Crate structure cleanup

[jackh726]

There are a few different crate structure cleanups we want to do.

• Move recursive solver out of chalk-solve
  • A lot of the functionality here overlaps with chalk-engine, so it would be nice try to set up the crate structure in a way that we only need to compile one or the other.
• Merge chalk-rust-ir into chalk-solve
  • This is a pretty small crate and there's not really a benefit to keeping it split.
• Move chalk-macros into chalk-engine
  • chalk-macros is also pretty small. This might be a little different, depending on the how the first bullet point goes. But basically, this doesn't need to be all by itself.

This issue has been assigned to @Mcat12 via this comment.

[AzureMarker]

I can take this, though I have a question about "Merge chalk-rust-ir into chalk-solve". Will it need to be moved out again once a common type library is shared with rustc? Keeping it out of chalk-solve would therefore be closer to the target state.

@rustbot claim

[AzureMarker]

Is the SLG solver wholly contained within chalk-engine, or is it also in chalk-solve? It may make sense to have a structure where chalk-engine provides the shared types between the SLG and recursive solvers. The solvers would be in their own chalk-slg-solver/chalk-recursive-solve crates and depend on chalk-engine. Then chalk-solve would depend on all three crates. Does this make sense?

[jackh726]

I think @zaharidichev wanted to work on this: https://rust-lang.zulipchat.com/#narrow/stream/144729-wg-traits/topic/coinduction.20in.20the.20recursive.20solver.20chalk.23399/near/197735953

[jackh726]

But to answer your questions @Mcat12, the types that we need to share are all in chalk-ir.

As far as SLG, the core logic is all inchalk-engine, but the Slg context and aggregate logic is in chalk-solve. I was also think that there would be a shared crate that both slg and recursive solvers depend on. And solve depends on both. But I don't really know if about the recursive solver to say how much is actually shared.

[zaharidichev]

Yeah I wanted to but life is first come first served :) So take a stab @Mcat12! @jackh726 Anything else that I can dig into (something hard)?

[AzureMarker]

That's fine, I can unassign myself assuming he's still interested.

Do you know of other issues that might help me learn the codebase? I'm at an "intermediate" level with the codebase, but I found some issues that may be doable. Perhaps you would have a better idea which would be most suitable? #313 #429

@rustbot release-assignment

Edit: Doh! 10 seconds too late

[jackh726]

Let me open a topic on Zulip for a few things people can work on!

[AzureMarker]

Thanks for the Zulip thread @jackh726 (link)! I'll re-claim this issue; it looks like @zaharidichev will be working on #313.

@rustbot claim

[AzureMarker]

Based on a Zulip discussion, we'll keep the recursive solver in chalk-solve for now but still have the "chalk-engine-base" crate which holds the few shared solver types (replacing chalk-macros). There will be crate features to disable the SLG/recursive solvers, to avoid unnecessary compilation.