Re-implement abbreviations in Markdown reader.

Note that with this new implementation, you can defeat the
abbreviationization by using two spaces after the period.

This commit removes support for moving abbreviations
after soft breaks (#4635), so the abbreviation support won't
work for abbreviations occuring at the end of a line.
It seems better not to mess with the user's soft breaks,
especially now that we have `--wrap=preserve`.

MANUAL.txt

@@ -691,10 +691,13 @@ header when requesting a document from a URL:
     directory or fall back on a system default.  To see the
     system default, use
     `pandoc --print-default-data-file=abbreviations`.  The only
-    use pandoc makes of this list is in the Markdown reader.
-    Strings found in this list will be followed by a nonbreaking
-    space, and the period will not produce sentence-ending space
-    in formats like LaTeX.  The strings may not contain spaces.
+    use pandoc makes of this list is in the Markdown reader,
+    and only if the `smart` extension is enabled.
+    A single space following a string on this list will be
+    transformed into a nonbreaking space (so, to defeat this
+    feature you can use two spaces after a period).
+    As a result, the period will not produce sentence-ending space
+    in formats like LaTeX.  The abbreviations may not contain spaces.
 
 [`pandocfilters`]: https://github.com/jgm/pandocfilters
 [PHP]: https://github.com/vinai/pandocfilters-php

src/Text/Pandoc/Readers/Markdown.hs

@@ -2,6 +2,7 @@
 {-# LANGUAGE TupleSections       #-}
 {-# LANGUAGE OverloadedStrings   #-}
 {-# LANGUAGE ViewPatterns        #-}
+{-# LANGUAGE FlexibleContexts    #-}
 {- |
    Module      : Text.Pandoc.Readers.Markdown
    Copyright   : Copyright (C) 2006-2021 John MacFarlane
@@ -1731,35 +1732,33 @@ nonEndline = satisfy (/='\n')
 
 str :: PandocMonad m => MarkdownParser m (F Inlines)
 str = do
-  result <- mconcat <$> many1
-             ( T.pack <$> (many1 (satisfy (\c -> isAlphaNum c ||
+  abbrevs <- getOption readerAbbreviations
+  isSmart <- extensionEnabled Ext_smart <$> getOption readerExtensions
+  let tryAbbrev t =
+         if t `Set.member` abbrevs
+            then try (do char ' ' <* notFollowedBy (char ' ')
+                         return $ t <> "\160")
+  --               <|>
+  --               try (do lookAhead newline
+  --                       guardDisabled Ext_hard_line_breaks
+  --                       guardDisabled Ext_ignore_line_breaks
+  --                       endline
+  --                       -- move soft break before abbrev (#4635)
+  --                       return $ "\n" <> t <> "\160")
+                 <|> return t
+            else return t
+  let nonSpaceChunk = do
+        t <- T.pack <$> many1 (satisfy (\c -> isAlphaNum c ||
                                                   c == ',' || c == '?' ||
                                                   c == '(' || c == ')' ||
-                                                  c == '/' )) <*
-                              updateLastStrPos)
-              <|> try (T.pack <$> many1 spaceChar <* notFollowedBy newline)
-              <|> try (T.singleton <$> char '.' <*
-                           notFollowedBy (char '.') <* updateLastStrPos) )
--- TODO: handle abbreviations as an AST transformation?
--- Then they could work on other formats, too.
---  (do guardEnabled Ext_smart
---      abbrevs <- getOption readerAbbreviations
---      if result `Set.member` abbrevs
---         then try (do ils <- whitespace
---                      notFollowedBy (() <$ cite <|> () <$ note)
---                      -- ?? lookAhead alphaNum
---                      -- replace space after with nonbreaking space
---                      -- if softbreak, move before abbrev if possible (#4635)
---                      return $ do
---                        ils' <- ils
---                        case B.toList ils' of
---                             [Space] ->
---                                 return (B.str result <> B.str "\160")
---                             _ -> return (B.str result <> ils'))
---               <|> return (return (B.str result))
---         else return (return (B.str result)))
---     <|>
-  return (return (B.str result))
+                                                  c == '/' )
+                        <|> try (char '.' <* notFollowedBy (char '.')))
+        updateLastStrPos
+        if isSmart
+           then tryAbbrev t
+           else return t
+  let spaceChunk = T.pack <$> (try (many1 spaceChar <* notFollowedBy newline))
+  return . B.text . mconcat <$> many1 (nonSpaceChunk <|> spaceChunk)
 
 -- an endline character that can be treated as a space, not a structural break
 endline :: PandocMonad m => MarkdownParser m (F Inlines)

test/command/4635.md

@@ -1,31 +0,0 @@
-```
-% pandoc -f markdown -t native
-(cf.
-foo)
-^D
-[ Para [ Str "(cf." , SoftBreak , Str "foo)" ] ]
-```
-
-```
-% pandoc -f markdown -t native
-a (cf.
-foo)
-^D
-[ Para [ Str "a (cf." , SoftBreak , Str "foo)" ] ]
-```
-
-```
-% pandoc -f markdown -t native
-cf.
-foo
-^D
-[ Para [ Str "cf." , SoftBreak , Str "foo" ] ]
-```
-
-```
-% pandoc -f markdown -t native
-a cf.
-foo
-^D
-[ Para [ Str "a cf." , SoftBreak , Str "foo" ] ]
-```