Bug 127237

Summary: No prototype for elements
Product: [Applications] konqueror Reporter: jordan.osete
Component: kjsAssignee: Konqueror Developers <konq-bugs>
Severity: normal CC: maksim
Priority: NOR    
Version: unspecified   
Target Milestone: ---   
Platform: unspecified   
OS: Linux   
Latest Commit: Version Fixed In:

Description jordan.osete 2006-05-13 10:28:30 UTC
Version:           3.5.2 (using KDE 3.5.2, Debian Package 4:3.5.2-2+b1 (testing/unstable))
Compiler:          Target: i486-linux-gnu
OS:                Linux (i686) release 2.6.15-1-k7

Actually, it is between bug and feature request. The default elements of a document (like HTMLElement, Element, etc.) should have an extendable prototype. If you add functions to the prototype, it should be added to elements. A very useful example (works in FF & Opera):

HTMLElement.prototype.hasClass=function (className){
    return true && this.className.match(RegExp('\\b'+className+'\\b'));}
HTMLElement.prototype.delClass=function (className){
    var c=this.className.replace(RegExp('\\b'+className+'\\b'), '');
    var ret=(c!=this.className);
    if(ret)this.className=c.replace(/ {1,}/, ' ');
    return ret;}
HTMLElement.prototype.addClass=function (className){
        return false;
        this.className=this.className+' '+className;
        this.className=this.className.replace(/ {1,}/, ' ');
        return true;}}

Also note (related to bug 127236) that it should also work for events handlers, like that:
        alert ("image loaded, url: "+this.src);}
No browser currently does that.
Actually the best way to do it all would be to resolve the event handler only when it is called. Ie.: on click, check if the item has an onclick member, then if not, check if its prototype has one.
Comment 1 Maksim Orlovich 2006-05-13 16:02:18 UTC
I might add the mozillaish constructor object thing, but I can't think of any sensible way for the later construct to work or any reason for it to work the way you want it --- all this is setting is a fallback to a dynamic property, which is always there, just sometimes null, and hence is a no-op
Comment 2 Maksim Orlovich 2006-07-03 00:41:38 UTC
On Sunday 02 July 2006 18:11, Maks Orlovich wrote:
> SVN commit 557310 by orlovich:

Forgot to forward it to bugs... And note that this will probably need some 
additions to support everything, but they should be pretty easy...

> A bunch of DOM changes we'll need to support the new yahoo mail
> and ATLAS stuff; this is neccessary but not sufficient, since they
> also heavily use setters/getters, which I am not sure we want to support in
> 3.5.x... (But this stuff is standalone)
> 1. Provide some emulation of mozillaisms like Node.prototype = ... .
> These are incomplete, but should hopefully cover the important stuff.
> Related to that, fix HTMLDocument to use a proper prototype.
> 2. Merge from Apple's tree: support the no-op debugger; statement,
> it at least seems to be used by yahoo...
> 3. Readonly support for characterSet property.
> 4. Fix crash on reading some properties of mouse events that weren't
> dispatched --- noticed on testing this stuff.
>  M  +49 -0     khtml/ecma/kjs_binding.h
>  M  +2 -0      khtml/ecma/kjs_css.cpp
>  M  +2 -0      khtml/ecma/kjs_css.h
>  M  +30 -74    khtml/ecma/kjs_dom.cpp
>  M  +8 -12     khtml/ecma/kjs_dom.h
>  M  +33 -51    khtml/ecma/kjs_events.cpp
>  M  +1 -11     khtml/ecma/kjs_events.h
>  M  +20 -21    khtml/ecma/kjs_html.cpp
>  M  +2 -0      khtml/ecma/kjs_html.h
>  M  +20 -2     khtml/ecma/kjs_window.cpp
>  M  +3 -1      khtml/ecma/kjs_window.h
>  M  +864 -841  kjs/grammar.cpp
>  M  +57 -55    kjs/grammar.h
>  M  +13 -0     kjs/grammar.y
>  M  +1 -1      kjs/keywords.table
>  M  +19 -1     kjs/lookup.h

Comment 3 jordan.osete 2006-07-03 09:46:27 UTC
Thank you.
Actually, though prototypes are really useful, event inheritance might not be that much needed. I only proposed this because if it is not implemented, there can be some inconsistenties, for example if you try to define some event handler for anchors:
  HTMLAnchorElement.prototype.onclick=function () {alert(this.href);};
It will not be activated when the user clicks on any link, BUT it will be activated if you try to simulate a click by doing, for example:

Well, that is not that bad, just wanted to point the case.
Comment 4 Maksim Orlovich 2006-07-03 15:12:08 UTC
As I said above, that does not make sense, since onclick is a dynamic property. Also note that we do not have pseudo-constructors for subclasses of HTMLElement.