Best-effort support of Qualified Imports in GHC 9.4

[konn]

As discussed in #3473, we cannot provide import qualified feature in GHC 9.4 due to the regression in GHC, at least based on the textual information given by error messages.

On the other hand, we can still determine qualified module name from the information of error-range and textual source code.
As I am heavily using Import Qualified feature everyday life, so I try to follow this path to tentatively support it in GHC 9.4.
Even though there can be some corner cases that the method implemented in this PR cannot handle correctly, but even a partial support of this could be still helpful.

TODOs

• Check if this works with multibyte symbols
• (Optional) revive adding new import item to existing list along this line This cannot be handled as GHC 9.4 won't give us suggested imports.

[konn]

It seems that regression #22130 in GHC removes import suggestion for qualified symbols. This cannot be fixed even with our approach, so I decided to concentrate on reviving add qualified import feature in this PR.

[konn]

I think this PR is ready for review now.

[fendor]

LGTM! Thank you very much, this is a great addition!

Small documentation things, otherwise, gtg.

[konn]

@fendor @michaelpj Thank you for your reviews and sorry for my late response! I had been a bit busy this week. I'm starting responding to the reviews from now.

[konn]

I think I addressed all the reviews so far. Thank you for your feedbacks! I mark some threads that as resolved with some changes, but please feel free to unresolve threads if you are unsatisfied with changes.

[fendor]

LGTM, thank you very much!

Two small comments

[fendor]

Is this function used for anything else? If not, we can likely turn it into a local function (e.g. in the where block) and remove the parameter f with the thing it is supposed to do (adding a ".")

[konn]

Good point! I made it local, changed to more speic name qualify, and made it taking qualifier q instead of general mapping function f.
Addressed in e0eb90a.

[konn]

Test failure with GHC 9.2 on Win seems like a genuine regression:

Test suite func-test: RUNNING...
haskell-language-server
  code actions
    redundant import code actions
      remove solitary redundant imports: FAIL (6.48s)
        test\functional\FunctionalCodeAction.hs:259:
        expected: Nothing
         but got: Just (Command {_title = "Execute Code Action", _command = "4884:hlint:codeActionResolve", _arguments = Just [Object (fromList [("data",Object (fromList [("_uri",String "file:///D:/a/haskell-language-server/haskell-language-server/test/testdata/redundantImportTest/src/CodeActionRedundant.hs"),("_value",Object (fromList [("ignoreHintTitle",String "Use module export list"),("tag",String "IgnoreHint"),("verTxtDocId",Object (fromList [("uri",String "file:///D:/a/haskell-language-server/haskell-language-server/test/testdata/redundantImportTest/src/CodeActionRedundant.hs"),("version",Number 0.0)]))]))])),("diagnostics",Array [Object (fromList [("code",String "Use module export list"),("message",String "Use module export list\nFound:\n  module CodeActionRedundant where\nWhy not:\n  module CodeActionRedundant (\n        module CodeActionRedundant\n    ) where\nan explicit list is usually better\n"),("range",Object (fromList [("end",Object (fromList [("character",Number 0.0),("line",Number 0.0)])),("start",Object (fromList [("character",Number 0.0),("line",Number 0.0)]))])),("severity",Number 3.0),("source",String "hlint")])]),("isPreferred",Bool False),("kind",String "quickfix"),("title",String "Ignore hint \"Use module export list\" in this module")])]})

haskell-language-server/test/functional/FunctionalCodeAction.hs

Line 259 in 8c22b84

forM_ allActions $ \a -> a ^. L.command @?= Nothing

https:/haskell/haskell-language-server/actions/runs/5629658777/job/15256519280?pr=3712#step:10:211

It seems rather strange as all other versions of GHC on Win passes tests.
Does anybody guess the root cause, or is it just a flaky test?

[fendor]

I presume flaky error, I rerun the failing CI run.

[konn]

@fendor Now it seems that all Testings are green even with GHC 9.4 on Win. Thanks!

It smees Nix build on ubuntu also failed, but it seems due to the storage shortage. So I suppose everything is OK, right?

[fendor]

@konn CI and I are happy, but you'll need to rebase the PR 😅