.08%

@@             master      #1446   diff @@
==========================================
  Files           244        245     +1   
  Lines          3623       3660    +37   
  Methods        3499       3539    +40   
  Messages          0          0          
  Branches        124        121     -3   
==========================================
+ Hits           3331       3368    +37   
  Misses          292        292          
  Partials          0          0

Powered by Codecov. Last update c9521e1...52067df

adelbertc · 2016-10-28T22:23:22Z

free/src/main/scala/cats/free/Cofree.scala

+  * A tree with data at the branches, as opposed to Free which is a tree with data at the leaves.
+  * Not an instruction set functor made into a program monad as in Free, but an instruction set's outputs as a
+  * functor made into a tree of the possible worlds reachable using the instruction set.
+  */


Was this derived/based on Scalaz's Cofree or from scratch? If former could you add attribution to Scalaz, similar to what we have here ?

Totally derived from the one at Scalaz. I did not know there was an example of such an attribution somewhere in the code to crib from, I'll add it in.

adelbertc · 2016-10-28T22:24:45Z

free/src/test/scala/cats/free/CofreeTests.scala

+
+  checkAll("Cofree[Option, ?]", ComonadTests[Cofree[Option, ?]].comonad[Int, Int, Int])
+  checkAll("Cofree[Option, ?]", FoldableTests[Cofree[Option, ?]].foldable[Int, Int])
+  checkAll("Cofree[Option, ?]", TraverseTests[Cofree[Option, ?]].traverse[Int, Int, Int, Int, Option, Option])


I think Traverse checks Foldable already

Check Serializability of the Traverse instance?

I don't need to check that the Traverse and Foldable instances are valid separately? They are defined separately. I will add that serializable check.

Are they? Where did you see that? Seem to be together https://github.com/typelevel/cats/blob/master/laws/src/main/scala/cats/laws/discipline/TraverseTests.scala#L9

You are absolutely right. Took the foldable tests out.

adelbertc · 2016-10-28T22:25:28Z

free/src/test/scala/cats/free/CofreeTests.scala

+  checkAll("Comonad[Cofree[Option, ?]]", SerializableTests.serializable(Comonad[Cofree[Option, ?]]))
+  checkAll("Cofree[List, ?]", ComonadTests[Cofree[List, ?]].comonad[Int, Int, Int])
+  checkAll("Cofree[List, ?]", FoldableTests[Cofree[List, ?]].foldable[Int, Int])
+  checkAll("Cofree[List, ?]", TraverseTests[Cofree[List, ?]].traverse[Int, Int, Int, Int, Option, Option])


Do we need to test these given the tests above? Seems the only difference is s/Option/List

If we're comfy with not testing these, absolutely. I don't see any precedent in other tests in doing the same law checks on different instances, so I'll go ahead and remove.

edmundnoble · 2016-10-31T06:27:59Z

Is this alright for inclusion or should it perhaps have more operators?

adelbertc · 2016-11-01T02:08:51Z

Could you write some functions for the other functions, maybe like unfold and unfoldStrategy ?

Also looks like the Foldable-specific instance isn't being tested because the test is getting the Foldable through the Traverse test. You can see an example here of how we test these kinds of "if F[_]: TC then MT[F, ?]: TC law checks.

Other than that, LGTM 👍 Thanks @edmundnoble !

xuwei-k · 2016-11-02T22:32:54Z

free/src/main/scala/cats/free/Cofree.scala

+}
+
+sealed abstract class CofreeInstances2 {
+  /** low priority `Foldable1` instance */


This is a Foldable, not Foldable1. There is scalaz.Foldable1. but cats does not have Foldable1.

#145

@xuwei-k We have Foldable1 but we call it Reducible #298. No Traverse1 though like you said below.

xuwei-k · 2016-11-02T22:34:52Z

free/src/main/scala/cats/free/Cofree.scala

+}
+
+sealed abstract class CofreeInstances1 extends CofreeInstances2 {
+  /** low priority `Traverse1` instance */


This is a Traverse, not Traverse1

edmundnoble · 2016-11-04T03:00:17Z

I removed unfoldStrategy because after trying to write a test, I'm not convinced it has any use. If anybody can see one, please suggest it. Otherwise, I believe I have addressed all of your comments.

dwhitney · 2016-12-02T22:09:34Z

Bump. This looks interesting and it looks like @edmundnoble has addressed @adelbertc's concerns. Any reason not to merge?

johnynek · 2016-12-02T23:55:25Z

free/src/main/scala/cats/free/Cofree.scala

+
+  /** Evaluate just the tail. */
+  def forceTail: Cofree[S, A] =
+    Cofree[S, A](head, Eval.now(tailEval.value))


this is not tested, can we add a test?

Yeppers. They'll use side effects but meh, I'll steal them from the eval tests.

johnynek · 2016-12-02T23:55:35Z

free/src/main/scala/cats/free/Cofree.scala

+
+  /** Evaluate the entire Cofree tree. */
+  def forceAll(implicit S: Functor[S]): Cofree[S, A] =
+    Cofree[S, A](head, Eval.now(tailEval.map(S.map(_)(_.forceAll)).value))


this is not tested, can we add a test?

Absotively.

johnynek · 2016-12-03T00:02:55Z

free/src/main/scala/cats/free/Cofree.scala

+  override final def reduceLeftTo[A, B](fa: Cofree[F, A])(z: A => B)(f: (B, A) => B): B =
+    F.foldLeft(fa.tailForced, z(fa.head))((b, cof) => foldLeft(cof, b)(f))
+
+  override def reduceRightTo[A, B](fa: Cofree[F, A])(z: A => B)(f: (A, Eval[B]) => Eval[B]): Eval[B] = {


why did you do this vs something like:

f(ta.head, fa.tailEval.map { reduceRightTo(_)(z)(f) })

Try doing that yourself. Doesn't typecheck as far as I can see; you're expecting tailEval to hold an Eval[Cofree[F, A]] instead of an Eval[F[Cofree[F, A]]].

adelbertc · 2016-12-03T01:07:34Z

free/src/test/scala/cats/free/CofreeTests.scala

+  implicit val iso = CartesianTests.Isomorphisms.invariant[Cofree[Option, ?]]
+
+  checkAll("Cofree[Option, ?]", ComonadTests[Cofree[Option, ?]].comonad[Int, Int, Int])
+  locally {


woah what's this locally thing do?

http://stackoverflow.com/questions/3237727/what-does-predef-locally-do-and-how-is-it-different-from-predef-identity

edmundnoble · 2016-12-10T09:14:43Z

I've added tests for the methods @johnynek pointed out and fixed the tests for mapBranchingS/mapBranchingT/mapBranchingRoot/unfold which weren't running (but looked to be).

peterneyens · 2016-12-10T11:30:30Z

free/src/main/scala/cats/free/Cofree.scala

+
+}
+
+sealed abstract class CofreeInstances2 {


Maybe you could make the CofreeInstancesX classes and the CofreeComonad, CofreeReducible and CofreeTraversable traits package private (private[free]) (mostly so they don't show up in the ScalaDoc) ?

Very right. One more feature to add to the eventual pedantic-scala tool that will come out once scala.meta has a semantic API ;)

adelbertc · 2016-12-12T05:57:31Z

LGTM 👍

xuwei-k · 2016-12-12T06:08:48Z

free/src/main/scala/cats/free/Cofree.scala

+}
+
+sealed private[free] abstract class CofreeInstances1 extends CofreeInstances2 {
+  /** low priority `Traverse` instance */


I wrote the comments in scalaz.Cofree because there are two methos in scalaz.

scalaz/scalaz@a05ab2c0053d54b37474

https://github.com/scalaz/scalaz/blob/v7.2.8/core/src/main/scala/scalaz/Cofree.scala#L128

/** low priority `Traverse1` instance */ implicit def cofreeTraverse[F[_]: Traverse]: Traverse1[Cofree[F, ?]] =

https://github.com/scalaz/scalaz/blob/v7.2.8/core/src/main/scala/scalaz/Cofree.scala#L155

/** high priority `Traverse1` instance. more efficient */ implicit def cofreeTraverse1[F[_]: Traverse1]: Traverse1[Cofree[F, ?]] =

Cats have only one method for Traverse[Cofree[F, ?]]

Ahhh I was wondering why. Removed.

adelbertc · 2016-12-14T19:03:30Z

@xuwei-k Looks like @edmundnoble removed it - what do you think?

travisbrown

👍 from me.

travisbrown · 2016-12-21T00:03:48Z

free/src/main/scala/cats/free/Cofree.scala

+  * [[https://github.com/scalaz/scalaz/blob/series/7.3.x/core/src/main/scala/scalaz/Cofree.scala Scalaz's Cofree]],
+  * originally written by Rúnar Bjarnason.
+  */
+final case class Cofree[S[_], A](head: A, tailEval: Eval[S[Cofree[S, A]]]) {


Could we just call this tail? In general it seems like the convention is to add the suffix only when it's needed for disambiguation (although afaik that's not documented anywhere).

The suffix is there to disambiguate it with tailForced, but I think it's reasonable for tail to be the safe version.

travisbrown · 2016-12-21T00:05:18Z

free/src/main/scala/cats/free/Cofree.scala

+package free
+
+/**
+  * A free comonad for some branching functor `S`. Branching is done lazily using Eval.


This is super nitpicky but could you put [[]] around Eval and the Free in the next line? When I'm reading Scaladocs I always get pissed off if the authors make me search to follow a reference to another class.

tpolecat · 2016-12-21T00:16:18Z

free/src/main/scala/cats/free/Cofree.scala

+  /** Cofree anamorphism, lazily evaluated. */
+  def unfold[F[_], A](a: A)(f: A => F[A])(implicit F: Functor[F]): Cofree[F, A] =
+    Cofree[F, A](a, Eval.later(F.map(f(a))(unfold(_)(f))))
+


It would be nice to have the monadic ana as well.

def unfoldM[M[_]: Monad, F[_]: Traverse, A](a: A)(f: A => M[F[A]]): M[Cofree[F, A]]

tpolecat · 2016-12-21T00:18:13Z

free/src/main/scala/cats/free/Cofree.scala

+  /**
+    * A stack-safe algebraic recursive fold out of the cofree comonad.
+    */
+  def cata[F[_], A, B](cof: Cofree[F, A])(folder: (A, F[B]) => Eval[B])(implicit F: Traverse[F]): Eval[B] =


I am unhappy that Eval forces Traverse[F] when the strict version would only need Functor.

Let the record show that I am also made unhappy by this.

tpolecat

I had a couple comments but this looks good to me. 👍

edmundnoble · 2016-12-21T00:35:28Z

All cleaned up @tpolecat and @travisbrown, thanks for the reviews :)

tpolecat · 2016-12-21T00:43:30Z

free/src/main/scala/cats/free/Cofree.scala

+    Cofree[F, A](a, Eval.later(F.map(f(a))(unfold(_)(f))))
+
+  /** Cofree monadic anamorphism, lazily evaluated. */
+  def unfoldM[F[_], M[_], A](a: A)(f: A => M[F[A]])(implicit F: Traverse[F], M: Monad[M]): M[Cofree[F, A]] = {


Can you convince me this is lazy? I don't quite see it.

(Corrected: I meant to say lazy.)

I cannot for the life of me figure out how to make it lazy. I have removed it for now just because it's unsafe otherwise, but I'd like to add it back soon. Edit: Maybe for CofreeT. Edit two: I believe the reason for this is we want to get all the effects on top of the type, but we want to keep the values inside the type. Can't bind if you haven't done the work yet.

Removed Cofree[List, A] tests and added attribution Remove unused imports Add Reducible instance, separate tests, fix Traverse1 comment Removed unfoldStrategy, added a test for unfold Tests for Cofree.mapBranchingRoot/mapBranchingS/T Add Cofree.tailForced and Cofree.forceAll tests, fix notrunning tests Make instances classes package-private Added Cofree.cata/cataM Add docs Add test for forceTail Remove comments tailEval=>tail, bracket doc references Add unfoldM Removed unfoldM

dwhitney · 2016-12-21T03:12:55Z

unfoldM is correct ;)

…

On Tue, Dec 20, 2016 at 10:00 PM, Rob Norris ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In free/src/main/scala/cats/free/Cofree.scala <#1446>: > + Cofree[S, A](head, Eval.now(tail.value)) + + /** Evaluate the entire Cofree tree. */ + def forceAll(implicit S: Functor[S]): Cofree[S, A] = + Cofree[S, A](head, Eval.now(tail.map(S.map(_)(_.forceAll)).value)) + +} + +object Cofree extends CofreeInstances { + + /** Cofree anamorphism, lazily evaluated. */ + def unfold[F[_], A](a: A)(f: A => F[A])(implicit F: Functor[F]): Cofree[F, A] = + Cofree[F, A](a, Eval.later(F.map(f(a))(unfold(_)(f)))) + + /** Cofree monadic anamorphism, lazily evaluated. */ + def unfoldM[F[_], M[_], A](a: A)(f: A => M[F[A]])(implicit F: Traverse[F], M: Monad[M]): M[Cofree[F, A]] = { FINE — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1446>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AACCoYHpgRKN5utktU8H5CLi8jFGlGEEks5rKJZBgaJpZM4Kj4mH> .

edmundnoble · 2016-12-21T03:21:43Z

@dwhitney I tried running it and it looped forever eagerly. It's correct in that it produces the right answer, but it's incorrect in that it makes Cofree's laziness useless.

peterneyens · 2016-12-21T10:38:01Z

Thanks @edmundnoble !

adelbertc reviewed Oct 28, 2016

View reviewed changes

edmundnoble force-pushed the cofree-comonad-2 branch 3 times, most recently from df17df2 to 4a4db93 Compare October 29, 2016 01:34

xuwei-k suggested changes Nov 2, 2016

View reviewed changes

edmundnoble force-pushed the cofree-comonad-2 branch 2 times, most recently from 0db086d to 85f6852 Compare November 4, 2016 02:59

edmundnoble force-pushed the cofree-comonad-2 branch from 85f6852 to 1842303 Compare November 4, 2016 14:56

johnynek reviewed Dec 3, 2016

View reviewed changes

adelbertc approved these changes Dec 3, 2016

View reviewed changes

edmundnoble force-pushed the cofree-comonad-2 branch from 1842303 to 12e8d14 Compare December 10, 2016 09:12

peterneyens reviewed Dec 10, 2016

View reviewed changes

edmundnoble force-pushed the cofree-comonad-2 branch 3 times, most recently from 7dd8e06 to 56fc0c4 Compare December 11, 2016 09:46

xuwei-k suggested changes Dec 12, 2016

View reviewed changes

edmundnoble force-pushed the cofree-comonad-2 branch from 56fc0c4 to 07c89c9 Compare December 12, 2016 19:19

travisbrown reviewed Dec 21, 2016

View reviewed changes

tpolecat reviewed Dec 21, 2016

View reviewed changes

tpolecat approved these changes Dec 21, 2016

View reviewed changes

edmundnoble force-pushed the cofree-comonad-2 branch from 07c89c9 to e5e7ab8 Compare December 21, 2016 00:34

tpolecat reviewed Dec 21, 2016

View reviewed changes

edmundnoble force-pushed the cofree-comonad-2 branch from e5e7ab8 to b3cb2ff Compare December 21, 2016 01:35

edmundnoble force-pushed the cofree-comonad-2 branch from b3cb2ff to 52067df Compare December 21, 2016 02:26

peterneyens merged commit d49985d into typelevel:master Dec 21, 2016

kailuowang mentioned this pull request Jan 10, 2017

Cofree comonad #1422

Closed

Cofree comonad #1446

Cofree comonad #1446

Conversation

edmundnoble commented Oct 28, 2016

codecov-io commented Oct 28, 2016 • edited Loading

Current coverage is 92.02% (diff: 100%)

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

Choose a reason for hiding this comment

Choose a reason for hiding this comment

edmundnoble commented Oct 31, 2016

adelbertc commented Nov 1, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

edmundnoble commented Nov 4, 2016

dwhitney commented Dec 2, 2016

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

Choose a reason for hiding this comment

edmundnoble commented Dec 10, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

adelbertc commented Dec 12, 2016

xuwei-k Dec 12, 2016 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

adelbertc commented Dec 14, 2016

travisbrown left a 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

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tpolecat left a comment

Choose a reason for hiding this comment

edmundnoble commented Dec 21, 2016

tpolecat Dec 21, 2016 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

edmundnoble Dec 21, 2016 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dwhitney commented Dec 21, 2016 via email

edmundnoble commented Dec 21, 2016

peterneyens commented Dec 21, 2016

codecov-io commented Oct 28, 2016 •

edited

Loading

xuwei-k Dec 12, 2016 •

edited

Loading

tpolecat Dec 21, 2016 •

edited

Loading

edmundnoble Dec 21, 2016 •

edited

Loading