Remove empty haddocks and remove empty comments.

[philderbeast]

Can we suggest removing empty comments and removing empty haddocks? I've seen examples of these littering code, like the following:

-- |
foo = 1

I had a crack at doing this.

• If I run cabal run hlint -- --refactor tests/empty-comment-test.hs the comments are removed but single line comments are replaced by an empty line. I'd really like to remove the whole line. Is there a way to do that? I tried advanceSrcLoc (realSrcSpanEnd realSpan) '\n' for the refactor span but this didn't work as I'd hoped.
• I'm not quite sure what is the best way to convert tests/empty-comment-test.hs into tests.

$ cabal run hlint -- --refactor --refactor-options=inplace tests/empty-comment-test.hs
$ git diff

diff --git a/tests/empty-comment-test.hs b/tests/empty-comment-test.hs
index 33b38157..84811b24 100644
--- a/tests/empty-comment-test.hs
+++ b/tests/empty-comment-test.hs
@@ -1,11 +1,11 @@
--- |
+
 haddockAboveSingle = 1
 
 haddockBelowSingle = 1
--- ^
 
--- |
---
+
+
+
 haddockAboveSingles = 1
 
 -- |
@@ -24,51 +24,45 @@ haddockBelowIntoNonEmpty = 1
 --
 -- foo
 
-{--}
+
 commentMultiEmpty = 1
 
-{- -}
+
 commentMultiEmptySpace = 1
 
-{-
 
-    -}
 commentMultiEmptyLines = 1
 
-{- |-}
+
 haddockAboveMultiEmpty = 1
 
-{- | -}
+
 haddockAboveMultiEmptySpace = 1
 
-{- |
 
-    -}
 haddockAboveMultiEmptyLines = 1
 
 haddockBelowMultiEmpty = 1
-{- ^-}
+
 
 haddockBelowMultiEmptySpace = 1
-{- ^-}
+
 
 haddockBelowMultiEmptyLines = 1
-{- ^
 
-    -}
 
 data HaddockSingle = HaddockSingle
   { haddockFieldEmptyBelow :: Int
-  -- ^
-  -- |
+  
+  
   , haddockFieldAboveEmpty :: Int
-  , haddockFieldEmptyAfter :: Int -- ^
+  , haddockFieldEmptyAfter :: Int 
   }
 
 data HaddockMulti = HaddockMulti
   { haddockFieldEmptyBelow :: Int
-  {- ^-}
-  {- |-}
+  
+  
   , haddockFieldAboveEmpty :: Int
-  , haddockFieldEmptyAfter :: Int {- ^-}
-  }
\ No newline at end of file
+  , haddockFieldEmptyAfter :: Int 
+  }

[ndmitchell]

No idea on the refactor thing - @zliu41 is the expert there.

Can you turn your test script into something that runs via the .test files in the tests directory?

[philderbeast]

Can you turn your test script into something that runs via the .test files in the tests directory?

Thanks for the suggestion @ndmitchell, I've done that now.

[ndmitchell]

I note this is still in draft - is there something else you are hoping to do here before I take a look?

[philderbeast]

Thanks @ndmitchell but this is not ready for review (and apologies for reworking it on this PR branch and not on another branch).

I had an initial implementation that detected simple empty comments, like these;

• the empty multi-line haddock comment.
• the empty single line haddock comment or run of comments.

As I did more testing I found more types of empty comment that I'd like to include, doctest comments in particular;

• the empty doctest leader >>>
• the empty multi-line doctest :{ .. :}

If an otherwise empty haddock comment includes a empty doctest comment then which new HLint suggestion should take precedence or should I report both?

-- |
-- >>>
--

I'm also considering reporting leading or trailing blank lines in haddocks. For instance, in the Agda source there are many trailing blank lines in its haddocks.

-- | A symbolic library name.
--
type LibName = String

Should there be separate HLint suggestions for trimming leading or trailing blank lines? Should the same go for trimming a run of blank lines down to single blank line in otherwise non-empty comments?

I'm in the process of redoing the implementation to first partition comment types and then do the detection as before.

In the earlier implementation I processed single line comments one at a time but I think I could provide more nuanced empty comment detection if I treated a run of comments (with a haddock leader) like a multi-line haddock comment.