Implement type inference of method invocations in Fasta.

Implement type inference of method invocations in Fasta.

R=scheglov@google.com

Committed: https://github.com/dart-lang/sdk/commit/6e499b16cf7dbf3332d134999b35ece30d23bb65

[Paul Berry, 2017-05-09 23:03:23]

paulberry@google.com changed reviewers:
+ jmesserly@google.com, leafp@google.com, scheglov@google.com

[scheglov, 2017-05-09 23:45:57]

LGTM

https://codereview.chromium.org/2873813002/diff/1/pkg/front_end/lib/src/fasta...
File pkg/front_end/lib/src/fasta/type_inference/type_inferrer.dart (right):

https://codereview.chromium.org/2873813002/diff/1/pkg/front_end/lib/src/fasta...
pkg/front_end/lib/src/fasta/type_inference/type_inferrer.dart:414: /// is the
name fo the method.  [explicitTypeArguments] is the set of type
/fo/of/s

[Paul Berry, 2017-05-09 23:51:09]

Description was changed from

==========
Implement type inference of method invocations in Fasta.
==========

to

==========
Implement type inference of method invocations in Fasta.

R=scheglov@google.com

Committed:
https://github.com/dart-lang/sdk/commit/6e499b16cf7dbf3332d134999b35ece30d23bb65
==========

[Paul Berry, 2017-05-09 23:51:09]

Committed patchset #1 (id:1) manually as
6e499b16cf7dbf3332d134999b35ece30d23bb65 (presubmit successful).

[Leaf, 2017-05-10 04:23:10]

https://codereview.chromium.org/2873813002/diff/1/pkg/front_end/lib/src/fasta...
File pkg/front_end/lib/src/fasta/type_inference/type_inferrer.dart (right):

https://codereview.chromium.org/2873813002/diff/1/pkg/front_end/lib/src/fasta...
pkg/front_end/lib/src/fasta/type_inference/type_inferrer.dart:432: //
inferConstructorInvocation?
The specification is definitely intended to treat these equivalently.  That is,
a constructor call is just like a call to a static function with the appropriate
generic type signature.

https://codereview.chromium.org/2873813002/diff/1/pkg/front_end/lib/src/fasta...
pkg/front_end/lib/src/fasta/type_inference/type_inferrer.dart:459:
Substitution.fromPairs(memberTypeParameters, inferredTypes);
Is this right?  Looking at class Substitution from kernel, it doesn't look like
it does anything reasonable with a null inferredTypes.  Should this be
explicitTypeArguments?

https://codereview.chromium.org/2873813002/diff/1/pkg/front_end/lib/src/fasta...
pkg/front_end/lib/src/fasta/type_inference/type_inferrer.dart:587: FunctionType
_getCalleeFunctionType(DartType receiverType, Name methodName) {
Surely this should be factored out somewhere?  Or does nothing else really ever
need to do this after inference has happened (presumably because the type is
attached to the node or somesuch)?

https://codereview.chromium.org/2873813002/diff/1/pkg/front_end/lib/src/fasta...
pkg/front_end/lib/src/fasta/type_inference/type_inferrer.dart:628: //
TODO(paulberry): is there a kernel function that does this binary search?
There's just got to be...otherwise something is very wrong.

[Paul Berry, 2017-05-10 16:24:29]

https://codereview.chromium.org/2873813002/diff/1/pkg/front_end/lib/src/fasta...
File pkg/front_end/lib/src/fasta/type_inference/type_inferrer.dart (right):

https://codereview.chromium.org/2873813002/diff/1/pkg/front_end/lib/src/fasta...
pkg/front_end/lib/src/fasta/type_inference/type_inferrer.dart:414: /// is the
name fo the method.  [explicitTypeArguments] is the set of type
On 2017/05/09 23:45:57, scheglov wrote:
> /fo/of/s

Done.

https://codereview.chromium.org/2873813002/diff/1/pkg/front_end/lib/src/fasta...
pkg/front_end/lib/src/fasta/type_inference/type_inferrer.dart:432: //
inferConstructorInvocation?
On 2017/05/10 04:23:09, Leaf wrote:
> The specification is definitely intended to treat these equivalently.  That
is,
> a constructor call is just like a call to a static function with the
appropriate
> generic type signature.

Acknowledged.  The TODO comment is mostly because I wanted to get stuff working
before I was sure exactly what code should be shared.  I'll refactor this stuff
in a follow-up CL.

https://codereview.chromium.org/2873813002/diff/1/pkg/front_end/lib/src/fasta...
pkg/front_end/lib/src/fasta/type_inference/type_inferrer.dart:459:
Substitution.fromPairs(memberTypeParameters, inferredTypes);
On 2017/05/10 04:23:09, Leaf wrote:
> Is this right?  Looking at class Substitution from kernel, it doesn't look
like
> it does anything reasonable with a null inferredTypes.  Should this be
> explicitTypeArguments?

Whoops, you're right.  I wonder why this didn't cause any exceptions in my test
runs--I'll investigate and follow up with a fix.

https://codereview.chromium.org/2873813002/diff/1/pkg/front_end/lib/src/fasta...
pkg/front_end/lib/src/fasta/type_inference/type_inferrer.dart:587: FunctionType
_getCalleeFunctionType(DartType receiverType, Name methodName) {
On 2017/05/10 04:23:09, Leaf wrote:
> Surely this should be factored out somewhere?  Or does nothing else really
ever
> need to do this after inference has happened (presumably because the type is
> attached to the node or somesuch)?

Hmm, I'm not sure.  Kernel has a slot for interfaceTarget which I'm not yet
filling.  Presumably once I fill it, it will carry enough information so that
clients can derive whatever information they need.

I'll investigate and follow up.

https://codereview.chromium.org/2873813002/diff/1/pkg/front_end/lib/src/fasta...
pkg/front_end/lib/src/fasta/type_inference/type_inferrer.dart:628: //
TODO(paulberry): is there a kernel function that does this binary search?
On 2017/05/10 04:23:09, Leaf wrote:
> There's just got to be...otherwise something is very wrong.

I dug a little deeper and all I've found is see some code in
TypeCheckingVisitor.handleFunctionCall that does a linear search.  I'll make a
follow up CL that moves the binary search into kernel and refactors
TypeCheckingVisitor.handleFunctionCall to use it.