648355 - Function.prototype.isGenerator

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Dave Herman [:dherman]]

Function.isGenerator(f) -> boolean

Indicates whether a given function is a generator function. As specified on the attached TC39 wiki page.

Should be easy -- I'll try to do today.

Dave

Comment 1

[Dave Herman [:dherman]]

Attachment: Function.isGenerator

This version tests for generator functions by testing whether a function a) is interpreted and b) starts its script with JSOP_GENERATOR. This seems a bit brittle. Is it worth nabbing a JSFUN_XXX bit and setting that on the JSScript flags?

Dave

Comment 2

[Dave Herman [:dherman]]

Brendan: what was the reason we couldn't put isGenerator on Function.prototype as a (non-enumerable) zero-argument self-test? Does it create web incompatibilities?

Dave

Comment 3

[Dave Herman [:dherman]]

Attachment: moved to Function.prototype, added tests

Ready for review.

Dave

Comment 4

[Brendan Eich [:brendan]]

Comment on attachment 524713 [details] [diff] [review]
moved to Function.prototype, added tests

>From: Dave Herman <[e-mail]>
>Function.isGenerator (bug 648355, r=?)
>
>diff --git a/js/src/jsfun.cpp b/js/src/jsfun.cpp
>--- a/js/src/jsfun.cpp
>+++ b/js/src/jsfun.cpp
>@@ -2331,16 +2331,39 @@ CallOrConstructBoundFunction(JSContext *
>         return false;
> 
>     *vp = args.rval();
>     return true;
> }
> 
> }
> 
>+    bool rval = false;

Nit: rval is a canonical name for a jsval or js::Value local. Suggest result or answer instead.

>+    if (fun->isInterpreted()) {
>+        JSScript *script = fun->u.i.script;
>+        rval = script->length > 0 && script->code[0] == JSOP_GENERATOR;

Nit: prevailing style prefers != 0 for unsigned tests, not > 0, but:

Non-nit: we don't need a length check here, scripts by definition have at least JSOP_STOP as a bytecode. So you could get rid of the script local, or keep it and assert that length != 0.

>+script regress-648355.js

I'm glad to have a regression test but this space in the suite is for actual bugs we don't want to reopen, not features being added and covered. IIRC from discussion with Waldo, something like tests/js1_8_5/extensions/is-generator.js would be better.

Best to update the strawman at the bug's URL to match what this patch implements.

r=me with these addressed; happy to look again if you like.

/be

Comment 5

[Dave Herman [:dherman]]

http://hg.mozilla.org/tracemonkey/rev/90fa43dcb844

Comment 6

[Dave Herman [:dherman]]

Waldo caught a missing #if/#endif. Pushed here:

http://hg.mozilla.org/tracemonkey/rev/dbecfa435101

Comment 7

[Chris Leary [:cdleary] (not checking bugmail)]

http://hg.mozilla.org/mozilla-central/rev/90fa43dcb844
http://hg.mozilla.org/mozilla-central/rev/dbecfa435101

Comment 8

[Eric Shepherd [:sheppy]]

Now documented:

https://developer.mozilla.org/en/JavaScript/Reference/Global_Objects/Function/isGenerator

With a link from the Function page here:

https://developer.mozilla.org/en/JavaScript/Reference/Global_Objects/Function#Methods_2

And mentioned on Firefox 5 for developers.