Roff tokenizer: check for first-column before parsing macro.

Also add SourcePos as argument to lexRoff, so we can pass in
current source pos when parsing a table cell.

Closes #5025.

2 files changed, 10 insertions, 6 deletions

diff --git a/src/Text/Pandoc/Readers/Man.hs b/src/Text/Pandoc/Readers/Man.hs
index 665679c4d..3644050c7 100644
--- a/src/Text/Pandoc/Readers/Man.hs
+++ b/src/Text/Pandoc/Readers/Man.hs
@@ -52,7 +52,7 @@ import Text.Pandoc.Shared (crFilter)
 import Text.Pandoc.Readers.Roff  -- TODO explicit imports
 import Text.Parsec hiding (tokenPrim)
 import qualified Text.Parsec as Parsec
-import Text.Parsec.Pos (updatePosString)
+import Text.Parsec.Pos (updatePosString, initialPos)
 import qualified Data.Foldable as Foldable
 
 data ManState = ManState { readerOptions :: ReaderOptions
@@ -69,7 +69,7 @@ type ManParser m = ParserT [RoffToken] ManState m
 -- | Read man (troff) from an input string and return a Pandoc document.
 readMan :: PandocMonad m => ReaderOptions -> T.Text -> m Pandoc
 readMan opts txt = do
-  tokenz <- lexRoff (crFilter txt)
+  tokenz <- lexRoff (initialPos "input") (crFilter txt)
   let state = def {readerOptions = opts} :: ManState
   eitherdoc <- readWithMTokens parseMan state
      (Foldable.toList . unRoffTokens $ tokenz)
diff --git a/src/Text/Pandoc/Readers/Roff.hs b/src/Text/Pandoc/Readers/Roff.hs
index a98678a30..e83821ed4 100644
--- a/src/Text/Pandoc/Readers/Roff.hs
+++ b/src/Text/Pandoc/Readers/Roff.hs
@@ -311,6 +311,7 @@ lexComment = do
 lexMacro :: PandocMonad m => RoffLexer m RoffTokens
 lexMacro = do
   pos <- getPosition
+  guard $ sourceColumn pos == 1
   char '.' <|> char '\''
   skipMany spacetab
   macroName <- many (satisfy (not . isSpace))
@@ -369,7 +370,9 @@ lexTableRows = do
   return $ zip aligns rows
 
 tableCell :: PandocMonad m => RoffLexer m RoffTokens
-tableCell = (enclosedCell <|> simpleCell) >>= lexRoff . T.pack
+tableCell = do
+  pos <- getPosition
+  (enclosedCell <|> simpleCell) >>= lexRoff pos . T.pack
   where
   enclosedCell = do
     try (string "T{")
@@ -642,9 +645,10 @@ linePartsToString = mconcat . map go
   go _ = mempty
 
 -- | Tokenize a string as a sequence of groff tokens.
-lexRoff :: PandocMonad m => T.Text -> m RoffTokens
-lexRoff txt = do
-  eithertokens <- readWithM (mconcat <$> many manToken) def (T.unpack txt)
+lexRoff :: PandocMonad m => SourcePos -> T.Text -> m RoffTokens
+lexRoff pos txt = do
+  eithertokens <- readWithM (do setPosition pos
+                                mconcat <$> many manToken) def (T.unpack txt)
   case eithertokens of
     Left e       -> throwError e
     Right tokenz -> return tokenz