I'm stating reasons / ways to work on tools to get the generated files out of the source tree so they don't need to be tracked.
why?
they hinder resolving diffs (even when resolved, it would need to be re-resolved in the future).
they hinder concurrent development (harder to solve diffs when more people work on code at the same time).
they hinder development by having a automatized process to be a developers responsibility.
tools / ways to handle and distribute generated files
testing could be done using a Makefile (or any kind of dependency resolver) to do tree-sitter generate before tree-sitter test (no need to keep track of them).
generated source files (eg src/parser.c) can still be offered as a download (for consumers of "a Tree-sitter language") by adding a CI release package that includes generated source files (CI runs tree-sitter generate and packs them).
could re-use such CI to also offer compiled Tree-sitter language objects (.so, .dylib, .dll)
I can do the work myself if people are patient enough to wait. I would like to hear everyone else's opinion on this matter.
well, I just found out that this is a discussion that should be scaled to the main tree-sitter project since all tree-sitter-{lang} repos share this structure
I'm stating reasons / ways to work on tools to get the generated files out of the source tree so they don't need to be tracked. why?
tools / ways to handle and distribute generated files
tree-sitter generate
beforetree-sitter test
(no need to keep track of them).tree-sitter generate
and packs them).I can do the work myself if people are patient enough to wait. I would like to hear everyone else's opinion on this matter.