IndexedDB: Use [Default] in IDLs to reduce overloads in implementations

IndexedDB: Use [Default] in IDLs to reduce overloads in implementations

By default, if the IDL for interface XXX has foo(optional int x) then
the implementation must have XXX::foo() and XXX::foo(int x). Adding
[Default=Undefined] lets the implementation drop the overload and (the
coerced value of) |undefined| will be passed in. Use this in IndexedDB
IDLs to reduce the overload clutter in implementations.

This does allow add() and put() to be called with |undefined| as a
literal second element which is ignored, but there is gathering
consensus in the webapps group that this is desirable to match ES6
semantics. The linked bug tracks making all optional arguments accept
|undefined|.

BUG=163959
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=148581

[jsbell, 2013-04-16 21:38:04]

dgrogan@, alecflett@ - what do you think?

[dgrogan, 2013-04-16 22:47:02]

On 2013/04/16 21:38:04, jsbell wrote:
> dgrogan@, alecflett@ - what do you think?

What are ES6 semantics?

[dgrogan, 2013-04-16 22:47:53]

Also, a cursory search left me unable to find threads about the undefined / ES6
issue. Do you have any links handy?

[jsbell, 2013-04-16 23:00:16]

On 2013/04/16 22:47:53, dgrogan wrote:
> Also, a cursory search left me unable to find threads about the undefined /
ES6
> issue. Do you have any links handy?

The big thread starts here:
http://lists.w3.org/Archives/Public/public-webapps/2012OctDec/0085.html

In short, ES6 is adding default parameters, so you can write:

function f(arg1 = "foo", arg2 = "bar") { ... }

The question was raised: what happens if you call this as:

f("first", undefined);

Should arg2 be "bar" or |undefined| ? After some pontificating, TC39 settled on
"bar", i.e. if you pass an explicit |undefined| value then the default is used.
This also allows:

f(undefined, "second");

WebIDL *currently* specifies that if you call with an explicit |undefined|
argument you get funky coercion (e.g. a for an optional DOMString argument you'd
get a string-literal "undefined"), but it's on heycam@'s list to change that.

[dgrogan, 2013-04-16 23:40:22]

lgtm

[alecflett, 2013-04-17 00:01:49]

lgtm

yay!

[jsbell, 2013-04-17 00:11:53]

haraken@ - could you take a look?

[haraken, 2013-04-17 00:35:58]

LGTM

[jsbell, 2013-04-17 21:35:28]

Committed patchset #3 manually as r148581 (presubmit successful).