Warning
This project is under active development, and has not yet been released. That said, you might already get some usable results out of it; if something breaks, please check the issue tracker to see if the problem is already known, and open an issue if not.
The goal of the low-level API is to generate Haskell bindings from C headers, in "fire and forget" mode.
The most important existing tools for the generation of Haskell bindings from C
headers, hsc2hs and c2hs, require a lot of user input (see Alternative
generators for
a full review): they assist in writing bindings by filling in details about the
C code when requested, but the process is still driven by the programmer. The
goal of hs-bindgen, inspired by the Rust
bindgen tool, is to have the
entire process be driven by the C header(s) themselves.
The bindings generated in this mode are too low-level for convenient usage in
Haskell applications; for example, char* in C translates to Ptr CChar in
Haskell, instead of String (or Text, ByteString, ..). Tools like c2hs
can be used to define higher-level bindings, but not automatically, and at the
cost of having to learn a rather arcane bespoke syntax. Using hs-bindgen, even
if we do not use it to generate high-level Haskell bindings, users can write
their own high-level bindings on top of the low-level bindings, just as
regular Haskell code. This should also improve integration with tooling such as
HLS. Writing such bindings
could potentially assisted by regular Haskell functions to capture common
patterns.
The goal of the high-level API is to be able to generate higher-level bindings using heuristics to recognize particular patterns, with a good set of defaults that can be extended by users for specific applications.
Note
We have not yet begun work on the high level API.
You can find the (currently still very incomplete) manual in manual/.