Hint `Redundant <&>` suggests code obfuscation

[andreasabel]

main = do
  e <- getContents <&> pExp . myLexer >>= \case
    Left err -> do
      putStrLn err
      exitFailure
    Right e -> pure e

This triggers this linter warning:

Warning: Redundant <&>
Found:
  getContents <&> pExp . myLexer
    >>=
      \case
        Left err
          -> do putStrLn err
                exitFailure
        Right e -> pure e
Perhaps:
  getContents
    >>=
      (\case
         Left err
           -> do putStrLn err
                 exitFailure
         Right e -> pure e)
        . pExp . myLexer

Clearly, you do not want such a refactoring. (Imagine the case distinction was even larger.)

In fact the whole point of using <&> in my original expression is that I can write a monadic case in a readable way without introducing a throw-away variable.

There might be similar problems with similar hints.
They hardly take the size of expressions into account, or pipelining styles.

[googleson78]

In general, I tend to ignore all hints that suggest stuff like "fuse map into foldr" or any form of "fuse fmap into bind", etc, as they all often have similar issues

[andreasabel]

This may point to a systematic problem of hlint: it suggests mathematically correct refactorings, but code has two consumers, the compiler, and the human reader, and a refactoring must be sound for both of them.

[zliu41]

Perhaps a reasonable rule is: if the suggestion requires extra brackets, avoid it. This is however not very trivial to implement, and I'm not sure if it should always be applied.