Essentially, the purpose of this is to use only pointers for raylib structs, for use in FFI that only allows pointers, and not pass-by-value.
To use it, head over to releases and grab the latest for your platform. With this, make your FFI bindings however you want. I am making experimental bunjs bindings to test it out.
This explores a new direction for some ideas in raybun and raylib-pointers-ideas.
# build native raylib-pointers (in build/librlptr.[dylib|so|dll])
bun run build
# generate the raylib API JSON (requires build first)
bun run gen:raylib
# generate tools/api.json (requires raylib JSON)
bun run gen:api
# generate src/lib.c from tools/api.json
bun run gen:lib
# generate raylib_bun.js from tools/api.json
bun run gen:bun
# run test.js
bun run test
# clean up code-style
bun run lint
Each one calls it's dependencies, so you can just do bun run build, for example, which will call bun run gen:raylib, then bun run gen:lib.
You can also build without the scripts, if you need to customize things:
# build pointer-only raylib
cmake -G Ninja -B build
cmake --build build
# test it in bunjs
bun test.js
I check in the generated files (tools/api.json, src/lib.c, src/struct.h) so you can build any step without needing previous steps.
I also include api.json, which is the same as raylib's parser output, but includes my additions, and seperates things into categories. You can use this in your own codegen to wrap this lib easier.
unwrapped # original raylib functions that are just exported
wrapped # exported functions that have been wrapped to use pointers
struct_utils # size/get/set util-functions for structs
# these are all original fields:
defines
structs
aliases
enums
callbacks
functions # you should probly use wrapped/unwrapped above, since it's the same info (but categorized, with updated types)