Tracking issue for RFC 2451, "Re-Rebalancing Coherence"

[Centril]

This is a tracking issue for the RFC "Re-Rebalancing Coherence" (rust-lang/rfcs#2451).

Steps:

• Implement the RFC – Implement the Re-rebalance coherence RFC #56145
• Adjust documentation (see instructions on forge)
• Stabilization proposal
• Stabilization PR (see instructions on forge)

Unresolved questions:

• Are there additional implementations which are clearly acceptable under the current restrictions, which are disallowed with this extension? Should we allow them if so?

[dwijnand]

Implemented in #56145.

[weiznich]

To just add a implementation example:
We've used this feature to improve the implementation of diesel-oci. For this use case it seems to work fine.

@Centril Somebody needs to update the boxes in the opening comment and change the label to B-RFC-implemented.

[nikomatsakis]

We discussed this some time back and the plan was for me to write a stabilization report. But I've not done so. I do expect to do so, but perhaps not for a week or two.

[pmarks]

@nikomatsakis in case it's useful: sparsemat/sprs#160

ndarray::Dot is a trait similar to Mul but for matrix multiplication. ndarray provides dense matrices and Dot impls for dense matrices. The sprs crate for sparse matrices wants to interoperate with ndarray and provide generic sparse.dot(dense) and dense.dot(sparse) impls. Both are needed since matrix multiplication isn't commutative. We're able to provide impl Dot<sprs::CsMat<_,_>> for ndarray::ArrayBase<_,_> from the sprs crate with this feature. We're looking forward to having it stabilized!!

[djc]

What's the holdup for moving this towards stabilization? Still waiting for a lang team member to write the stabilization report? Is there a template that a non-team member could follow to assist with this work?

[Centril]

@djc Niko is on vacation and generally busy so he hasn't had time to write up the report yet. If you want to help out with that, there's inspiration to find in #61682 (comment) (which is my most recent and elaborate report hitherto), #61347 (comment), and #57175. Feel free to reach out to me on Discord's #design channel to discuss your progress / draft.

[nikomatsakis]

@djc see #63599

[nikomatsakis]

Er, I could have worded that a bit more broadly for everyone. =) #63599 contains a stabilization proposal. Thoughts etc welcome in that issue.

[djc]

@nikomatsakis awesome, thanks! Sorry I couldn't contribute anything in the end.

[tesuji]

Stabilized in #65879

[QuineDot]

This RFC stated that

RFC #1105 is amended to remove the statement that implementing a non-fundamental trait is a minor breaking change, and states that adding any blanket impl for an existing trait is a major breaking change, using the definition of blanket impl given above.

However, the RFC has not been updated. (The reference was updated in PR 703.)