Docx reader: improve index reference support.

Support crossrefs.
Clean up and unify switch parsing for fields.

2 files changed, 28 insertions, 27 deletions

diff --git a/src/Text/Pandoc/Readers/Docx.hs b/src/Text/Pandoc/Readers/Docx.hs
index 16451aaed..8bd4aa8b5 100644
--- a/src/Text/Pandoc/Readers/Docx.hs
+++ b/src/Text/Pandoc/Readers/Docx.hs
@@ -472,7 +472,10 @@ parPartToInlines' (OMathPara exps) =
 parPartToInlines' (Field info children) =
   case info of
     HyperlinkField url -> parPartToInlines' $ ExternalHyperLink url children
-    IndexrefField entry -> pure $ spanWith ("",["indexref"],[("entry",entry)]) mempty
+    IndexrefField entry mbsee ->
+      pure $ spanWith ("",["indexref"],
+                           (("entry",entry) :
+                             maybe [] (\x -> [("crossref",x)]) mbsee)) mempty
     PagerefField fieldAnchor True -> parPartToInlines' $ InternalHyperLink fieldAnchor children
     EndNoteCite t -> do
       formattedCite <- smushInlines <$> mapM parPartToInlines' children
diff --git a/src/Text/Pandoc/Readers/Docx/Fields.hs b/src/Text/Pandoc/Readers/Docx/Fields.hs
index 4a9f88d46..f21ad425e 100644
--- a/src/Text/Pandoc/Readers/Docx/Fields.hs
+++ b/src/Text/Pandoc/Readers/Docx/Fields.hs
@@ -25,7 +25,7 @@ type Anchor = T.Text
 data FieldInfo = HyperlinkField URL
                 -- The boolean indicates whether the field is a hyperlink.
                | PagerefField Anchor Bool
-               | IndexrefField T.Text
+               | IndexrefField T.Text (Maybe T.Text) -- second is optional 'see'
                | CslCitation T.Text
                | CslBibliography
                | EndNoteCite T.Text
@@ -45,7 +45,7 @@ fieldInfo = do
     <|>
     ((uncurry PagerefField) <$> pageref)
     <|>
-    (IndexrefField <$> indexref)
+    ((uncurry IndexrefField) <$> indexref)
     <|>
     addIn
     <|>
@@ -97,49 +97,47 @@ unquotedString :: Parser T.Text
 unquotedString = T.pack <$> manyTill anyChar (try $ void (lookAhead space) <|> eof)
 
 fieldArgument :: Parser T.Text
-fieldArgument = quotedString <|> unquotedString
-
--- there are other switches, but this is the only one I've seen in the wild so far, so it's the first one I'll implement. See §17.16.5.25
-hyperlinkSwitch :: Parser (T.Text, T.Text)
-hyperlinkSwitch = do
-  sw <- string "\\l"
-  spaces
-  farg <- fieldArgument
-  return (T.pack sw, farg)
+fieldArgument = do
+  notFollowedBy (char '\\') -- switch
+  quotedString <|> unquotedString
 
 hyperlink :: Parser URL
 hyperlink = do
   string "HYPERLINK"
   spaces
   farg <- option "" $ notFollowedBy (char '\\') *> fieldArgument
-  switches <- spaces *> many hyperlinkSwitch
-  let url = case switches of
-              ("\\l", s) : _ -> farg <> "#" <> s
-              _              -> farg
+  switches <- many fieldSwitch
+  let url = case [s | ('l',s) <- switches] of
+              [s] -> farg <> "#" <> s
+              _   -> farg
   return url
 
 -- See §17.16.5.45
-pagerefSwitch :: Parser (T.Text, T.Text)
-pagerefSwitch = do
-  sw <- string "\\h"
+fieldSwitch :: Parser (Char, T.Text)
+fieldSwitch = try $ do
+  spaces
+  char '\\'
+  c <- anyChar
   spaces
   farg <- fieldArgument
-  return (T.pack sw, farg)
+  return (c, farg)
 
 pageref :: Parser (Anchor, Bool)
 pageref = do
   string "PAGEREF"
   spaces
   farg <- fieldArgument
-  switches <- spaces *> many pagerefSwitch
-  let isLink = case switches of
-              ("\\h", _) : _ -> True
-              _              -> False
+  switches <- many fieldSwitch
+  let isLink = any ((== 'h') . fst) switches
   return (farg, isLink)
 
-indexref :: Parser T.Text
+-- second element of tuple is optional "see".
+indexref :: Parser (T.Text, Maybe T.Text)
 indexref = do
   string "XE"
   spaces
-  fieldArgument
-
+  farg <- fieldArgument
+  switches <- spaces *> many fieldSwitch
+  case [see | ('t', see) <- switches] of
+     [see] -> pure (farg, Just see)
+     _     -> pure (farg, Nothing)