Partially revert 4530d3aa93131ec28096968a3c903ed1016dbf1b. Add back v… #1506

marcin-rzeznicki · 2016-12-29T12:16:58Z

…ariance on methods in EitherT and Either syntax.
Current signatures make it impossible to write, for example, for-comprehensions with 'lefts' not being of the same type. The following code does not compile with current signatures, but it compiles after the change (compiler is able to infer the common supertype):

sealed abstract class AppError
case object Error1 extends AppError
case object Error2 extends AppError

val either1: Future[Either[Error1.type , String]] = Future.successful(Right("hi"))
val either2: Future[Either[Error2.type , String]] = Future.successful(Right("bye"))

for {
  s1 <- EitherT(either1)
  s2 <- EitherT[Future, AppError, String](either2)
} yield s1 ++ s2

Error:(17, 7) type mismatch;
found : cats.data.EitherT[scala.concurrent.Future,Error2.type,String]
required: cats.data.EitherT[scala.concurrent.Future, Error1.type,?]
s2 <- EitherT(either2)
^

This change has been discussed on gitter with @travisbrown and @edmundnoble

… Either syntax.

codecov-io · 2016-12-29T12:54:36Z

Current coverage is 92.32% (diff: 100%)

Merging #1506 into master will increase coverage by 0.24%

@@             master      #1506   diff @@
==========================================
  Files           246        246          
  Lines          3659       3659          
  Methods        3534       3540     +6   
  Messages          0          0          
  Branches        125        119     -6   
==========================================
+ Hits           3369       3378     +9   
+ Misses          290        281     -9   
  Partials          0          0

Powered by Codecov. Last update 974b840...5cafcdd

travisbrown · 2016-12-29T22:14:08Z

👍 from me. We'll need to fix the docs (for example) to match, and I'd prefer to see that happen in this PR, I guess, but it could be a follow-up.

johnynek · 2016-12-30T00:11:09Z

core/src/main/scala/cats/syntax/either.scala

      }
  }

-  def partialCompare(that: Either[A, B])(implicit A: PartialOrder[A], B: PartialOrder[B]): Double = eab match {
+  def partialCompare[AA >: A, BB >: B](that: Either[AA, BB])(implicit AA: PartialOrder[AA], BB: PartialOrder[BB]): Double = eab match {


this method is untested. Is it too much to ask to write a test of some kind during this PR?

Not at all. Will do

Actually this one should be covered by order laws at https://github.com/typelevel/cats/blob/master/tests/src/test/scala/cats/tests/EitherTests.scala#L44 , right?

EDIT: Ah, it isn't because implementation of partialCompare in EitherInstances is duplicated in EitherOps. Is it better to dedup these two?

johnynek · 2016-12-30T00:11:24Z

core/src/main/scala/cats/syntax/either.scala

-  def show(implicit A: Show[A], B: Show[B]): String = eab match {
-    case Left(a)  => s"Left(${A.show(a)})"
-    case Right(b) => s"Right(${B.show(b)})"
+  def show[AA >: A, BB >: B](implicit AA: Show[AA], BB: Show[BB]): String = eab match {


this is untested. Can we add a simple test?

johnynek · 2016-12-30T00:11:39Z

core/src/main/scala/cats/syntax/either.scala

  }

-  def ap[C](that: Either[A, B => C]): Either[A, C] = (new EitherOps(that)).flatMap(this.map)
+  def ap[AA >: A, BB >: B, C](that: Either[AA, BB => C]): Either[AA, C] = new EitherOps(that).flatMap(this.map)


this is untested. Can we add a simple test?

marcin-rzeznicki · 2016-12-30T00:51:37Z

I'll take care of docs update as well, Thanks for pointing out.

johnynek · 2016-12-30T04:49:27Z

My codecov plugin seems to say that codecov thinks it is uncovered.

…

On Thu, Dec 29, 2016 at 16:39 Marcin Rzeźnicki ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In core/src/main/scala/cats/syntax/either.scala <#1506>: > } } - def partialCompare(that: Either[A, B])(implicit A: PartialOrder[A], B: PartialOrder[B]): Double = eab match { + def partialCompare[AA >: A, BB >: B](that: Either[AA, BB])(implicit AA: PartialOrder[AA], BB: PartialOrder[BB]): Double = eab match { Actually this one should be covered by order laws at https://github.com/typelevel/cats/blob/master/tests/src/test/scala/cats/tests/EitherTests.scala#L44 , right? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1506>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAEJdtPkInW2SvIicSWzHBYtoqqMjzcDks5rNG7GgaJpZM4LXi16> .

johnynek · 2016-12-30T04:50:27Z

Maybe we aren't exercising this call site. The laws can't be calling this one since PartialOrder is invariant. On Thu, Dec 29, 2016 at 18:49 P. Oscar Boykin <oscar.boykin@gmail.com> wrote:

…

My codecov plugin seems to say that codecov thinks it is uncovered. On Thu, Dec 29, 2016 at 16:39 Marcin Rzeźnicki ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In core/src/main/scala/cats/syntax/either.scala <#1506>: > } } - def partialCompare(that: Either[A, B])(implicit A: PartialOrder[A], B: PartialOrder[B]): Double = eab match { + def partialCompare[AA >: A, BB >: B](that: Either[AA, BB])(implicit AA: PartialOrder[AA], BB: PartialOrder[BB]): Double = eab match { Actually this one should be covered by order laws at https://github.com/typelevel/cats/blob/master/tests/src/test/scala/cats/tests/EitherTests.scala#L44 , right? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1506>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAEJdtPkInW2SvIicSWzHBYtoqqMjzcDks5rNG7GgaJpZM4LXi16> .

marcin-rzeznicki · 2016-12-30T14:57:30Z

Missing tests added in 5c86c44

marcin-rzeznicki · 2016-12-30T15:10:12Z

@travisbrown It looks like the whole 'Either in the small...' chapter can be removed from the docs since flatMap you get from enrichment is aligned with what you get on 2.12. WDYT?

travisbrown · 2016-12-30T15:17:33Z

@marcin-rzeznicki Yep, that makes sense to me.

…`flatMap` now behaves identically to 2.12

adelbertc · 2016-12-30T20:42:31Z

docs/src/main/tut/datatypes/either.md

@@ -212,132 +212,6 @@ magic("123") match {
 }
 ```

-## Either in the small, Either in the large
-Once you start using `Either` for all your error-handling, you may quickly run into an issue where


Why is this being removed?

@adelbertc The idea is that all the stuff about the left sides needing to be the same will be a lot less relevant (or not at all relevant, in the form here) once the syntax flatMap for Either behaves the same as the real flatMap method on Either in 2.12.

This section talks about using local ADTs for local errors and wrapping as you go wider and wider in your application. Variance is tangentially relevant to Solution 1 but does not help with Solution 2.

Okay, you've convinced me the removal should be more surgical. :)

Nonetheless, the whole section was presented as a 'solution' - but the problem being solved does not exist anymore. To quote the relevant part: 'If you're on Scala 2.12, this line will compile and work as expected (...)'. Everything that followed seemed to have been written as a guide to deal with issues on previous versions of Scala. It of course does not invalidate ideas contained therein. Obviously, organizing your errors as ADTs is a valid and useful technique. But a rather generic one and not really tied to usage of Either in cats. Of course I'm open to any suggestions, but I must admit that I simply do not know what should be there instead.

We should strive to present certain patterns (for lack of a better word) for using these data types - in this case I wrote it to show how one could use Either throughout their code to represent possibly failing computations.

adelbertc · 2016-12-30T21:31:15Z

docs/src/main/tut/datatypes/either.md

@@ -212,132 +212,6 @@ magic("123") match {
 }
 ```

-## Either in the small, Either in the large


Please be more careful about removing this section - variance does not subsume the concerns here.

Deleted section is still relevant - it was meant to present use of `Either` to reprsent failing computations. Variance or not, it is more aout how to use this data type. I tried to rewrote parts that are no longer true in presence of variance. Some other parts have been rearranged.

marcin-rzeznicki · 2016-12-31T01:23:03Z

@adelbertc I've come up with a new version of docs. Only some parts that had been no longer true in presence of variance were rewritten. Some other parts were rearranged. Please check if you like it. ( Since Contributors' Guide discourages squashing commits, it may be awkward to compare :-( )

adelbertc · 2016-12-31T01:36:14Z

Cool @marcin-rzeznicki , LGTM, thanks!

Partially revert 4530d3a. Add back variance on methods in EitherT and…

d084def

… Either syntax.

johnynek requested changes Dec 30, 2016

View reviewed changes

Add missing tests for Either syntax

5c86c44

Fix the docs to match added variance to Either's enriched flatMap. …

95d96fa

…`flatMap` now behaves identically to 2.12

travisbrown mentioned this pull request Dec 30, 2016

Prepare for 0.9.0 release #1508

Closed

11 tasks

adelbertc reviewed Dec 30, 2016

View reviewed changes

adelbertc suggested changes Dec 30, 2016

View reviewed changes

adelbertc approved these changes Dec 31, 2016

View reviewed changes

johnynek approved these changes Dec 31, 2016

View reviewed changes

travisbrown mentioned this pull request Jan 3, 2017

Draft of 0.9.0 release notes #1511

Merged

kailuowang approved these changes Jan 3, 2017

View reviewed changes

johnynek merged commit 4c3e244 into typelevel:master Jan 3, 2017

kailuowang mentioned this pull request Apr 24, 2017

improve EitherT inference for use in for-comprehension #1627

Closed

jtjeferreira mentioned this pull request Apr 24, 2017

add unit test for variance on methods in EitherT #1629

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Partially revert 4530d3aa93131ec28096968a3c903ed1016dbf1b. Add back v… #1506

Partially revert 4530d3aa93131ec28096968a3c903ed1016dbf1b. Add back v… #1506

marcin-rzeznicki commented Dec 29, 2016

codecov-io commented Dec 29, 2016 •

edited

Loading

travisbrown commented Dec 29, 2016

johnynek Dec 30, 2016

marcin-rzeznicki Dec 30, 2016

marcin-rzeznicki Dec 30, 2016 •

edited

Loading

johnynek Dec 30, 2016

marcin-rzeznicki Dec 30, 2016

johnynek Dec 30, 2016

marcin-rzeznicki commented Dec 30, 2016

johnynek commented Dec 30, 2016 via email

johnynek commented Dec 30, 2016 via email

marcin-rzeznicki commented Dec 30, 2016

marcin-rzeznicki commented Dec 30, 2016

travisbrown commented Dec 30, 2016

adelbertc Dec 30, 2016

travisbrown Dec 30, 2016

adelbertc Dec 30, 2016

travisbrown Dec 30, 2016

marcin-rzeznicki Dec 30, 2016 •

edited

Loading

adelbertc Dec 30, 2016

adelbertc Dec 30, 2016

marcin-rzeznicki commented Dec 31, 2016

adelbertc commented Dec 31, 2016

Partially revert 4530d3aa93131ec28096968a3c903ed1016dbf1b. Add back v… #1506

Partially revert 4530d3aa93131ec28096968a3c903ed1016dbf1b. Add back v… #1506

Conversation

marcin-rzeznicki commented Dec 29, 2016

codecov-io commented Dec 29, 2016 • edited Loading

Current coverage is 92.32% (diff: 100%)

travisbrown commented Dec 29, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

marcin-rzeznicki Dec 30, 2016 • edited Loading

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

marcin-rzeznicki commented Dec 30, 2016

johnynek commented Dec 30, 2016 via email

johnynek commented Dec 30, 2016 via email

marcin-rzeznicki commented Dec 30, 2016

marcin-rzeznicki commented Dec 30, 2016

travisbrown commented Dec 30, 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

marcin-rzeznicki Dec 30, 2016 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

marcin-rzeznicki commented Dec 31, 2016

adelbertc commented Dec 31, 2016

codecov-io commented Dec 29, 2016 •

edited

Loading

marcin-rzeznicki Dec 30, 2016 •

edited

Loading

marcin-rzeznicki Dec 30, 2016 •

edited

Loading