tritania / sdl_svg Goto Github PK

------------------------------------------------------------------------------
David Ashley (DA) Sun Feb 20 20:25:32 PST 2005
[email protected]

To build do
./configure
make
make install

It might work. It's my first attempt at the automake stuff.

The ./clean.sh can be used to wipe out everything, but then you need to run
./autogen.sh
to recreate the configure script.

There usually are other targets like
make clean
and
make distclean

The svgtest program can be used to load and display an SVG file.
------------------------------------------------------------------------------
David Ashley Sun Jan  9 11:33:33 PST 2005
[email protected]

This code was taken from the SDL_svg project hosted by sourceforge:
http://sourceforge.net/projects/sdlsvg/
That project was created by Enzo <[email protected]>
I also pulled code from the cairo project (specifically the libsvg tree).
Putting here means there's at least a known good copy that works properly with
SDL_svg. As that project evolves the embedded copy here could be updated.

Enzo's SDL_svg project had what I consider to be fundamental flaws in its
architecture. I didn't like the dependency on SDL_gfx for drawing primitives,
I think SDL_gfx probably doesn't support enough options to handle all SVG's
requirements. I also didn't like how the closure argument to all the calls
isn't used as it should, instead state is stored in global variables and
tables. For performance the tactic of malloc/copy/free on every single
new LineTo point was really...questionable. However probably a lot of people
would find fault in my own code here.

I liked the approach of leveraging the libsvg out of cairo, to parse the xml
and handle callbacks. However I think the division point of what libsvg
does and what the rendering engine has to do puts too much burden on the
rendering engine.

I think a better approach might be a middle layer that would go between libsvg
and the rendering engine. This middle layer would translate all CurveTo's to
LineTo's, it would handle the bezier curve stuff. All primitives like
rectangles and ovals and circles would get translated into CurveTo's and
LineTo's. All the rendering engine would have to handle is the working out
of what is the actual set of spans to render, then for each span figure out
the value of each pixel based on the gradient + paint that is being applied.

The path itself would be handled by the middle layer also, it would just
call the renderer once for the solid part of the element, then a second time
to stroke the path with its own path + dotted lines + whatnot.

All of this is just too much rework for every single rendering engine.
Even working out the spans could be done in the middle layer. So in the
bottom layer the renderer would just have to deal with the pixel operations on
the frame buffer.

This code development was sponsored by NXTV, Inc. I'm an employee of NXTV.
We took it as far as was necessary for our requirements. It's further than
it was when Enzo left it, and the foundation is there for building out a lot
more functionality. However there is still a lot to do, and not everything
that I've implemented is 100% correct.

I don't want to get stuck maintaining this. Maybe Enzo will pull this into
his sourceforge project and other people will push it further from there.
An SDL implementation of SVG would be extremely useful in a lot of projects
and games. An artist can build an image with inkscape
http://www.inkscape.org/
which is a very impressive SVG editor. It's got a few quirks but overall is
a fantastic program, very useful. One big complaint I have is the little
dialogue boxes that pop up don't have dragbars (at least in my window manager)
nor do they have a close icon.

Anyway read the CHANGES file for some comments on architecture + probems.
Read the TODO file for a list of things I think need doing on this project.