The goal of wondermagick is to advance the Rust imaging ecosystem as a whole. We want to enable people to migrate other imaging libraries to Rust backends. Making wondermagick bypass the common libraries with bespoke solutions is not the goal.
To that end we have a policy of upstream-first contributions. That is, if whatever you want to implement can be implemented in a dependency crate, such as image, it should be done there and not in wondermagick. To facilitate that, wondermagick may depend on unreleased versions of dependency crates (pre-releases or versions from Git).
We don't want any bespoke unsafe code in wondermagick. We declare #![forbid(unsafe_code)] to make the compiler yell at you if you try to add any.
We are also cautious with pulling in dependencies that use unsafe code. Any unsafe code that enhances performance should be disabled, with the exception of SIMD intrinsics. Any unsafe code that does get included despite the previous restriction, should be fuzzed with cargo fuzz or an equally capable fuzzer (e.g. AFL) and a fuzzing harness contributed upstream. The project must also have a test suite and run it in CI under Address Sanitizer or miri. If it does not, please add such a CI setup upstream before adding the dependency to wondermagick.
There are two main ways to contribute code:
- Improve crates such as
imagethatwondermagickrelies on. See here for a list of things we need. - Add new commands to
wondermagick
wondermagick commands should be backed by algorithms available elsewhere, not just in wondermagick. If you are contributing e.g. a blur implementation, ideally it should be in image or imageproc. If you really need to write it yourself and contributing to an existing crate is not an option (e.g. the maintainers rejected it), please structure it as a separate crate within this repository.
To add a new command to wondermagick, add your argument to the Arg enum in src/args.rs and follow the trail of compiler errors. Once it compiles, you're done!