IDL: overload resolution for string and enumeration arguments

IDL: overload resolution for string and enumeration arguments

Pretty straightforward.
Main complexity is avoiding unnecessary tests related to automatic type
conversion.

I've also slightly restructured the test generation to make explicit
when we can only have one match (use |next()|), and when we can have multiple
matches (use |for|).
Ihis was previously an issue, and this makes it clear.

This lets us remove 2 of the 3 uses of [LegacyOverloadString]
(b/c now this is an ok overload)!
(Will remove last in followup, when handle optional args.)

Also cleanup IDL files:
* lets us order methods more naturally,
* align IDL order with spec
  (CanvasRenderingContext2D esp. was v. hard to read before)

R=haraken
BUG=293561
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=174012

[Jens Widell, 2014-05-14 05:19:30]

https://codereview.chromium.org/282873004/diff/1/Source/bindings/scripts/v8_i...
File Source/bindings/scripts/v8_interface.py (right):

https://codereview.chromium.org/282873004/diff/1/Source/bindings/scripts/v8_i...
Source/bindings/scripts/v8_interface.py:680: try:
You could skip the 'try' and "move" the controlled block into the preceding
'if', which checks for the same condition. That would avoid checking for a
string-or-enum argument twice using two different types of checks.

[haraken, 2014-05-14 06:09:03]

LGTM. (I'm assuming that you didn't make any behavior change to IDL files in
this CL.)

https://codereview.chromium.org/282873004/diff/1/Source/bindings/scripts/v8_i...
File Source/bindings/scripts/v8_interface.py (right):

https://codereview.chromium.org/282873004/diff/1/Source/bindings/scripts/v8_i...
Source/bindings/scripts/v8_interface.py:655: yield test, method

Why do you not need a try-catch block here?

https://codereview.chromium.org/282873004/diff/1/Source/bindings/scripts/v8_i...
Source/bindings/scripts/v8_interface.py:663: if any(idl_type.name == 'String' or
idl_type.is_enum

I wonder why you need to check string or enum here.

https://codereview.chromium.org/282873004/diff/1/Source/bindings/scripts/v8_i...
Source/bindings/scripts/v8_interface.py:690: # • a numeric type

I'm curious why the step 11 and 12 are separated. Aren't they doing the same
thing? (just generating "true"). Probably you're planning to make it different
in a follow-up.

[Nils Barth (inactive), 2014-05-14 06:28:15]

https://codereview.chromium.org/282873004/diff/1/Source/bindings/scripts/v8_i...
File Source/bindings/scripts/v8_interface.py (right):

https://codereview.chromium.org/282873004/diff/1/Source/bindings/scripts/v8_i...
Source/bindings/scripts/v8_interface.py:655: yield test, method
On 2014/05/14 06:09:03, haraken wrote:
> Why do you not need a try-catch block here?

Because this can return multiple matches.
Added comment to clarify.

https://codereview.chromium.org/282873004/diff/1/Source/bindings/scripts/v8_i...
Source/bindings/scripts/v8_interface.py:663: if any(idl_type.name == 'String' or
idl_type.is_enum
On 2014/05/14 06:09:03, haraken wrote:
> I wonder why you need to check string or enum here.

It's only needed if there's a match in step 11, otherwise it's redundant.
Added comment to address this.

https://codereview.chromium.org/282873004/diff/1/Source/bindings/scripts/v8_i...
Source/bindings/scripts/v8_interface.py:680: try:
On 2014/05/14 05:19:31, Jens Lindström wrote:
> You could skip the 'try' and "move" the controlled block into the preceding
> 'if', which checks for the same condition. That would avoid checking for a
> string-or-enum argument twice using two different types of checks.

While that seems simpler, I'm following the order in the spec;
the added |if| check above is to remove redundancy in the CG.
It gets trickier when we also have booleans.

https://codereview.chromium.org/282873004/diff/1/Source/bindings/scripts/v8_i...
Source/bindings/scripts/v8_interface.py:690: # • a numeric type
On 2014/05/14 06:09:03, haraken wrote:
> I'm curious why the step 11 and 12 are separated. Aren't they doing the same
> thing? (just generating "true").

Because they're different in the spec, and splitting makes a difference:
it means we *first* resolve to DOMString, *then* to numeric, *regardless*
of the order in the IDL file.
(This is to handle automatic type conversion.)

This is explained in the comment above:
# (Perform automatic type conversion, in order. If any of these match,
# that’s the end, and no other tests are needed.)

[Nils Barth (inactive), 2014-05-14 06:28:49]

On 2014/05/14 06:09:03, haraken wrote:
> LGTM. (I'm assuming that you didn't make any behavior change to IDL files in
> this CL.)

Nope, no behavior changes.
(Might get different error messages and need to rebase tests, but that's it.)

[Nils Barth (inactive), 2014-05-14 06:31:59]

Thanks for comments:
the handling of automatic type conversion during overload resolution is very
confusing, particularly b/c we want to avoid cruft in the CG, so I've added
comments to clarify; hope they help!

[Nils Barth (inactive), 2014-05-14 06:32:10]

The CQ bit was checked by nbarth@chromium.org

[commit-bot: I haz the power, 2014-05-14 06:32:18]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/nbarth@chromium.org/282873004/20001

[Nils Barth (inactive), 2014-05-14 07:36:06]

The CQ bit was checked by nbarth@chromium.org

[commit-bot: I haz the power, 2014-05-14 07:36:15]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/nbarth@chromium.org/282873004/40001

[commit-bot: I haz the power, 2014-05-14 09:38:01]

Change committed as 174012