Refine the type of a local after it has been used as the receiver of a call.

Refine the type of a local after it has been used as the receiver of a call.

R=karlklose@google.com

Committed: https://code.google.com/p/dart/source/detail?r=26342

[karlklose, 2013-08-19 10:52:33]

LGTM.

https://codereview.chromium.org/22947008/diff/10001/sdk/lib/_internal/compile...
File sdk/lib/_internal/compiler/implementation/universe/function_set.dart
(right):

https://codereview.chromium.org/22947008/diff/10001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/universe/function_set.dart:210:
TypeMask computeMask(Compiler compiler) {
Please add a comment to describe what this function computes.

https://codereview.chromium.org/22947008/diff/10001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/universe/function_set.dart:216: ?
[[cls], compiler.world.mixinUses[cls]].expand((x) => x)
How about writing this as
  'new List.from(compiler.world.mixinUses[cls])..add(cls)'
or
  '[cls].addAll(compiler.world.isUsedAsMixin(cls)
                ? compiler.world.mixinUses[cls] : [])'
?

https://codereview.chromium.org/22947008/diff/10001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/universe/function_set.dart:220:
Set<ClassElement> subclasses = compiler.world.subclassesOf(cls);
Consider adding a helper for this test:
 hasSubclasses(ClassElement cls) {
   Set<ClassElement> subclasses = compiler.world.subclassesOf(cls);
   return subclasses == null || subclasses.isEmpty
 }

[ngeoffray, 2013-08-20 08:00:48]

Thanks Karl.

https://codereview.chromium.org/22947008/diff/10001/sdk/lib/_internal/compile...
File sdk/lib/_internal/compiler/implementation/universe/function_set.dart
(right):

https://codereview.chromium.org/22947008/diff/10001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/universe/function_set.dart:210:
TypeMask computeMask(Compiler compiler) {
On 2013/08/19 10:52:33, karlklose wrote:
> Please add a comment to describe what this function computes.

Done.

https://codereview.chromium.org/22947008/diff/10001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/universe/function_set.dart:216: ?
[[cls], compiler.world.mixinUses[cls]].expand((x) => x)
On 2013/08/19 10:52:33, karlklose wrote:
> How about writing this as
>   'new List.from(compiler.world.mixinUses[cls])..add(cls)'
> or
>   '[cls].addAll(compiler.world.isUsedAsMixin(cls)
>                 ? compiler.world.mixinUses[cls] : [])'
> ?

I did not want to create a new list again. By using those methods I hope the
underlying implementation just caches them.

https://codereview.chromium.org/22947008/diff/10001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/universe/function_set.dart:220:
Set<ClassElement> subclasses = compiler.world.subclassesOf(cls);
On 2013/08/19 10:52:33, karlklose wrote:
> Consider adding a helper for this test:
>  hasSubclasses(ClassElement cls) {
>    Set<ClassElement> subclasses = compiler.world.subclassesOf(cls);
>    return subclasses == null || subclasses.isEmpty
>  }

Done.

[ngeoffray, 2013-08-20 09:13:34]

https://codereview.chromium.org/22947008/diff/10001/sdk/lib/_internal/compile...
File sdk/lib/_internal/compiler/implementation/universe/function_set.dart
(right):

https://codereview.chromium.org/22947008/diff/10001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/universe/function_set.dart:216: ?
[[cls], compiler.world.mixinUses[cls]].expand((x) => x)
On 2013/08/20 08:00:48, ngeoffray wrote:
> On 2013/08/19 10:52:33, karlklose wrote:
> > How about writing this as
> >   'new List.from(compiler.world.mixinUses[cls])..add(cls)'
> > or
> >   '[cls].addAll(compiler.world.isUsedAsMixin(cls)
> >                 ? compiler.world.mixinUses[cls] : [])'
> > ?
> 
> I did not want to create a new list again. By using those methods I hope the
> underlying implementation just caches them.

After further discussions and talking to the library team, I changed it to use
addAlll instead. More readable, and we'll have to rewrite it anyway if we enver
end up having a lazy concat.

[ngeoffray, 2013-08-20 09:36:50]

Committed patchset #5 manually as r26342 (presubmit successful).