Introduce RefCell::{try_replace, try_replace_with, try_swap}

[daboross]

This adds Result-based equivalents to RefCell::replace, RefCell::replace_with and RefCell::swap.

This was proposed in #54493, and has not gone through an ACP or RFC process.

Concerns

I believe the design of these functions should be relatively uncontroversial, but their presence at all may be. It's unclear how much these would be used, and they do clutter the documentation of RefCell. The main purpose is completeness: offering a Result-based method for every panicking RefCell method.

Looking at history, RFC-2057 (PR, rendered), which added replace and swap, simply didn't address the possibility of try_replace and try_swap. The author dismissed them as not very useful.

Error types

#54493 proposed adding unique error types for each function. I've opted instead to reuse BorrowMutError, matching try_borrow_unguarded reusing BorrowError.

Process?

As I haven't gone through the rust-lang/rust PR process much yet, I'd love guidance on what to do with this. I figured writing this PR couldn't hurt, but it's unclear whether an ACP or RFC would be best here.

The prior additions of RefCell::{try_borrow, try_borrow_mut} and RefCell::{swap, replace} both used RFCs, but it's unclear to me if things have changed since then.

[rustbot]

r? @scottmcm

rustbot has assigned @scottmcm.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[scottmcm]

These days you don't need full RFCs just to add methods; and ACP is the right choice.

(Library RFCs are more for traits that lots of the ecosystem should implement, for example, or large additions of new functionality that set new precedents or scopes.)

[daboross]

Great, makes sense!

I'll make an ACP issue now & link it.

[daboross]

Finally got around to writing that ACP - rust-lang/libs-team#472

I think it's given me a new perspective - I'm not sure this is something we want. But I think it's likely enough that I'm glad to have written that, and that we can decide on it.

[daboross]

Per rust-lang/libs-team#472 (comment), the ACP was closed as without sufficient justification, so I'll close this as well.