Switch mapping of context bounds to using clauses in 3.6 #21257
Conversation
|
This needs a change in scaladoc to recognize using clauses with evidence parameters as context bounds. Right now it seems the logic is that a context bound is recognized only if the evidence parameter is an implicit. @Florian3k can you take this on? Alternatively, who else is working on scaladoc these days? |
|
Mikołaj is currently having ~2 weeks off to focus on his thesis, @Gedochao we might need to look for some else to handle the issue |
Nobody else AFAIK. So it's fair game for anyone, if Mikołaj isn't available. |
|
Shouldn't this be a quick fix around https://github.com/scala/scala3/blob/main/scaladoc/src/dotty/tools/scaladoc/tasty/ClassLikeSupport.scala#L633 ? Last PR that touched it was #13172 We can either wait for him or try to fix it here (I think you have most context about the change and could probably change it quickest though) |
Was future before.
odersky
force-pushed
the
change-ctx-bounds-mapping
branch
from
ba82d69 to
b8f9c2c
Compare
If we have a function like ```scala def foo[X: CB](...)(implicit x: T) ``` always map context bounds to implicit parameters, irrespective of version. Likewise, if we have a function ``scala def foo[X: CB](...)(using x: T) ``` always map context bounds to "using" parameters, irrespective of version. This avoids mixing implicit and using in one parameter list.
|
@KacperFKorban Thanks for the pointers! |
Was future before.
We can roll this out now, since we already made it an error in 3.5 to pass a normal argument to a using clause. It would be good to roll this out now since otherwise context bounds would represent an exception to the implicit priority inversion in #19300.
See discussion at the end of #19300 for details.