Register type for literal list/map.

Register type for literal list/map.


BUG=

Committed: https://code.google.com/p/dart/source/detail?r=21136

[karlklose, 2013-04-08 14:00:37]

LGTM.

[Johnni Winther, 2013-04-09 08:06:14]

Committed patchset #1 manually as r21136 (presubmit successful).

[ngeoffray, 2013-04-10 11:35:51]

What is the net result of this change? Do we know have the right runtime type
for the literal lists and maps? If yes, shouldn't we have tests that pass now?

https://codereview.chromium.org/13467021/diff/1/sdk/lib/_internal/compiler/im...
File sdk/lib/_internal/compiler/implementation/dart_types.dart (right):

https://codereview.chromium.org/13467021/diff/1/sdk/lib/_internal/compiler/im...
sdk/lib/_internal/compiler/implementation/dart_types.dart:455: message: () =>
'Invalid type argument count on ${element.thisType}. '
Why this change?

[Johnni Winther, 2013-04-10 11:43:21]

On 2013/04/10 11:35:51, ngeoffray wrote:
> What is the net result of this change? Do we know have the right runtime type
> for the literal lists and maps? If yes, shouldn't we have tests that pass now?
> 
>
https://codereview.chromium.org/13467021/diff/1/sdk/lib/_internal/compiler/im...
> File sdk/lib/_internal/compiler/implementation/dart_types.dart (right):
> 
>
https://codereview.chromium.org/13467021/diff/1/sdk/lib/_internal/compiler/im...
> sdk/lib/_internal/compiler/implementation/dart_types.dart:455: message: () =>
> 'Invalid type argument count on ${element.thisType}. '
> Why this change?

The change is made in preparation for typechecking. The missing computation of
the raw type of a map was revealed during this. Pulling the a [null] as the raw
type is silently handled by the registration code, so it currently causes no
error.

[Johnni Winther, 2013-04-10 11:43:27]

https://codereview.chromium.org/13467021/diff/1/sdk/lib/_internal/compiler/im...
File sdk/lib/_internal/compiler/implementation/dart_types.dart (right):

https://codereview.chromium.org/13467021/diff/1/sdk/lib/_internal/compiler/im...
sdk/lib/_internal/compiler/implementation/dart_types.dart:455: message: () =>
'Invalid type argument count on ${element.thisType}. '
On 2013/04/10 11:35:51, ngeoffray wrote:
> Why this change?

Eager evaluation of the message could provoke a crash (through the evaluation of
${element.thisType} or $typeArguments).

[ngeoffray, 2013-04-10 11:49:47]

On 2013/04/10 11:43:21, Johnni Winther wrote:
> On 2013/04/10 11:35:51, ngeoffray wrote:
> > What is the net result of this change? Do we know have the right runtime
type
> > for the literal lists and maps? If yes, shouldn't we have tests that pass
now?
> > 
> >
>
https://codereview.chromium.org/13467021/diff/1/sdk/lib/_internal/compiler/im...
> > File sdk/lib/_internal/compiler/implementation/dart_types.dart (right):
> > 
> >
>
https://codereview.chromium.org/13467021/diff/1/sdk/lib/_internal/compiler/im...
> > sdk/lib/_internal/compiler/implementation/dart_types.dart:455: message: ()
=>
> > 'Invalid type argument count on ${element.thisType}. '
> > Why this change?
> 
> The change is made in preparation for typechecking. The missing computation of
> the raw type of a map was revealed during this. Pulling the a [null] as the
raw
> type is silently handled by the registration code, so it currently causes no
> error.

OK. Looking at builder.dart, it looks like we're not using the type variables
for list and map literals for codegen. So this change should help implement that
too, right?

[Johnni Winther, 2013-04-10 13:16:38]

On 2013/04/10 11:49:47, ngeoffray wrote:
> On 2013/04/10 11:43:21, Johnni Winther wrote:
> > On 2013/04/10 11:35:51, ngeoffray wrote:
> > > What is the net result of this change? Do we know have the right runtime
> type
> > > for the literal lists and maps? If yes, shouldn't we have tests that pass
> now?
> > > 
> > >
> >
>
https://codereview.chromium.org/13467021/diff/1/sdk/lib/_internal/compiler/im...
> > > File sdk/lib/_internal/compiler/implementation/dart_types.dart (right):
> > > 
> > >
> >
>
https://codereview.chromium.org/13467021/diff/1/sdk/lib/_internal/compiler/im...
> > > sdk/lib/_internal/compiler/implementation/dart_types.dart:455: message: ()
> =>
> > > 'Invalid type argument count on ${element.thisType}. '
> > > Why this change?
> > 
> > The change is made in preparation for typechecking. The missing computation
of
> > the raw type of a map was revealed during this. Pulling the a [null] as the
> raw
> > type is silently handled by the registration code, so it currently causes no
> > error.
> 
> OK. Looking at builder.dart, it looks like we're not using the type variables
> for list and map literals for codegen. So this change should help implement
that
> too, right?

Yes. The type (with the correct type arguments) should now be in
elements.getType(node).