ginormous performance issue

[Marco Martin]

On Friday 17 July 2009, Aaron J. Seigo wrote:
> On Friday 17 July 2009, Marco Martin wrote:
> > and still, in the case of an applet with many subwidgets, independent
> > timers can do really a big signal storm, so i still kinda like more the
> > pointer approach :p
>
> yes, having a single timer for this is a nice win .... the pointer thing is
> as bit of a hack (and yes, a pointer is always an int, so that's fine in
> that previous patch). the pointer bit would need to be stripped before
> actually putting it into the cache; i wonder if having a set "syntax" for
> it would make more sense, e.g. 
<id>:path_details_separated_by_underscores
> allowing the id to be stripped off.
>
> then the "don't cache this yet" collection could be a map of ids to
> entries.
>
> really, this is just a way around adding API. i wonder if it wouldn't make
> more sense to just add a new addToCache method in Theme that takes an
> additional "id" argument for this purpose. avoids all the string parsing
> and reliance on "well formed" entries and makes it clear what's going on
> internally.
>
> i'll think more on this over the weekend (back to work on monday for me!
> wee!
>
> :)) but the more i look at the various patches the more i think that a new
>
> method is the cleanest approach. thoughts?
+1 from me for the new method
so the version without id will insert immediately into the cache, the version 
with id will be delayed, that will need to be well documented

thinking about it besides the pointer as id we need also in case of Svg the 
element id (if any) and in FrameSvg the prefix(if any)
it's fine just to concatenate the two things in  a single string i think

tomorrow i will give a try (buaaah, /me want giiit:p)

-- 
Marco Martin