Comment #0 by qs.il.paperinik — 2022-11-23T10:37:48Z
Currently, `with` can only have one type or expression argument. When multiple resolvers are needed, one has to nest the `with` statements. Among the semantics is that one has priority over the other. If that is not intended, there’s no way to get around it.
Proposed: `with (A, B, C)`.
When an identifier `id` is encountered and among `A.id`, `B.id`, and `C.id` more than one resolve, it is an ambiguity error. It expresses the intent of the programmer that either ambiguity is expected to be impossible or that ambiguity is might be impossible, but the code should explicitly disambiguate to avoid confusion on the side of the reader. When members are added or renamed in `A`, `B`, or `C`, this construct is resistant to silent change of behavior.
When `Ts` is a compile-time sequence (AliasSeq), `with (Ts)` does not compile, even if the sequence has exactly 1 element.
Proposed:
When `Ts` is a compile-time sequence (AliasSeq), `with (A, Ts, C)` is as if `with (A, Ts[0], ..., Ts[$-1], C)`, i.e. `Ts` will be expanded as one would expect.
Comment #1 by robert.schadek — 2024-12-13T19:25:52Z