Re: [Gnash-dev] Questions about trapezoids, triangle stripper etc.

[strk]

On Mon, Sep 25, 2006 at 06:23:32PM +0200, Udo Giacomozzi wrote:

> I skip the cache part until basic rendering works. However, how do you
> suggest the cache to be stored? It really seems to be the best if it
> is stored in the *definition* of the character, but which would be a
> clean way to store cache objects? I'd like to let the renderer decide
> how the cache should look like. But I don't like to put a simple void
> pointer there...

I faced the same problem with bitmap_info.

Basically, bitmap_info are associated with bitmap_character_def, which
is read at parse time.
The problem is that bitmap_info are created trough a call to the
renderer, so, like meshsets, should be seen as renderer cache instead.
I think that leaving the bitmap_character_def ignorant about the cache
is a good thing, and we could pass bitmap_character_def references
whenever needed (in fill styles for example, but also when calling
the renderer to draw them). In turn, the renderer, whenever gets
to a bitmap_character_def to be drawn, shoudl chekc wheter a cache
item is already associated with it, and if not create one and associate
it.

I guess the same applied to mesh sets and friends.

About the void ptr, a possible implmentation might be defining an abstract
class called RendererCache and store a pointer to it in the definition
classes that allow some form of caching (like bitmap_character_Def or
shape or whatever). The renderer would then define a subclass for it
and dynamic_cast back when getting the cache value.

What do you think ?

--strk;