thindx - Thin DirectX wrappers

Thin DirectX wrappers for Rust.

Why not winapi directly?

• This crate aims to make fns safe/sound/slightly rusty when possible
• Attempts to verify API soundness through mass unit tests, even if they mostly test DirectX's behavior.
• Doc comments for one-stop intellisense, safety documentation, etc.

Why not <other graphics crate>?

• Most other graphics crates focus on hiding the underlying graphics API as much as possible, improving application portability.
• This crate surfaces the underlying graphics API as much as possible, improving the potential for interoperability with other graphics code / use in middleware applications.

❌ This crate is probably unsound! ❌

I'm exposing a huge legacy C++ API to Rust. Mistakes will happen.

That said, soundness is a very high priority goal. thindx will add things like extra bounds checks, parameter validation, extra init, etc. if need be in order to ensure soundness in safe fns whenever possible. When it's not possible to validate unsoundness away, the fns in question should be marked unsafe. This crate strives to be sounder than whatever manual FFI you'd write yourself would be, and that's a high bar.

But there are some practical limits to this. If a background driver thread invokes UB if it fails to allocate memory, without any direct correlation to specific API misuse like a large integer overflowing, that's a bug I can't sanely mitigate via safe fns. I mean, theoretically I could write a pure-software clone of the entire DirectX runtime... but no.

Additionally, while I'm seeking to validate my APIs via testing, older implementations of the APIs in question may have more bugs / unchecked parameters / ??? that I'll fail to mitigate due to being unable to trigger them myself. While I'm happy to investigate, accept pull requests, expand test coverage, etc. it's worth assuming this crate is unsound on older versions unless you've tested yourself.

⚠️ API major version churn ⚠️

Individual fns are likely to gain/lose unsafe, traits, etc. in a neverending attempt to make DirectX access sound. As such, thindx itself will likely always suffer from major version churn. This isn't too much of a problem until two crates wish to share / pass thindx types between themselves. It might be possible to somewhat stabilize some types by exiling them into subcrates, but this has not yet been done. Additionally, individual extension traits / functions / methods will likely never get the same treatment (no need?)

Project Status

examples • header coverage • docs.rs • lib.rs • crates.io

License

Licensed under either of

• Apache License, Version 2.0 (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
• MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)

at your option.

Contribution

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.