As I’ve finished the documentation-madness I stumbled over my GLSL shader. Doxygen, _the_ documentation tool of choice, reads them (as they are very c-like) but dumps them rather unfancy. Doxygen is not only very flexible – it is also extensible with filters to process other languages.
Idea
Write a small filter, that pads every shader into a class and set a namespace that acts kinda like a category (e.g. class Gauss
in namespace GLSL::FILTER::BLUR
). The result is my glslfilter.py python-script.
Usage
- Make Doxygen aware of the filter and the newly supported file extensions. To do this, edit your Doxyfile:
- Add
FILE_PATTERNS: *.frag, *.vert
- Add
FILTER_PATTERNS: "*.frag=./glslfilter.py", "*.vert=./glslfilter.py"
- AddÂ
EXTENSION_MAPPING=.frag=C++, .vert=C++
(thanks the_summer)
- If you want, add annotations to your shader:
- Use
@class
to set the class name
- Use
@namespace
to set the namespace – a category
If you set no name, the script will use the bare filename and the default namespace is “GLSL”.
Of course you can further document your shader with doxytags – they are fully processed by Doxygen. The only limitation here (blame my lazyness or my thought it just wasn’t worth it) is that you have to put the “class” comment (= the shader information) at the very beginning in one big blockcomment starting with /*
or /**
. But seriously – why would you comment it in any other way? 
To better illustrate the procedure here’s a little GLSL fragment shader:
Example
/**
* A simple 3x3 gaussian convolution filter, non-separated version
* @author Sebastian Schaefer
* @date 2012
* @namespace GLSL::FILTER::BLUR
* @class Gauss3x3
*/
#version 150 core
uniform sampler2D image; ///< the input image
in vec2 tex; ///< texture coordinated
out vec4 color; ///< the color output
/**
* The main routine: read the 3x3 neighbours and multiply with kernel
*/
void main()
{
...
}
Click here to see how the above example can look like.
Download
Grab the filter at my github-repository github.com/numb3r23/glslAdditions
This page might get updated so please link to this page only, not the direct download. thank you!
—
edit: moved source to github