This page is WIP! - KrzesimirNowak
gmmproc is a tool used for creating c++ source codes from templates. The whole process has two stages:
- Generating defs files and xml documentation of a wrapped C library - done by user.
- Generating source code using template, defs and xml documentation - done by gmmproc.
I'm using two names on this page:
- gmmproc - current generator.
- mm-proc - rewritten generator.
Problems with gmmproc
These are not written in order of importance.
- The second stage of generation is run per file. That means that defs and xml are parsed again and again, when they could be parsed once.
- Gmmproc now uses perl and m4. Their interdependence causes adding features to be a hard task.
- Defs file are not perfect - they are not always correct and does not contain all needed information.
- To generate defs file several third-party tools need to be used and some have to be written by hand.
- Docs parser sometimes has problems with converting xml documentation to doxygen comments.
- Because I didn't studied gmmproc internals very very deep, I suppose that some points could be written yet. - krnowak
What we want from mm-proc
- Easy extensibility.
- Use information gir files gives.
- Instead of generating defs (.gir) it would be better to maybe download it?
Two stage process:
- Rework defs and xml parser to use gir files. Xml documentation file would stay, because gir contains all information about C API, but not all documentation.
- Rewrite gmmproc itself to use only one language - perl or python.
- I suppose that moving to gir files should be the first step, so new gmmproc could utilize some information from gir. - krnowak
- I would use perl. - krnowak
- I think that new gmmproc (or mm-proc) should be placed in mm-common. - krnowak
- Should mm-proc's workflow differ much from original? I mean - first parsing defs (.gir) and xml, then parsing templates and them outputting source code. - krnowak
- Working on all templates would allow us to analyze them to find what is wrapped into what, so for example docs could be properly transformed. - krnowak
- Find out what else is wrong with current gmmproc, so we could have that in mind while designing new generator.
- Find out what else we want from rewritten mm-proc, which gmmproc does not have.