Re: [AGG] SpanGenerators

[Ivan N. <in...@pd...>]

Maxim, 

What I am suggesting is to allow for some kind of integration of the
compound rasterizer with other AGG pieces (e.g. other span rasterizers,
scan-bool clipping, etc). 

For example, since compound rasterizer generates scanlines, it may be a good
idea to wrap the compound rasterizer in a new SpanGenerator? Then the
compound rasterizer could be used to fill spans generated using
rasterizer_aa, rasterizer_aa_solid, or 'scan-bool' shapes (e.g. by caching
the current scanline to fill all the spans generated by the span generator,
then moving to the next scanline, etc). This way it would be possible to use
compound rasterizer together with other rasterizers, clippers,
SpanGenerators, etc. 

What do you think?

Ivan

-----Original Message-----
Message: 7
Date: Wed, 26 Jul 2006 10:19:28 -0700 (PDT)
From: Maxim Shemanarev <mcs...@ya...>
Subject: Re: [AGG] SpanGenerators
To: Anti-Grain Geometry <vec...@li...>
Message-ID: <200...@we...>
Content-Type: text/plain; charset=iso-8859-1

> I am doing some tests with the new layered compound rasterizer. Would it
be
> possible to use this rasterizer in conjunction with other AGG pieces? In
> particular, I am not completely clear on how to clip the rendered output
> against a 'scan-bool' region. sbool_combine_shapes_aa expects the last
> parameter to be a 'renderer', but currently there is no compound renderer
> (analogous to renderer_scanline_aa) and I am not sure how to write one
(e.g.
> render_scanlines_compound_layered accepts rasterizer). 

Well, it's more complex. The problem is that a scanline is in general
multi-color. So, render_scanlines_compound_layered pulls scanlines that
correspond to certain styles and composes a combined multi-color span. A
single
span may consist of may scanlines, one per style. So, there's no way to
collect
them in a scanline storage. It's technically possible to apply some boolean
operation after "if(ras.sweep_scanline(sl_aa, i))" in this function, but I'm
afraid the design may become too complex. Besides, scanline boolean operates
with scanline shapes, not separate scanlines, so, it also requires some
redesign and it's not that obvious how to do that. 

As an alternative for the existing version you can use alpha-masking;
agg_pixfmt_amask_adaptor.h will work, but it's restricted. 

BTW, there's an idea. It's technically possible to perform any boolean
and/or
arithmetic compositing operations directly in
render_scanlines_compound_layered. Now I use only rgba8::add(), which
performs
"geometry flattening", or just "multicolor union" operation. What we need
is:
1. A possibility to substitute the logical/compositing operation.
2. A possibility to specify the operation for each layer.

That would provide all the Constructive Solid Geometry functionality (and
even
more) done in the rasterizer, but it requires serious redesign of some
parts,
like color compositing.

McSeem