The approach to this crate can best be summarized as wanting to expose all functionality, even if not idiomatically. The libretro API has quite a bit of functionality, and implementing all of that up front would just cause delays. Instead, all of the raw API is made available to you. If you run into something that isn't explicitly supported, you can always use the raw API.

Another design note is to try and include data that would be available to a C implementation. For example, emulators cannot reasonably be expected to be constructed without parameters (à la Default). Therefore, there is a RetroCore::init function that can acccess the RetroEnvironment, so things like asset paths can be used to construct the emulator. With that in mind, care will be taken to hide API functionality where it isn't allowed.

After the above requirements are met, the last goal is to make the bindings logical and ergonomic. If you're feeling pain anywhere in the API then definitely let us know!

A reference implementation is available in the example folder, and should be enough to get you started. At a high level, you need to modify your Cargo.toml:

[dependencies]
libretro-rs = "0.1"

[lib]
crate-type = ["cdylib"]

Then implement a trait and call a macro:

use libretro_rs::*;

struct Emulator {
  // ...
}

impl RetroCore for Emulator {
  // ...
}

libretro_core!(Emulator);

et voilà! Running cargo build will produce a shared library (.so, .dll, etc) that you can use with a libretro front-end:

$ retroarch --verbose -L libemulator.so /path/to/game.rom

The example folder contains a working CHIP-8 emulator. You can run it using this command:

$ cargo build -p example
$ retroarch --verbose -L ./target/debug/libexample.so <game path>

Replace .so with .dll if running on Windows.