Never fully collapse string gaps

Author: amesgen

CHANGELOG.md

@@ -1,3 +1,9 @@
+## Unreleased
+
+* Correctly format edge cases where fully collapsing string gaps changes the
+  string represented by a string literal. [Issue
+  1160](https://github.com/tweag/ormolu/issues/1160).
+
 ## Ormolu 0.8.0.0
 
 * Format multiple files in parallel. [Issue

data/examples/declaration/value/function/multiline-strings-0-out.hs

@@ -7,7 +7,7 @@ s =
   """
 
 s_2 =
-  """Line 1
+  """\ \Line 1
      Line 2
   Line 3
   """

data/examples/declaration/value/function/multiline-strings-1-out.hs

@@ -2,9 +2,11 @@
 
 s =
   """
-  a b c d e
+  a b\ \ c d e
   f g
   """
 
 -- equivalent to
 s' = "a b c d e\nf g"
+
+weirdGap = """\65\ \0"""

data/examples/declaration/value/function/multiline-strings-1.hs

@@ -9,3 +9,5 @@ s =
 
 -- equivalent to
 s' = "a b c d e\nf g"
+
+weirdGap = """\65\ \0"""

data/examples/declaration/value/function/strings-out.hs

@@ -2,9 +2,11 @@
 
 foo = "foobar"
 
-bar = "foo\&barbaz"
+bar = "foo\&bar\ \baz"
 
 baz =
   "foo\
   \bar\
   \baz"
+
+weirdGap = "\65\ \0"

data/examples/declaration/value/function/strings.hs

@@ -5,3 +5,5 @@ bar = "foo\&bar\   \baz"
 baz = "foo\
       \bar\
     \baz"
+
+weirdGap = "\65\ \0"

src/Ormolu/Printer/Meat/Declaration/StringLiteral.hs

@@ -28,7 +28,7 @@ p_stringLit src = case parseStringLiteral $ T.pack $ unpackFS src of
     case stringLiteralKind of
       RegularStringLiteral -> do
         let singleLine =
-              txt $ T.concat segments
+              txt $ intercalateMinimalStringGaps segments
             multiLine =
               sep breakpoint f (attachRelativePos segments)
               where
@@ -105,9 +105,9 @@ parseStringLiteral = \s -> do
     splitMultilineString :: Text -> [Text]
     splitMultilineString =
       splitGaps
-        -- There is no reason to use gaps with multiline string literals, so
-        -- we collapse them.
-        >>> T.concat
+        -- There is no reason to use gaps with multiline string literals just to
+        -- emulate multi-line strings, so we replace them with "\\ \\".
+        >>> intercalateMinimalStringGaps
         >>> splitNewlines
         >>> fmap expandLeadingTabs
         >>> rmCommonWhitespacePrefixAndBlank
@@ -150,3 +150,12 @@ parseStringLiteral = \s -> do
             | otherwise = (Just $ Min leadingSpace, T.drop commonWs l)
             where
               leadingSpace = T.length $ T.takeWhile is_space l
+
+-- | Add minimal string gaps between string literal chunks. Such string gaps
+-- /can/ be semantically meaningful (so we preserve them for simplicity); for
+-- example:
+--
+-- >>> "\65\ \0" == "\650"
+-- False
+intercalateMinimalStringGaps :: [Text] -> Text
+intercalateMinimalStringGaps = T.intercalate "\\ \\"