I see. One of the leading performance issues Henry and i saw from profiling 
swf10 looked to be method closures, at lest from a gc/memory churn standpoint.  
Approved for now!

On Nov 12, 2010, at 6:11 AM, P T Withington <[email protected]> wrote:

> The line you are complaining about is unchanged from the original code.  What 
> is it that is troubling about it?
> 
> Only as3 supports method closures, so the only way you could avoid saying 
> `context[curvemethod]` would be to duplicate the code.  At this point in 
> time, it seems better to keep a single implementation.  If it turns out this 
> is a performance bottleneck, we can change it later.
> 
> On 2010-11-12, at 04:56, Max Carlson wrote:
> 
>> This line is troubling: 
>>   var curvemethod = context['curveTo'] ? 'curveTo' : 'quadraticCurveTo'. 
>> 
>> Can you look up the method reference once ahead of time instead of computing 
>> its name and looking it up, e.g. context[curvemethod]?
>> Perhaps use an if ($dhtml) {...} else {...} clause that stores a reference 
>> to the method once?
>> 
>> Otherwise approved!
>> 
>> On 11/11/10 1:35 PM, P T Withington wrote:
>>> Change ptw-20101111-LWK by [email protected] on 2010-11-11 16:26:03 EST
>>>    in /Users/ptw/OpenLaszlo/trunk-3
>>>    for http://svn.openlaszlo.org/openlaszlo/trunk
>>> 
>>> Summary: Generalize rectangle drawing to support oval borders, CSS 
>>> defaulting
>>> 
>>> Bugs Fixed: LPP-9484 Strokes in drawbutton shoulld be stylable (partial)
>>> 
>>> Technical Reviewer: [email protected] (pending)
>>> QA Reviewer: [email protected] (pending)
>>> 
>>> Details:
>>> 
>>>    Made a new method, roundrect, that takes separate horizontal and
>>>    vertical radii for the corners.  All radii must be supplied,
>>>    defaulting should be done at a higher level.  Uses the CSS
>>>    algorithm for scaling radii to fit small boxes.  This low-level
>>>    method is needed to be able to draw different width borders that
>>>    smoothly transition around corners.
>>> 
>>>    Made rect default its radii arguments according to the CSS
>>>    defaulting algorithm and call roundrect.
>>> 
>>> Tests:
>>>    smokecheck, is there a test for rect?
>>> 
>>> Files:
>>> M       WEB-INF/lps/lfc/kernel/LzKernelUtils.lzs
>>> 
>>> Changeset: http://svn.openlaszlo.org/openlaszlo/patches/ptw-20101111-LWK.tar
> 

Reply via email to