Tucker - thanks for writing this up. Without carefully following the
various threads it was difficult to know what exactly had been
proposed. A summary like this is vital for us all to be on the same
page.
And - I like the proposal! Once there's consensus, can you make sure
there's bug coverage for the various tasks involved?
jim
On May 16, 2006, at 10:03 AM, P T Withington wrote:
> Henry and I are proposing:
>
> 1) Use a sentinel rather than null for declared events. This is
> typesafe, will not crash Javascript, and can be optimized as
> necessary.
>
> 2) setAttribute, which is the primary caller of sendEvent and
> likely to be impacted performance-wise, will be re-written to be
> careful about undeclared events and to not make function calls to
> the sentinel event. (Safe, will not crash, can be optimized.)
>
> 3) The compiler will inline setAttribute and sendEvent. (It used
> to inline setAttribute, that was turned off because it conflicted
> with registers, is fairly easy to reintroduce. It should be easy
> to inline enough of sendEvent to avoid unnecessary function calls.)
>
> On 2006-05-16, at 12:43 EDT, Jim Grandy wrote:
>
>> On May 16, 2006, at 7:42 AM, Amy Muntz wrote:
>>
>>> Task 2
>>> Then, please use the DeclareEvent pattern, and remove the if's as
>>> you no longer have to check for null. Please do this in the same
>>> set of files you are currently working with, and then send that
>>> out for review (Henry has volunteered to review this next
>>> changeset - thanks, Henry!)
>>
>> Woah. The DeclareEvent isn't where the if's are -- it's in the
>> sendEvent idiom we use. We'd need a SendEvent standalone function
>> to get rid of the if's.
>>
>> SendEvent = function(evt, arg) {
>> if (evt != null) evt.sendEvent(arg);
>> }
>>
>> Then we can rewrite the body of SendEvent if the semantics change.
>>
>> Or am I confused? Can someone please write up a mini-proposal for
>> what we're doing?
>>
>> Thanks,
>>
>> jim
>
_______________________________________________
Laszlo-dev mailing list
[email protected]
http://www.openlaszlo.org/mailman/listinfo/laszlo-dev
