The plan for what to do is here:
http://jira.openlaszlo.org/jira/browse/LPP-6617
In general, I think we should be taking more and more of the transformations
that the tag compiler does out. Many of them were created to give the
impression that the runtime was smarter than it really was, but they can only
work on constant initial values, which leads to this whacky dichotomy where you
can have a fancy initial value, but nothing works at runtime. E.g., it used
to be that you could use any CSS color spec at init time, but could only use a
numeric color at run time. We've kludged around that by having the color
setters invoke the presentation-type parser (and by making the
presentation-type parser accept non-string representations by passing them
straight through). We have a similarly broken situation in the CSS type, e.g.,
for the layout attribute. At compile/init time you set the layout using a CSS
spec like "axis: y; spacing: 5", whereas at runtime you have to use a hash like
"{axis: 'y', spacing: 5}". It's pretty messy. We should be cleaning these!
things up.
On 2010-09-11, at 16:54, Henry Minsky wrote:
> This might be our opportunity also to remove the kludgy HTML massaging we do
> for <text> view, and
> just pass the user's XML content verbatim (maybe with whitespace normalized
> out), to the runtime?
>
>
>
>
> On Fri, Sep 10, 2010 at 6:43 PM, P T Withington <[email protected]> wrote:
>
>> That would be great! Seems like a big waste if it is all for nothing.
>>
>> On 2010-09-10, at 18:27, Henry Minsky wrote:
>>
>>> That was supposed to propagate the font style info at compile time where
>>> possible as an optimization (and to compute text widths for flash 5 input
>>> text fields). I don't know if it even is working at all. Now that Max
>> put
>>> in runtime inheritance of font properties, I am pretty sure we can get
>> rid
>>> of it. I will see what happens if I remove
>>> it...
>>>
>>>
>>>
>>> On Fri, Sep 10, 2010 at 6:12 PM, P T Withington <[email protected]> wrote:
>>>
>>>> I'm trying to remodularize some stuff in the schema to pass around
>>>> classModels everwhere (instead of tagnames that get looked up over and
>> over)
>>>> and this is the only caller of `schema.getBaseClassname`. If that is
>> all
>>>> crud, I'd like to rip it out.
>>>
>>>
>>>
>>>
>>> --
>>> Henry Minsky
>>> Software Architect
>>> [email protected]
>>
>>
>
>
> --
> Henry Minsky
> Software Architect
> [email protected]
