Review Request: KJS: Implement Object.prototypeOf, Object.keys, Object.getOwnPropertyNames

[Bernd Buschinski]

> On April 10, 2012, 2:38 a.m., Maks Orlovich wrote:
> > kjs/object.h, line 431
> > <http://git.reviewboard.kde.org/r/104511/diff/1/?file=56158#file56158line431>
> >
> >     It's OK to make API changes here; just make sure you fix every call site. Frankly, past experience shows that overloaded virtual groups like this are a mistake and hard to get right (there have been bugs caused with the getters, for example).
> >

Oh good :)
I would love to clean it, I was told that I should avoid API changes, but if you are ok with it, I will do it.


> On April 10, 2012, 2:38 a.m., Maks Orlovich wrote:
> > kjs/object_object.cpp, line 241
> > <http://git.reviewboard.kde.org/r/104511/diff/1/?file=56160#file56160line241>
> >
> >     Faster: save return value of getObject() and check that for null. (Note that it matters here..)

done


> On April 10, 2012, 2:38 a.m., Maks Orlovich wrote:
> > kjs/object_object.cpp, line 251
> > <http://git.reviewboard.kde.org/r/104511/diff/1/?file=56160#file56160line251>
> >
> >     So this one shows the DontEnum ones? Joy.

Yes that also includes the DontEnum ones. Is this just comment or is there something I overlook? (As this is an Issue that stops the "Ship it!")


> On April 10, 2012, 2:38 a.m., Maks Orlovich wrote:
> > kjs/object.cpp, line 496
> > <http://git.reviewboard.kde.org/r/104511/diff/1/?file=56159#file56159line496>
> >
> >     This might actually show up on some benchmarks. I bet you can speed this up by making an appropriate choice of value for PropertyMap::IncludeDontEnumProperties -- basically making it a mask on attr, but that may be a bit brittle maintenance wise

uhm really? for maintenance I would really say I am against it. And I doubt that it would really show up on some benchmarks... unless you are benchmarking huge objects/arrays with millions of properties


> On April 10, 2012, 2:38 a.m., Maks Orlovich wrote:
> > kjs/function.cpp, line 301
> > <http://git.reviewboard.kde.org/r/104511/diff/1/?file=56157#file56157line301>
> >
> >     How should this react when the length property is altered?
> >

good question, I don't know, not explicitly covered as I can see. But I guess it would make sense that it shows 0 to length-1 parameters, if the they are all mapped.


> On April 10, 2012, 2:38 a.m., Maks Orlovich wrote:
> > kjs/function.cpp, line 308
> > <http://git.reviewboard.kde.org/r/104511/diff/1/?file=56157#file56157line308>
> >
> >     Why convert to ident when isMapped will convert back to number again? Just add  aversion that takes a number; similarly [] works on numbers already. (And perhaps fix up some of the sloppiness with signs there)

Nononono!, I figured out that I have to be very careful with the number versions. The problem is the numbers have attributes too. If I overwrite it, it will only return the properties of the mapped object. In other places I need the properties for the numbers too. (defineOwnProperty for example)
So I must not add a number version.


- Bernd


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://git.reviewboard.kde.org/r/104511/#review12281
-----------------------------------------------------------


On April 8, 2012, 9:14 p.m., Bernd Buschinski wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> http://git.reviewboard.kde.org/r/104511/
> -----------------------------------------------------------
> 
> (Updated April 8, 2012, 9:14 p.m.)
> 
> 
> Review request for kdelibs.
> 
> 
> Description
> -------
> 
> KJS: Implement Object.prototypeOf, Object.keys, Object.getOwnPropertyNames
> 
> NOTE: Array was left out on purpose, as currentl imeplementation does not save attributes (next patch will fix this)
> 
> keys&GetOwnPropertyNames requieres to change the JSObject::getOwnPropertyNames implementation,
> for future use a enum is better than than a bool, maybe there will be more ways to include/exclude properties.
> 
> All changes for khtml/ecma/ are to silense the -Woverloaded-virtual warnings
> 
> 
> Diffs
> -----
> 
>   khtml/ecma/kjs_css.h aba44b8 
>   khtml/ecma/kjs_css.cpp e3e7417 
>   khtml/ecma/kjs_dom.h d0433c3 
>   khtml/ecma/kjs_dom.cpp 5fff7e3 
>   khtml/ecma/kjs_html.h 0f3f544c 
>   khtml/ecma/kjs_html.cpp e3da95c 
>   khtml/ecma/kjs_scriptable.h af5343c 
>   khtml/ecma/kjs_scriptable.cpp 5d4ea68 
>   kjs/JSVariableObject.h a8f01eb 
>   kjs/JSVariableObject.cpp b00ef76 
>   kjs/array_instance.h 3f2b630 
>   kjs/function.h 3757fe8 
>   kjs/function.cpp 5f39ae6 
>   kjs/object.h 047c242 
>   kjs/object.cpp c19122f 
>   kjs/object_object.cpp 986f03f 
>   kjs/property_map.h 6b127ff 
>   kjs/property_map.cpp b2ff08e 
>   kjs/string_object.h e890d52 
>   kjs/string_object.cpp 170e2f7 
> 
> Diff: http://git.reviewboard.kde.org/r/104511/diff/
> 
> 
> Testing
> -------
> 
> ecma script & daily surfing
> 
> 
> Thanks,
> 
> Bernd Buschinski
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kde-core-devel/attachments/20120413/e19c5db8/attachment.htm>