Unused warnings

[finnteegen]

An expression like count = [let { _ = print [i]; i = getLine } in ()] yields a warning that i is unused (which is not the case apparently).

[Ziharrk]

i is only relevant for the effect and not the actual result (). Should we disable those warnings in the SML plugin completely? Or is it still sensible to warn a user about this?
Maybe we could just try to mark values with type () (e.g. the _ in this case) as always used. That would prevent most insensible warnings, while not removing the warning in sensible cases.

[finnteegen]

I do not have a good answer for that question. Due to the strict evaluation, everything is evaluated at some point, which makes it hard to refer to things as unused. Nevertheless, you are right that these warnings point the programmer to common errors - at least most of the time. In this case, I would argue that i is not unused because it is cleary referenced in another statement. Perhaps marking values of type () is plausible.

[Ziharrk]

As a reminder

Someone has noticed a related issue in GHC: https://gitlab.haskell.org/ghc/ghc/-/issues/19003