Generalize with-chain wrapping for defs and type definitions #1132

iantabolt · 2018-04-11T20:51:44Z

This fixes #1125 to handle wrapping of "with-chains".

For example, a long chain of

Foo with Bar with Baz

can be wrapped as

Foo
  with Bar
  with Baz

This was already implemented for self-types (trait Xyz { self: Foo with Bar with Baz => ). This PR expands the behavior to handle type definitions (type MyType = Foo with Bar with Baz) and function return types (def xyz: Foo with Bar with Baz).

iantabolt · 2018-04-11T20:52:42Z

scalafmt-core/shared/src/main/scala/org/scalafmt/internal/Router.scala

@@ -1021,7 +1021,7 @@ class Router(formatOps: FormatOps) {
              Split(Space, 0),
              Split(Newline, 1).withPolicy(policy)
            )
-          case t @ SelfAnnotation(top) =>
+          case t @ WithChain(top) =>


I don't love this name but I wanted to change it to something more general than SelfAnnotation. Any suggestions?

This name is good 👍

olafurpg

Thanks for this contribution! New name looks good, we should make this work for all result types.

olafurpg · 2018-04-11T22:23:50Z

scalafmt-core/shared/src/main/scala/org/scalafmt/util/TreeExtractors.scala

      case (top: Type.With) `:parent:` (_: Term.Param) `:parent:` (_: Template) =>
        Some(top)
+      // type definitions
+      case (top: Type.With) `:parent:` (_: Defn.Type) =>


I suspect this pattern can be extended to any Defn or Decl, that will cover val/var/def regardless if they're abstract or not.

I'm not sure when we don't want this to match for Type.With, there are other tree nodes like Term.ApplyType where this is equally relevant.

Yeah I agree for the most part.

If I open it up to match all Type.Withs then one test fails (with good reason IMO):

// Before: { def getAnnotationOwner( annotationOwnerLookUp: ScLiteral => Option[ PsiAnnotationOwner with PsiElement]): Option[PsiAnnotationOwner] = 1 } // After -- breaks in generic type: { def getAnnotationOwner( annotationOwnerLookUp: ScLiteral => Option[PsiAnnotationOwner with PsiElement]): Option[PsiAnnotationOwner] = 1 } // Expected: { def getAnnotationOwner( annotationOwnerLookUp: ScLiteral => Option[ PsiAnnotationOwner with PsiElement]): Option[PsiAnnotationOwner] = 1 }

From https://github.com/scalameta/scalafmt/blob/master/scalafmt-tests/src/test/resources/default/TypeWith.stat

But even in that scenario, if the type were long enough we probably want a newline instead of violating the column limit. Maybe we can make this split a low priority or high cost etc?

ceedubs · 2018-04-12T15:58:14Z

This looks like a great change! I'm adding some examples here based on @olafurpg's request in #1133.

In Cats when we have code like:

  implicit val catsDataInstancesForNonEmptyList: SemigroupK[NonEmptyList]
    with Reducible[NonEmptyList] with Bimonad[NonEmptyList] with NonEmptyTraverse[NonEmptyList] = ...

  implicit val catsDataInstancesForNonEmptyVector: SemigroupK[NonEmptyVector]
    with Reducible[NonEmptyVector] with Bimonad[NonEmptyVector]
    with NonEmptyTraverse[NonEmptyVector] = ...

   implicit val catsStdInstancesForOption
    : Traverse[Option] with MonadError[Option, Unit] with Alternative[Option]
      with CommutativeMonad[Option] with CoflatMap[Option] = ...

After formatting it ends up looking like:

  implicit val catsDataInstancesForNonEmptyList: SemigroupK[NonEmptyList] with Reducible[
    NonEmptyList] with Bimonad[NonEmptyList] with NonEmptyTraverse[NonEmptyList] = ...

  implicit val catsDataInstancesForNonEmptyVector: SemigroupK[NonEmptyVector] with Reducible[
    NonEmptyVector] with Bimonad[NonEmptyVector] with NonEmptyTraverse[NonEmptyVector] = ...

   implicit val catsStdInstancesForOption
    : Traverse[Option] with MonadError[Option, Unit] with Alternative[Option] with CommutativeMonad[
      Option] with CoflatMap[Option] = ...

It would be nice to see the style changes exhibited in this PR to extend to the return types of vals like these.

iantabolt · 2018-04-12T16:58:52Z

@ceedubs Thanks for the input! I'll add these as test cases.

I ran the examples through with this change and it ended up with

implicit val catsDataInstancesForNonEmptyList: SemigroupK[NonEmptyList]
  with Reducible[NonEmptyList]
  with Bimonad[NonEmptyList]
  with NonEmptyTraverse[NonEmptyList] = ???

implicit val catsDataInstancesForNonEmptyVector: SemigroupK[NonEmptyVector]
  with Reducible[NonEmptyVector]
  with Bimonad[NonEmptyVector]
  with NonEmptyTraverse[NonEmptyVector] = ???

implicit val catsStdInstancesForOption: Traverse[Option] with MonadError[Option, Unit]
  with Alternative[Option]
  with CommutativeMonad[Option]
  with CoflatMap[Option] = ???

Does that seem about right? I'm not sure why the last one didn't drop MonadError to the next line 🤔

Edit: Figured it out. I was running with { case With(top) => Some(top) } which doesn't actually enforce that it is the top with.

ceedubs · 2018-04-12T18:09:58Z

@iantabolt nice! On the last one, maybe it has to do with inlining until you hit the column limit? I'm not super familiar with the particulars of behavior. And does it depend on binPack.parentConstructors? Either way, I think that it's a really big improvement and I'd be happy with it.

iantabolt · 2018-04-12T19:51:32Z

scalafmt-tests/src/test/resources/unit/Type.stat

+  with HasThrottles
+  with HasThrowableNotifier
+<<< abstract def #1125
+trait NeedsServices {


This probably shouldn't be in Type.stat since there is no type. Maybe all these tests should be in TypeWith.stat?

TypeWith.stat would be good, this is important enough to merit a separate suite.

olafurpg

A couple more minor comments.

olafurpg · 2018-04-13T18:42:14Z

scalafmt-core/shared/src/main/scala/org/scalafmt/internal/Router.scala

@@ -1021,7 +1021,7 @@ class Router(formatOps: FormatOps) {
              Split(Space, 0),
              Split(Newline, 1).withPolicy(policy)
            )
-          case t @ SelfAnnotation(top) =>
+          case t @ WithChain(top) =>


This name is good 👍

olafurpg · 2018-04-13T18:45:14Z

scalafmt-tests/src/test/resources/default/TypeWith.stat

+   : Traverse[Option] with MonadError[Option, Unit] with Alternative[Option]
+       with CommutativeMonad[Option] with CoflatMap[Option] = ???
+>>>
+implicit val catsStdInstancesForOption: Traverse[Option]


This may be more ambitious that you were planning for (so "no" is an acceptable answer), but while were fixing with chains, could we add a test case for?

new A with B with C with D ...

I've seen surprising behavior in these scenarios and I suspect fixing it likewise is a matter of adding another case to WithChain

Yes no problem (I think)

I might need some help with this. I don't think this falls into the Type.With case and instead hits the Template case.

My thought is to treat this like the KwExtends case above and do

case tok @ FormatToken(_, right @ KwWith(), _) => rightOwner match { + // Anonymous template (new A with B { ... }) + case template: Template if template.parent.exists(_.is[Term.New]) => + val lastToken = template.tokens.last + binPackParentConstructorSplits( + Set(template), + lastToken, + 2 + ) case template: Template => val hasSelfAnnotation = template.self.tokens.nonEmpty

This almost works but it indents after every with

new A with B with C with D

Any ideas?

I pushed the non-working change mentioned above. I'm assuming the fix is something analogous to the isFirstWith logic but I'm not sure how to achieve it with the template

Hmm, the changes seem unrelated enough to merit a separate PR. Your change already improves on the state-of-the-art so I propose you drop the last commit, we merge and move this case to another ticket.

I'll need to locally play around with this case to see how would be best to solve it.

SGTM 👍. I rolled it back

olafurpg · 2018-04-13T18:46:05Z

scalafmt-tests/src/test/resources/unit/Type.stat

+  with HasThrottles
+  with HasThrowableNotifier
+<<< abstract def #1125
+trait NeedsServices {


TypeWith.stat would be good, this is important enough to merit a separate suite.

olafurpg

BTW, you can see what type nodes are for a given syntax with the ammonite repl like this

@ import $ivy.`org.scalameta:testkit_2.12:1.7.0`, scala.meta._, testkit._
import $ivy.$                                 , scala.meta._, testkit._

@ q"new A with B with C"
res1: Term.New = Term.New(
  Template(
    List(),
    List(Ctor.Ref.Name("A"), Ctor.Ref.Name("B"), Ctor.Ref.Name("C")),
    Term.Param(List(), _, None, None),
    None
  )
)

olafurpg

LGTM 👍 Thanks for this contribution @iantabolt !

olafurpg · 2018-04-14T11:25:43Z

I'll be attending a conference for the next week, but I hope to have v1.5.0 out this April. Until then, you can already use the pre-releases to Bintray as documented on the website but note those don't include the IntelliJ plugin.

tian000 · 2019-09-07T05:44:32Z

What configuration setting do you use to wrap with-chains? @olafurpg @iantabolt

olafurpg and others added 3 commits April 11, 2018 13:22

Treat Defn.Type with chains the same as self types.

2a04ef3

Add harder scalameta#1125 tests

1873e71

Better tests for scalameta#1125 and rename SelfAnnotation=>WithChain

3d06212

iantabolt commented Apr 11, 2018

View reviewed changes

olafurpg requested changes Apr 11, 2018

View reviewed changes

olafurpg mentioned this pull request Apr 12, 2018

Awkward line breaks with long/compound types with type parameters #1133

Closed

Open up with formats to all Defn/Decl; add cats test for scalameta#1133

d9a481c

iantabolt commented Apr 12, 2018

View reviewed changes

olafurpg requested changes Apr 13, 2018

View reviewed changes

olafurpg reviewed Apr 13, 2018

View reviewed changes

Move tests from Type.stat to TypeWith.stat

b60a53d

iantabolt force-pushed the 1125 branch from 90da50e to b60a53d Compare April 13, 2018 21:49

olafurpg approved these changes Apr 13, 2018

View reviewed changes

olafurpg merged commit 906ccd9 into scalameta:master Apr 14, 2018

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Generalize with-chain wrapping for defs and type definitions #1132

Generalize with-chain wrapping for defs and type definitions #1132

iantabolt commented Apr 11, 2018

iantabolt Apr 11, 2018 •

edited

olafurpg Apr 13, 2018

olafurpg left a comment

olafurpg Apr 11, 2018

olafurpg Apr 11, 2018

iantabolt Apr 11, 2018

iantabolt Apr 11, 2018

ceedubs commented Apr 12, 2018

iantabolt commented Apr 12, 2018 •

edited

ceedubs commented Apr 12, 2018

iantabolt Apr 12, 2018 •

edited

olafurpg Apr 13, 2018

olafurpg left a comment

olafurpg Apr 13, 2018

olafurpg Apr 13, 2018

iantabolt Apr 13, 2018

iantabolt Apr 13, 2018

iantabolt Apr 13, 2018

olafurpg Apr 13, 2018

iantabolt Apr 13, 2018

olafurpg Apr 13, 2018

olafurpg left a comment

olafurpg left a comment

olafurpg commented Apr 14, 2018

tian000 commented Sep 7, 2019 •

edited

Generalize with-chain wrapping for defs and type definitions #1132

Generalize with-chain wrapping for defs and type definitions #1132

Conversation

iantabolt commented Apr 11, 2018

iantabolt Apr 11, 2018 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

olafurpg 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

ceedubs commented Apr 12, 2018

iantabolt commented Apr 12, 2018 • edited

ceedubs commented Apr 12, 2018

iantabolt Apr 12, 2018 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

olafurpg 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

olafurpg left a comment

Choose a reason for hiding this comment

olafurpg left a comment

Choose a reason for hiding this comment

olafurpg commented Apr 14, 2018

tian000 commented Sep 7, 2019 • edited

iantabolt Apr 11, 2018 •

edited

iantabolt commented Apr 12, 2018 •

edited

iantabolt Apr 12, 2018 •

edited

tian000 commented Sep 7, 2019 •

edited