Caching of compilation of crates between different calls to load_workspace_at()

[nicola-gigante]

Hi all,

as you know I'm using RA to analyze Rust crates and their dependencies, so I'm going all through std and core. So even for small crates I'm doing a lot of work each time, even though core is not going to change very often (actually not at all, if the toolchain is the same) and the same can be said for any dependency that comes from crates.io or from a git repository with a specific commit hash.

So I was wondering if there is any way to cache the work that RA is doing when parsing those dependencies. From what I see, the Salsa database is not meant to be serializable, but maybe there's something I can do at a higher level?

I may also cache things on my side, i.e. cache the result of my extraction instead of caching RA's compilation, but before I'd like to know if there's anything simpler I can do from RA side.

Any hint?

[ChayimFriedman2]

Serialization for faster startup is planned but not done yet.

You can have your own cache, as you said.

[nicola-gigante]

I'm measuring right now and it seems most of the time is spent on the call of hir::Crate::root_module() on the first crate, which I would need to call anyway even if I cached results for dependencies. So it seems caching on the RA side is really what I'd need.

Fortunately it's not a blocker for me, I can wait. Any plans on when serialization will be worked on?

[ChayimFriedman2]

When someone will work on it (from the Salsa side first, although IIUC some rudimentary serialization is already implemented in Salsa). You can be that one if you want.

[nicola-gigante]

Oh I would be happy to have time to contribute, but I don't have this luxury at the moment :(

Anyway as I said this is not a blocker so I can go ahead with all the rest of the project.
Btw I'll tell you what I'm working on when it'll be ready. RA is a crucial part of what I'm doing so thanks! 🙏