Add some tests exhibiting meaning of generic type constraints #18213
Conversation
This captures some patterns we were discussing today that I hadn't internalized correctly. Hopefully writing these will help me keep the rules straight and lock in the current behavior if we don't already have tests covering them (we very well may). --- Signed-off-by: Brad Chamberlain <bradcray@users.noreply.github.com>
|
@e-kayrakli : Following up on our discussion and my confusion this morning, would you review these to see if they match what we were discussing and whether you think there are other key patterns to try and capture / lock in? (I put this together while writing up an "are we sure this is right?" issue). |
| @@ -0,0 +1,3 @@ | |||
| allGenericsDefaulted.chpl:43: error: unresolved call 'baz(R(real(64),2))' | |||
| allGenericsDefaulted.chpl:17: note: this candidate did not match: baz(r: R(int, ?rank)) | |||
| allGenericsDefaulted.chpl:17: note: because an argument was incompatible | |||
There was a problem hiding this comment.
Clearly not about this PR, but I am not sure if should try to make the error messages in {1,2} to be similar to {3,4}. The latter set seems more vague, and it can be less helpful if the function had many arguments.
There was a problem hiding this comment.
I don't disagree, but haven't felt sufficiently inspired to file the issue (yet). Let me know if you think it's important that I do so (vs. waiting for someone to truly be annoyed enough by it to get around to it themselves).
There was a problem hiding this comment.
Nope. This error message definitely falls into "good enough for now" category for me.
|
One thought I had given the exchange we just had on #18215 would be to expand the partially and completely generic cases to see what happens to the variable declarations if |
Happily, they are symmetrical with formal types. --- Signed-off-by: Brad Chamberlain <bradcray@users.noreply.github.com>
|
@e-kayrakli : As alluded to yesterday, I did end up adding another line to each test to explore what a declaration of the form |
|
Thanks! |
This captures some patterns we were discussing today regarding the
interpretation of
Rin the event that a generic record is fully, partially,or not at all defaulted. I hadn't internalized these rules correctly, so
wrote this both to improve my understanding and ensure we had a test
that locked in the expected behavior (though we likely do, but I have a
"the more tests the better" mindset).