Feature request: trait support #[refuse_impl] attribute, refuse outside implement

When we release some traits, due to lack of consideration, the structure may change in the future. In this case, a break change will occur, and we usually need to make it inherit a private trait to prevent other users from implementing it. I think it is necessary to give users a prompt before that, just like # [deprecated], Next, when a user tries to implement it, compiler give out a warning

This has been accepted by an RFC. It's a hard error, not a warning.

before change it, It is better to give users a prompt.

Please read the RFC. It answers a number of questions you likely have.

I think OP might be suggesting some kind of "user implementations on this trait are deprecated" annotation, to facilitate eventually turning a trait sealed in a next major version of their crate - or generally to mark a trait as "this was never intended to be user-implementable". The intended effect would be a warning when the user tries to implement the trait. Note "sealing" vs "sealed".

2 Likes

Yeah

Sounds like something that could be a more general attribute ("going to break this soon even though it's not deprecated").

  • Visibility changes
  • Meant to make this non_exhaustive
  • Taking away a visible field or variant
  • Need to change some variance, auto trait, or other leakage
  • Removing a default trait method body
  • Changing a blanket implementation
  • Etc

Incidentally, feature requests that are just thrown out there and not championed rarely go anywhere.

5 Likes

mybe, the name should be change to #[refuse_impl].

This topic was automatically closed 540 days after the last reply. New replies are no longer allowed.