Import disambiguation: Corrects handling of fully-applied and one-sided sectioned operators in qualifying strategy

[konn]

The current implementation of the import disambiguation (introduced in #1264) always qualifies operators with extra parens.
This can destroy the fully- or partially-applied infix operators, as follows:

op = (++)
fun xs ys = xs ++ ys
funL xs = (++ xs)
funR ys = (ys ++)

↓

op = (Prelude.++) -- As expected
fun xs ys = xs (Prelude.++) ys -- oops!
funL xs = ((Prelude.++) xs)  -- ooops!
funR ys = (ys (Prelude.++)) -- oooops!

This PR just fixes it, inspecting the buffer whether the code in question is already parenthesised or not.

[jneira]

lgtm, and good test coverage

[berberman]

If you don't mind, could you unify the filename extension of golden tests data to .hs? It would be better for review and can clarify that our codebase doesn't contain E language source code 😄

[Ailrun]

If you don't mind, could you unify the filename extension of golden tests data to .hs? It would be better for review and can clarify that our codebase doesn't contain E language source code 😄

Yes, especially when the "expected" version is better source, like in this case.

[konn]

Ah, I just followed the pattern used in Eval plugin, but, yes, those expected files in eval func-test doesn't have so much long extensions. I now renamed disambiguation-related test files to have .hs extensions. Thank you for your suggestions, @berberman @Ailrun!

[berberman]

Can we remove isSymOcc here?

Did you see my comment? I think it's enough to consider only parensed. If a non-symbolic identifier is parenthesized unnecessarily, the range reported in diagnostic won't include the parenthesis part, i.e. parensed will be False, so I think it doesn't matter if isSymOcc occSym is True. Correct me if I'm wrong :)

[konn]

Can we remove isSymOcc here?

Did you see my comment? I think it's enough to consider only parensed. If a non-symbolic identifier is parenthesized unnecessarily, the range reported in diagnostic won't include the parenthesis part, i.e. parensed will be False, so I think it doesn't matter if isSymOcc occSym is True. Correct me if I'm wrong :)

Oops, sorry I overlooked your comment 🙇 I think your suggestion makes sense. Thanks!

[jneira]

very happy to see more contributors with the commit bit, congrats and thanks for your great work here 🙂