Evaluate approaches for seamless interop between C++ and Rust

[nikomatsakis]

Metadata
Point of contact	@tmandry
Team(s)	compiler, lang, libs-api
Goal document	2025h1/seamless-rust-cpp

Summary

Seriously consider what it will take to enable Rust adoption in projects that must make use of large, rich C++ APIs. Map out the space of long-term solutions we are interested in. These solutions should enable interop between Rust and other languages in the future.

Tasks and status

• Discussion and moral support (lang, compiler, libs-api )
• Design meeting (lang, compiler, libs-api )
• Lang-team champion (lang )
• Author design doc (@tmandry)
• Author design doc ()
• Author design doc ()

[nikomatsakis]

This issue is intended for status updates only.

For general questions or comments, please contact the owner(s) directly.

[tmandry]

We had a lang team design meeting about C++ interop today. The outcome was very positive, with enthusiasm for supporting an ambitious vision of C++ interop: One where a large majority of real-world C++ code can have automated bindings to Rust and vice-versa.

At the same time, the team expressed a desire to do so in a way that remains in line with Rust's values. In particular, we would not make Rust a superset of Rust+C++, but instead would define extension points that can be used to express language interop boundaries that go beyond what Rust itself allows. As an example, we could allow template instantiation via a Rust "plugin" without adding templates to Rust itself. Similarly, we could allow calling overloaded C++ methods without adding function overloading to Rust itself. Other interop needs are more natural to enable with features in the Rust language, like custom reference types.

In either case, anything we do to support interop will need to be considered on its merits. Interop is a reason to support a feature, but it is never a "blank check" to add anything we might consider useful.

The discussion so far has been at a high level. Next steps will be:

• Discuss what significant-but-realistic milestones we might pursue as part of upcoming project goals, and what it would take to make them happen. Whether this happens as part of another lang team meeting or a more dedicated kickoff meeting for interested parties, I'll be sure to keep the lang team in the loop and will continue posting updates here.
• Dive into more specific proposals for use cases we would like to enable in meetings with the language, library, and compiler teams.

Notes: https://hackmd.io/2Ar_7CNoRkeXk1AARyOL7A?view

[tmandry]

Since our last update, there has been talk of dedicating some time at the Rust All Hands for interop discussion; @baumanj and @tmandry are going to work on fleshing out an agenda. @cramertj and @tmandry brainstormed with @oli-obk (who was very helpful) about ways of supporting a more ambitious "template instantiation from Rust" goal, and this may get turned into a prototype at some point.