Add ES5 Object.isExtensible and Object.preventExtensions.

Add ES5 Object.isExtensible and Object.preventExtensions.


Committed: http://code.google.com/p/v8/source/detail?r=5011

[Erik Corry, 2010-07-01 14:16:19]

LGTM

http://codereview.chromium.org/2819034/diff/1/4
File src/objects.cc (right):

http://codereview.chromium.org/2819034/diff/1/4#newcode2585
src/objects.cc:2585: // If there is fast elements we normalize.
is -> are

http://codereview.chromium.org/2819034/diff/1/4#newcode2596
src/objects.cc:2596: Map::cast(new_map)->set_is_extensible(false);
This is going to make a lot of maps.  We need a set of slow maps (one for each
flag combination) and then just pick the right one.  This will benefit other
parts of the system too.  No need to put it in this change though.

http://codereview.chromium.org/2819034/diff/1/4#newcode6237
src/objects.cc:6237: // the object into dictionary mode (and force it to stay
there).
Not the whole object, just the elements, right?

http://codereview.chromium.org/2819034/diff/1/8
File src/v8natives.js (right):

http://codereview.chromium.org/2819034/diff/1/8#newcode762
src/v8natives.js:762: !IS_UNDETECTABLE(obj)) {
You can't even ask an undetectable object whether it is extensible?  Deep
undercover...

http://codereview.chromium.org/2819034/diff/1/10
File test/mjsunit/object-prevent-extensions.js (right):

http://codereview.chromium.org/2819034/diff/1/10#newcode115
test/mjsunit/object-prevent-extensions.js:115: assertUnreachable
We don't allow semicolons to be omitted.

[Rico, 2010-07-02 09:16:11]

http://codereview.chromium.org/2819034/diff/1/4
File src/objects.cc (right):

http://codereview.chromium.org/2819034/diff/1/4#newcode2585
src/objects.cc:2585: // If there is fast elements we normalize.
On 2010/07/01 14:16:19, Erik Corry wrote:
> is -> are

Done.

http://codereview.chromium.org/2819034/diff/1/4#newcode2596
src/objects.cc:2596: Map::cast(new_map)->set_is_extensible(false);
On 2010/07/01 14:16:19, Erik Corry wrote:
> This is going to make a lot of maps.  We need a set of slow maps (one for each
> flag combination) and then just pick the right one.  This will benefit other
> parts of the system too.  No need to put it in this change though.
Good point, I guess we just need one for each flag, not every flag combination?

http://codereview.chromium.org/2819034/diff/1/4#newcode6237
src/objects.cc:6237: // the object into dictionary mode (and force it to stay
there).
On 2010/07/01 14:16:19, Erik Corry wrote:
> Not the whole object, just the elements, right?
Yes, corrected.

http://codereview.chromium.org/2819034/diff/1/8
File src/v8natives.js (right):

http://codereview.chromium.org/2819034/diff/1/8#newcode762
src/v8natives.js:762: !IS_UNDETECTABLE(obj)) {
On 2010/07/01 14:16:19, Erik Corry wrote:
> You can't even ask an undetectable object whether it is extensible?  Deep
> undercover...
I guess we could allow this - do we even allow people to add properties to
undetectable objects?

http://codereview.chromium.org/2819034/diff/1/10
File test/mjsunit/object-prevent-extensions.js (right):

http://codereview.chromium.org/2819034/diff/1/10#newcode115
test/mjsunit/object-prevent-extensions.js:115: assertUnreachable
On 2010/07/01 14:16:19, Erik Corry wrote:
> We don't allow semicolons to be omitted.

Done + all other places (I guess that the tests not failing is a good indication
that we did not hid this :-) )

[Mads Ager (chromium), 2010-07-02 09:21:01]

http://codereview.chromium.org/2819034/diff/1/8
File src/v8natives.js (right):

http://codereview.chromium.org/2819034/diff/1/8#newcode762
src/v8natives.js:762: !IS_UNDETECTABLE(obj)) {
On 2010/07/02 09:16:11, Rico wrote:
> On 2010/07/01 14:16:19, Erik Corry wrote:
> > You can't even ask an undetectable object whether it is extensible?  Deep
> > undercover...
> I guess we could allow this - do we even allow people to add properties to
> undetectable objects?
> 

Undetectable objects behave in a special way only in a few situation (such as
conversion to boolean). You can add and remove properties to undetectable
objects and extensible should be orthogonal to undetectability. You should allow
this for undetectable objects and we should have a test case for undetectable
objects. :)

[Rico, 2010-07-02 12:57:21]

http://codereview.chromium.org/2819034/diff/1/8
File src/v8natives.js (right):

http://codereview.chromium.org/2819034/diff/1/8#newcode762
src/v8natives.js:762: !IS_UNDETECTABLE(obj)) {
On 2010/07/02 09:21:01, Mads Ager wrote:
> On 2010/07/02 09:16:11, Rico wrote:
> > On 2010/07/01 14:16:19, Erik Corry wrote:
> > > You can't even ask an undetectable object whether it is extensible?  Deep
> > > undercover...
> > I guess we could allow this - do we even allow people to add properties to
> > undetectable objects?
> > 
> 
> Undetectable objects behave in a special way only in a few situation (such as
> conversion to boolean). You can add and remove properties to undetectable
> objects and extensible should be orthogonal to undetectability. You should
allow
> this for undetectable objects and we should have a test case for undetectable
> objects. :)
> 

As discussed offline this is actually correct behavior, since we explicitly do
not throw an exception if this is undetectable. This is, however, a pretty good
indication that this is not very readable. I will make another change that
cleans this up for all methods in v8natives.js and runtime.js (where there is
actually a bug because of this).

I added a test for undetectable objects and these two methods. I will add a bug
for creating tests for the remaining methods and do this when I return from
vacation.

[Mads Ager (chromium), 2010-07-02 13:55:23]

LGTM