Planning: Rendering backend

[memononen]

Since adding the text decorations, the rendering has gotten a bit more complicated. There's few more cases in the issues (e.g. inline objects), which will likely complicate things a bit further.

I'm thinking of adding a rendering backend that will simplify some of the rendering nuances. This would also fix the following issues: #14 #54 .

As part of the change I'm also thinking of renaming the existing "renderer" to "rasterizer":

• skb_renderer_t -> skb_rasterizer_t
• skb_render_cache_t -> skb_raster_cache_t
  And the backend would ne called renderer.

I dont know the full details yet, but the scope is more or less to lift out some parts of the debug_draw, the rendering loops from the examples, and add culling (maybe clipping too).

The backend will be optional, the current API will still be working.

If there are particular things that I should keep in mind, please let me know. I'm especially interested how to handle mixed rendering case, where you have 2d/3d rendering engine/system and want to integrate skribidi to it.

[memononen]

Took firsts steps towards nicer renderer in: 1f2ac0a

It made me think if we really need a renderer, as I was able to make the API simpler in earlier commits. I expect most use cases to have their own renderer where skribidi interfaces with. So I think making that integration simpler is far more important than having a wide range of backends.

I recently changed the skb_quad_t so that it can be used to hold the intermediate data. So it would be possible to have generic skb_layout_t -> array of skb_quad_t function. With things like culling that function might get more complicated over time.

I have a couple of things I want to still experiment before I make any calls:

1. add culling for render_draw_layout() it should output only (roughly) visible quads and request only visible glyphs from image atlas
2. out of band styling there are plenty of use cases, like hyperlinks, where we have multiple styles to render depending on some state Clickable runs #50, or maybe the user even wants to add custom effects, like outlines and shadows. It can be as simpler as few pieces of user data, but it needs to be tried out