Safe autogenerated interop between Rust and C++

10 releases (4 breaking)

0.5.1 Feb 20, 2021
0.4.1 Dec 16, 2020
0.4.0 Nov 10, 2020

#53 in FFI

Download history 8/week @ 2020-11-16 17/week @ 2020-11-23 11/week @ 2020-11-30 14/week @ 2020-12-07 22/week @ 2020-12-14 5/week @ 2020-12-21 9/week @ 2020-12-28 40/week @ 2021-01-04 15/week @ 2021-01-11 14/week @ 2021-01-18 22/week @ 2021-01-25 4/week @ 2021-02-01 17/week @ 2021-02-08 64/week @ 2021-02-15 12/week @ 2021-02-22 11/week @ 2021-03-01

83 downloads per month
Used in 3 crates




Docs.rs Crates.io

This project is a tool for calling C++ from Rust in a heavily automated, but safe, fashion.

The intention is that it has all the fluent safety from cxx whilst generating interfaces automatically from existing C++ headers using a variant of bindgen. Think of autocxx as glue which plugs bindgen into cxx.


namespace base {
  class Bob {
      Bob(std::string name);
      void do_a_thing() const;
use autocxx::include_cpp;

    #include "base/bob.h"

let a = ffi::base::Bob::make_unique("hello".into());

See demo/src/main.rs for a real example.

The existing cxx facilities are used to allow safe ownership of C++ types from Rust; specifically things like std::unique_ptr and std::string - so the Rust code should not typically require use of unsafe code, unlike with normal bindgen bindings.

How it works

Before building the Rust code, you must run a code generator (typically run in a build.rs for a Cargo setup.)


  • First, runs bindgen to generate some bindings (with all the usual unsafe, #[repr(C)] etc.)
  • Second, interprets and converts them to bindings suitable for cxx::bridge.
  • Thirdly, runs cxx::bridge to create the C++ bindings.
  • Fourthly, writes out a .rs file with the Rust bindings.

When building your Rust code, the procedural macro boils down to an include! macro that pulls in the generated Rust code.

Current state of affairs

There is an example of this macro working within the demo directory.

The project also contains test code which does this end-to-end, for all sorts of C++ types and constructs which we eventually would like to support.

Type Status
Primitives (u8, etc.) Works
Plain-old-data structs Works
std::unique_ptr of POD Works
std::unique_ptr of std::string Works
std::unique_ptr of opaque types -
Reference to POD Works
Reference to std::string Works
Classes Works, but with warnings
Methods Works (classes give warnings)
Int #defines Works
String #defines Works
Primitive constants Works
Enums Works, though more thought needed
#ifdef, #if etc. -
Typedefs Infinite permutations, some of which work
Structs containing UniquePtr Works
Structs containing strings Works (opaque only)
Passing opaque structs (owned by UniquePtr) into C++ functions which take them by value Works
Passing opaque structs (owned by UniquePtr) into C++ methods which take them by value Works, but with non-ideal syntax
Constructors/make_unique Works, though probably many problems
Destructors Works via cxx UniquePtr already
Inline functions Works
Construction of std::unique_ptrstd::string in Rust Works
Namespaces Works, but a known limitation
std::vector Works
Field access to opaque objects via UniquePtr -
Plain-old-data structs containing opaque fields Impossible by design, but may not be ergonomic so may need more thought
Reference counting, std::shared_ptr -
std::optional -
Function pointers -
Unique ptrs to primitives -
Inheritance from pure virtual classes -
Generic (templated) types Works though likely many problems

The plan is (roughly) to work through the above list of features and fix corner cases. This project is deliberately incremental. There are open questions about whether the end result is ergonomic and performant: specifically, whether it's acceptable to hold opaque C++ types always by UniquePtr in Rust. Until we know more, this project is considered experimental and we don't advise using it for anything in production.

On safety

This crate mostly intends to follow the lead of the cxx crate in where and when unsafe is required. But, this crate is opinionated. It believes some unsafety requires more careful review than other bits, along the following spectrum:

  • Rust unsafe code (requires most review)
  • Rust code calling C++ with raw pointers
  • Rust code calling C++ with shared pointers, or anything else where there can be concurrent mutation
  • Rust code calling C++ with unique pointers, where the Rust single-owner model nearly always applies (but we can't prove that the C++ developer isn't doing something weird)
  • Rust safe code (requires least review)

If your project is 90% Rust code, with small bits of C++, don't use this crate. You need something where all C++ interaction is marked with big red "this is terrifying" flags. This crate is aimed at cases where there's 90% C++ and small bits of Rust, and so we want the Rust code to be pragmatically reviewable without the signal:noise ratio of unsafe in the Rust code becoming so bad that unsafe loses all value.

See [safety!] in the documentation for more details.

Build environment

Because this uses bindgen, and bindgen may depend on the state of your system C++ headers, it is somewhat sensitive. It requires llvm to be installed due to bindgen

As with cxx, this generates both Rust and C++ side bindings code. You'll need to take steps to generate the C++ code: either by using the build.rs integration within autocxx_build, or the command line utility within autocxx_gen. Either way, you'll need to specify the Rust file(s) which have include_cpp macros in place, and suitable corresponding C++ and Rust code will be generated.

When you come to build your Rust code, it will expand to an include! macro which will pull in the generated Rust code. For this to work, you need to specify an AUTOCXX_RS environment variable such that the macro can discover the location of the .rs file which was generated. If you use the build.rs cargo integration, this happens automatically. You'll also need to ensure that you build and link against the C++ code. Again, if you use the Cargo integrationm and follow the pattern of the demo example, this is fairly automatic because we use cc for this.

You'll also want to ensure that the code generation (both Rust and C++ code) happens whenever any included header file changes. This is now handled automatically by our build.rs integration, but is not yet done for the standalone autocxx-gen tool.

Finally, this interop inevitably involves lots of fiddly small functions. It's likely to perform far better if you can achieve cross-language LTO. https://github.com/dtolnay/cxx/issues/371 may give some useful hints - see also all the build-related help in https://cxx.rs/ which all applies here too.

Directory structure

  • demo - a demo example
  • parser - code which parses a single include_cpp! macro. Used by both the macro (which doesn't do much) and the code generator (which does much more, by means of engine below)
  • engine - all the core code for actual code generation.
  • macro - the procedural macro which expands the Rust code.
  • gen/build - a library to be used from build.rs scripts to generate .cc and .h files from an include_cxx section.
  • gen/cmd - a command-line tool which does the same.
  • src (outermost project) - a wrapper crate which imports the procedural macro and a few other things.

Where to start reading

The main algorithm is in engine/src/lib.rs, in the function generate(). This asks bindgen to generate a heap of Rust code and then passes it into engine/src/conversion to convert it to be a format suitable for input to cxx.

However, most of the actual code is in engine/src/conversion/mod.rs.

At the moment we're using a slightly branched version of bindgen called autocxx-bindgen. It's hoped this is temporary; some of our changes are sufficiently weird that it would be presumptious to try to get them accepted upstream until we're sure autocxx has roughly the right approach.

How to develop

If you're making a change, here's what you need to do to get useful diagnostics etc. First of all, cargo run in the demo directory. If it breaks, you don't get much in the way of useful diagnostics, because stdout is swallowed by cargo build scripts. So, practically speaking, you would almost always move onto running one of the tests in the test suite. With suitable options, you can get plenty of output. For instance:

RUST_BACKTRACE=1 RUST_LOG=autocxx_engine=info cargo test  integration_tests::test_cycle_string_full_pipeline -- --nocapture

This is especially valuable to see the bindgen output Rust code, and then the converted Rust code which we pass into cxx. Usually, most problems are due to some mis-conversion somewhere in engine/src/conversion.


David Tolnay did much of the hard work here, by inventing the underlying cxx crate, and in fact nearly all of the parsing infrastructure on which this crate depends. bindgen is also awesome. This crate stands on the shoulders of giants!

License and usage notes

This is not an officially supported Google product.

Licensed under either of Apache License, Version 2.0 or MIT license at your option.


~137K SLoC