Improve parser performance

[thSoft]

No description provided.

[nittka]

Can you enumerate which aspects should be improved. Linking may still be an issue (caching of libraries that are unlikely to change).

[nittka]

• I think, collecting the imports transitively can be further optimized. Currently, the standard resource loading mechanism is used (potentially including linking), but we only need the plain parse tree in order to find include elements, so unnecessary work is done there.
• I found the refactoring operations extremely slow. I guess, they suffer from the same issue. In order to find affected files, they all have to be loaded. Here is some potential for improvement.
• Measuring the time used for validation during a build should indicate whether it is worth doing som optimization there.

[nittka]

I think optimizations in the GlobalScopeProvider should not be done at this point. Looking at the time spent for a full build of Mozart-KV254 (with a custom XtextResource measuring the time for each step), parsing is where most time is spent now.

parse: 2365 (52 invocations)
link: 373 (52 invocations)

parsedefaultImports: 469 (22 invocations)
linkdefaultImports: 63 (22 invocations)

Getting rid of a backtracking parser seems to be the most promising time saver.