Re: [Dynapi-Dev] (no subject)

[Robert R. <rai...@us...>]

There's also a bit of inconsistency with making the children array
null.   Because once you remove all the children, the array is not
null, but has a length of 0.  So, in your widget, you will have to
check for a null array and a array of length 0.  I don't understand
why the array can't be initialized to an empty array.


Later,
Robert <rai...@us...>

> I think I've found a fix for the problem with children arrays being shared
> between widgets made with the widgetX  model.
> 
> The problem:
> 
> Widget.prototype = new DynLayer()
> 
> When this is done, the properties of the base DynLayer are *shared* with
> every Widget made with the constructor.
> 
> Most of these shared properties are reassigned to unique ones when
> createElement is called on the Widget.  However, the children array is not,
> and it gives the effect that all Widgets share the same children array.
> 
> So I've posted a patch for DynLayer and DynDocument that assigns the
> this.children array only when you use addChild.  DynLayer-based widgets
> won't have a children array until you add at least one child.
> 
> I also made changes to removeChild, removeAllChildren, etc. that check for
> the existence of the child array before proceeding.
> 
> This should ensure that the children array is a unique object for each
> widget, and not the common one in the Widget prototype.  Then you can safely
> use WidgetX to simulate inheritance without the shared-properties problem.
> 
> The only consideration is if you build a widget that manipulates the
> children array, you should test it it exists first (it's initially set to
> null until you use addChild).
> 
> I've only tested it on a few IBS and GUI widgets, but it seems to work
> across the board.  I've posted the patch for review.
> 
> scottandrew
> 
> 
> 
> 
> 
> 
> ------
> Scott Andrew LePera
> DHTML / Scripting / CGI and other neat stuff
> sc...@sc...
> http://www.scottandrew.com