aboutsummaryrefslogtreecommitdiff
path: root/src/Text/Pandoc/Readers
diff options
context:
space:
mode:
Diffstat (limited to 'src/Text/Pandoc/Readers')
-rw-r--r--src/Text/Pandoc/Readers/DocBook.hs5
-rw-r--r--src/Text/Pandoc/Readers/HTML.hs4
-rw-r--r--src/Text/Pandoc/Readers/Haddock.hs4
-rw-r--r--src/Text/Pandoc/Readers/LaTeX.hs3
-rw-r--r--src/Text/Pandoc/Readers/Markdown.hs2
-rw-r--r--src/Text/Pandoc/Readers/MediaWiki.hs5
-rw-r--r--src/Text/Pandoc/Readers/Muse.hs3
-rw-r--r--src/Text/Pandoc/Readers/OPML.hs4
-rw-r--r--src/Text/Pandoc/Readers/Org.hs3
-rw-r--r--src/Text/Pandoc/Readers/RST.hs2
-rw-r--r--src/Text/Pandoc/Readers/TWiki.hs3
-rw-r--r--src/Text/Pandoc/Readers/Textile.hs4
-rw-r--r--src/Text/Pandoc/Readers/Txt2Tags.hs6
-rw-r--r--src/Text/Pandoc/Readers/Vimwiki.hs151
14 files changed, 105 insertions, 94 deletions
diff --git a/src/Text/Pandoc/Readers/DocBook.hs b/src/Text/Pandoc/Readers/DocBook.hs
index 6108aae7f..c1e4d742c 100644
--- a/src/Text/Pandoc/Readers/DocBook.hs
+++ b/src/Text/Pandoc/Readers/DocBook.hs
@@ -1,6 +1,6 @@
module Text.Pandoc.Readers.DocBook ( readDocBook ) where
import Data.Char (toUpper)
-import Text.Pandoc.Shared (safeRead)
+import Text.Pandoc.Shared (safeRead, crFilter)
import Text.Pandoc.Options
import Text.Pandoc.Definition
import Text.Pandoc.Builder
@@ -526,7 +526,8 @@ instance Default DBState where
readDocBook :: PandocMonad m => ReaderOptions -> Text -> m Pandoc
readDocBook _ inp = do
- let tree = normalizeTree . parseXML . handleInstructions $ T.unpack inp
+ let tree = normalizeTree . parseXML . handleInstructions
+ $ T.unpack $ crFilter inp
(bs, st') <- flip runStateT (def{ dbContent = tree }) $ mapM parseBlock $ tree
return $ Pandoc (dbMeta st') (toList . mconcat $ bs)
diff --git a/src/Text/Pandoc/Readers/HTML.hs b/src/Text/Pandoc/Readers/HTML.hs
index e203298b8..301afa207 100644
--- a/src/Text/Pandoc/Readers/HTML.hs
+++ b/src/Text/Pandoc/Readers/HTML.hs
@@ -45,7 +45,7 @@ import Text.Pandoc.Definition
import qualified Text.Pandoc.Builder as B
import Text.Pandoc.Builder (Blocks, Inlines, trimInlines, HasMeta(..))
import Text.Pandoc.Shared ( extractSpaces, addMetaField
- , escapeURI, safeRead )
+ , escapeURI, safeRead, crFilter )
import Text.Pandoc.Options (ReaderOptions(readerExtensions), extensionEnabled,
Extension (Ext_epub_html_exts,
Ext_raw_html, Ext_native_divs, Ext_native_spans))
@@ -82,7 +82,7 @@ readHtml :: PandocMonad m
readHtml opts inp = do
let tags = stripPrefixes . canonicalizeTags $
parseTagsOptions parseOptions{ optTagPosition = True }
- inp
+ (crFilter inp)
parseDoc = do
blocks <- (fixPlains False) . mconcat <$> manyTill block eof
meta <- stateMeta . parserState <$> getState
diff --git a/src/Text/Pandoc/Readers/Haddock.hs b/src/Text/Pandoc/Readers/Haddock.hs
index b22b71b96..a09ed8be9 100644
--- a/src/Text/Pandoc/Readers/Haddock.hs
+++ b/src/Text/Pandoc/Readers/Haddock.hs
@@ -27,7 +27,7 @@ import Text.Pandoc.Class (PandocMonad)
import Text.Pandoc.Definition
import Text.Pandoc.Error
import Text.Pandoc.Options
-import Text.Pandoc.Shared (splitBy, trim)
+import Text.Pandoc.Shared (splitBy, trim, crFilter)
-- | Parse Haddock markup and return a 'Pandoc' document.
@@ -35,7 +35,7 @@ readHaddock :: PandocMonad m
=> ReaderOptions
-> Text
-> m Pandoc
-readHaddock opts s = case readHaddockEither opts (unpack s) of
+readHaddock opts s = case readHaddockEither opts (unpack (crFilter s)) of
Right result -> return result
Left e -> throwError e
diff --git a/src/Text/Pandoc/Readers/LaTeX.hs b/src/Text/Pandoc/Readers/LaTeX.hs
index 1ac872933..090dc5fdb 100644
--- a/src/Text/Pandoc/Readers/LaTeX.hs
+++ b/src/Text/Pandoc/Readers/LaTeX.hs
@@ -63,7 +63,8 @@ readLaTeX :: PandocMonad m
-> Text -- ^ String to parse (assumes @'\n'@ line endings)
-> m Pandoc
readLaTeX opts ltx = do
- parsed <- readWithM parseLaTeX def{ stateOptions = opts } (unpack ltx)
+ parsed <- readWithM parseLaTeX def{ stateOptions = opts }
+ (unpack (crFilter ltx))
case parsed of
Right result -> return result
Left e -> throwError e
diff --git a/src/Text/Pandoc/Readers/Markdown.hs b/src/Text/Pandoc/Readers/Markdown.hs
index edb356b39..96885c9b1 100644
--- a/src/Text/Pandoc/Readers/Markdown.hs
+++ b/src/Text/Pandoc/Readers/Markdown.hs
@@ -74,7 +74,7 @@ readMarkdown :: PandocMonad m
-> m Pandoc
readMarkdown opts s = do
parsed <- (readWithM parseMarkdown) def{ stateOptions = opts }
- (T.unpack s ++ "\n\n")
+ (T.unpack (crFilter s) ++ "\n\n")
case parsed of
Right result -> return result
Left e -> throwError e
diff --git a/src/Text/Pandoc/Readers/MediaWiki.hs b/src/Text/Pandoc/Readers/MediaWiki.hs
index e371ff152..a7f073d50 100644
--- a/src/Text/Pandoc/Readers/MediaWiki.hs
+++ b/src/Text/Pandoc/Readers/MediaWiki.hs
@@ -58,7 +58,8 @@ import Text.Pandoc.Logging
import Text.Pandoc.Options
import Text.Pandoc.Parsing hiding (nested)
import Text.Pandoc.Readers.HTML (htmlTag, isBlockTag, isCommentTag)
-import Text.Pandoc.Shared (safeRead, stringify, stripTrailingNewlines, trim)
+import Text.Pandoc.Shared (safeRead, stringify, stripTrailingNewlines, trim,
+ crFilter)
import Text.Pandoc.Walk (walk)
import Text.Pandoc.XML (fromEntities)
@@ -77,7 +78,7 @@ readMediaWiki opts s = do
, mwLogMessages = []
, mwInTT = False
}
- (unpack s ++ "\n")
+ (unpack (crFilter s) ++ "\n")
case parsed of
Right result -> return result
Left e -> throwError e
diff --git a/src/Text/Pandoc/Readers/Muse.hs b/src/Text/Pandoc/Readers/Muse.hs
index 84121cabe..7eee064a7 100644
--- a/src/Text/Pandoc/Readers/Muse.hs
+++ b/src/Text/Pandoc/Readers/Muse.hs
@@ -57,6 +57,7 @@ import Text.Pandoc.Class (PandocMonad(..))
import Text.Pandoc.Definition
import Text.Pandoc.Logging
import Text.Pandoc.Options
+import Text.Pandoc.Shared (crFilter)
import Text.Pandoc.Parsing hiding (macro, nested)
import Text.Pandoc.Readers.HTML (htmlTag)
import Text.Pandoc.XML (fromEntities)
@@ -68,7 +69,7 @@ readMuse :: PandocMonad m
-> Text
-> m Pandoc
readMuse opts s = do
- res <- readWithM parseMuse def{ stateOptions = opts } (unpack s)
+ res <- readWithM parseMuse def{ stateOptions = opts } (unpack (crFilter s))
case res of
Left e -> throwError e
Right d -> return d
diff --git a/src/Text/Pandoc/Readers/OPML.hs b/src/Text/Pandoc/Readers/OPML.hs
index e9f876525..c25ace800 100644
--- a/src/Text/Pandoc/Readers/OPML.hs
+++ b/src/Text/Pandoc/Readers/OPML.hs
@@ -9,6 +9,7 @@ import Text.HTML.TagSoup.Entity (lookupEntity)
import Text.Pandoc.Builder
import Text.Pandoc.Class (PandocMonad)
import Text.Pandoc.Options
+import Text.Pandoc.Shared (crFilter)
import Text.Pandoc.Readers.HTML (readHtml)
import Text.Pandoc.Readers.Markdown (readMarkdown)
import Text.XML.Light
@@ -32,7 +33,8 @@ instance Default OPMLState where
readOPML :: PandocMonad m => ReaderOptions -> Text -> m Pandoc
readOPML _ inp = do
(bs, st') <- flip runStateT def
- (mapM parseBlock $ normalizeTree $ parseXML (unpack inp))
+ (mapM parseBlock $ normalizeTree $
+ parseXML (unpack (crFilter inp)))
return $
setTitle (opmlDocTitle st') $
setAuthors (opmlDocAuthors st') $
diff --git a/src/Text/Pandoc/Readers/Org.hs b/src/Text/Pandoc/Readers/Org.hs
index 5e0d67d10..eaccc251c 100644
--- a/src/Text/Pandoc/Readers/Org.hs
+++ b/src/Text/Pandoc/Readers/Org.hs
@@ -36,6 +36,7 @@ import Text.Pandoc.Definition
import Text.Pandoc.Error
import Text.Pandoc.Options
import Text.Pandoc.Parsing (reportLogMessages)
+import Text.Pandoc.Shared (crFilter)
import Control.Monad.Except (throwError)
import Control.Monad.Reader (runReaderT)
@@ -51,7 +52,7 @@ readOrg :: PandocMonad m
readOrg opts s = do
parsed <- flip runReaderT def $
readWithM parseOrg (optionsToParserState opts)
- (T.unpack s ++ "\n\n")
+ (T.unpack (crFilter s) ++ "\n\n")
case parsed of
Right result -> return result
Left _ -> throwError $ PandocParseError "problem parsing org"
diff --git a/src/Text/Pandoc/Readers/RST.hs b/src/Text/Pandoc/Readers/RST.hs
index fb5f6f2d4..d13f697b7 100644
--- a/src/Text/Pandoc/Readers/RST.hs
+++ b/src/Text/Pandoc/Readers/RST.hs
@@ -68,7 +68,7 @@ readRST :: PandocMonad m
-> m Pandoc
readRST opts s = do
parsed <- (readWithM parseRST) def{ stateOptions = opts }
- (T.unpack s ++ "\n\n")
+ (T.unpack (crFilter s) ++ "\n\n")
case parsed of
Right result -> return result
Left e -> throwError e
diff --git a/src/Text/Pandoc/Readers/TWiki.hs b/src/Text/Pandoc/Readers/TWiki.hs
index 91ee8d1f1..210d3e5aa 100644
--- a/src/Text/Pandoc/Readers/TWiki.hs
+++ b/src/Text/Pandoc/Readers/TWiki.hs
@@ -48,6 +48,7 @@ import Text.Pandoc.Options
import Text.Pandoc.Parsing hiding (enclosed, macro, nested)
import Text.Pandoc.Readers.HTML (htmlTag, isCommentTag)
import Text.Pandoc.XML (fromEntities)
+import Text.Pandoc.Shared (crFilter)
import Data.Text (Text)
import qualified Data.Text as T
@@ -58,7 +59,7 @@ readTWiki :: PandocMonad m
-> m Pandoc
readTWiki opts s = do
res <- readWithM parseTWiki def{ stateOptions = opts }
- (T.unpack s ++ "\n\n")
+ (T.unpack (crFilter s) ++ "\n\n")
case res of
Left e -> throwError e
Right d -> return d
diff --git a/src/Text/Pandoc/Readers/Textile.hs b/src/Text/Pandoc/Readers/Textile.hs
index 96b51feef..a80d75340 100644
--- a/src/Text/Pandoc/Readers/Textile.hs
+++ b/src/Text/Pandoc/Readers/Textile.hs
@@ -68,7 +68,7 @@ import Text.Pandoc.Options
import Text.Pandoc.Parsing
import Text.Pandoc.Readers.HTML (htmlTag, isBlockTag, isInlineTag)
import Text.Pandoc.Readers.LaTeX (rawLaTeXBlock, rawLaTeXInline)
-import Text.Pandoc.Shared (trim)
+import Text.Pandoc.Shared (trim, crFilter)
import Data.Text (Text)
import qualified Data.Text as T
@@ -79,7 +79,7 @@ readTextile :: PandocMonad m
-> m Pandoc
readTextile opts s = do
parsed <- readWithM parseTextile def{ stateOptions = opts }
- (T.unpack s ++ "\n\n")
+ (T.unpack (crFilter s) ++ "\n\n")
case parsed of
Right result -> return result
Left e -> throwError e
diff --git a/src/Text/Pandoc/Readers/Txt2Tags.hs b/src/Text/Pandoc/Readers/Txt2Tags.hs
index 260bb7fff..5708358f6 100644
--- a/src/Text/Pandoc/Readers/Txt2Tags.hs
+++ b/src/Text/Pandoc/Readers/Txt2Tags.hs
@@ -41,7 +41,7 @@ import qualified Text.Pandoc.Builder as B
import Text.Pandoc.Definition
import Text.Pandoc.Options
import Text.Pandoc.Parsing hiding (macro, space, spaces, uri)
-import Text.Pandoc.Shared (compactify, compactifyDL, escapeURI)
+import Text.Pandoc.Shared (compactify, compactifyDL, escapeURI, crFilter)
import Control.Monad (guard, void, when)
import Control.Monad.Reader (Reader, asks, runReader)
import Data.Default
@@ -95,7 +95,9 @@ readTxt2Tags :: PandocMonad m
-> m Pandoc
readTxt2Tags opts s = do
meta <- getT2TMeta
- let parsed = flip runReader meta $ readWithM parseT2T (def {stateOptions = opts}) (T.unpack s ++ "\n\n")
+ let parsed = flip runReader meta $
+ readWithM parseT2T (def {stateOptions = opts}) $
+ T.unpack (crFilter s) ++ "\n\n"
case parsed of
Right result -> return $ result
Left e -> throwError e
diff --git a/src/Text/Pandoc/Readers/Vimwiki.hs b/src/Text/Pandoc/Readers/Vimwiki.hs
index 0cfbec34d..98f04eda9 100644
--- a/src/Text/Pandoc/Readers/Vimwiki.hs
+++ b/src/Text/Pandoc/Readers/Vimwiki.hs
@@ -33,10 +33,10 @@ Conversion of vimwiki text to 'Pandoc' document.
* [X] header
* [X] hrule
* [X] comment
- * [X] blockquote
- * [X] preformatted
- * [X] displaymath
- * [X] bulletlist / orderedlist
+ * [X] blockquote
+ * [X] preformatted
+ * [X] displaymath
+ * [X] bulletlist / orderedlist
* [X] orderedlist with 1., i., a) etc identification.
* [X] todo lists -- not list builder with attributes? using span.
* [X] table
@@ -57,8 +57,8 @@ Conversion of vimwiki text to 'Pandoc' document.
* [X] sub- and super-scripts
* misc:
* [X] `TODO:` mark
- * [X] metadata placeholders: %title and %date
- * [O] control placeholders: %template and %nohtml -- %template added to
+ * [X] metadata placeholders: %title and %date
+ * [O] control placeholders: %template and %nohtml -- %template added to
meta, %nohtml ignored
--}
@@ -66,29 +66,29 @@ module Text.Pandoc.Readers.Vimwiki ( readVimwiki
) where
import Control.Monad.Except (throwError)
import Control.Monad (guard)
-import Data.Default
+import Data.Default
import Data.Maybe
import Data.Monoid ((<>))
import Data.List (isInfixOf, isPrefixOf)
import Data.Text (Text, unpack)
import Text.Pandoc.Builder (Blocks, Inlines, trimInlines, fromList, toList)
-import qualified Text.Pandoc.Builder
- as B (headerWith, str, space, strong, emph, strikeout, code, link, image,
- spanWith, para, horizontalRule, blockQuote, bulletList, plain,
- orderedList, simpleTable, softbreak, codeBlockWith, imageWith, divWith,
+import qualified Text.Pandoc.Builder
+ as B (headerWith, str, space, strong, emph, strikeout, code, link, image,
+ spanWith, para, horizontalRule, blockQuote, bulletList, plain,
+ orderedList, simpleTable, softbreak, codeBlockWith, imageWith, divWith,
setMeta, definitionList, superscript, subscript)
import Text.Pandoc.Class (PandocMonad(..))
-import Text.Pandoc.Definition (Pandoc(..), Inline(Space),
- Block(BulletList, OrderedList), Attr, nullMeta, Meta, ListNumberStyle(..),
+import Text.Pandoc.Definition (Pandoc(..), Inline(Space),
+ Block(BulletList, OrderedList), Attr, nullMeta, Meta, ListNumberStyle(..),
ListNumberDelim(..))
import Text.Pandoc.Options (ReaderOptions)
-import Text.Pandoc.Parsing (readWithM, ParserT, stateOptions, ParserState,
+import Text.Pandoc.Parsing (readWithM, ParserT, stateOptions, ParserState,
stateMeta', blanklines, registerHeader, spaceChar, emailAddress, uri, F, runF,
orderedListMarker, many1Till)
-import Text.Pandoc.Shared (splitBy, stripFirstAndLast, stringify)
-import Text.Parsec.Char (spaces, char, anyChar, newline, string, noneOf,
+import Text.Pandoc.Shared (splitBy, stripFirstAndLast, stringify, crFilter)
+import Text.Parsec.Char (spaces, char, anyChar, newline, string, noneOf,
alphaNum)
-import Text.Parsec.Combinator (eof, choice, many1, manyTill, count, skipMany1,
+import Text.Parsec.Combinator (eof, choice, many1, manyTill, count, skipMany1,
notFollowedBy, option)
import Text.Parsec.Prim (many, try, updateState, getState)
import Text.Parsec.Char (oneOf, space)
@@ -97,7 +97,8 @@ import Text.Parsec.Prim ((<|>))
readVimwiki :: PandocMonad m => ReaderOptions -> Text -> m Pandoc
readVimwiki opts s = do
- res <- readWithM parseVimwiki def{ stateOptions = opts } (unpack s)
+ res <- readWithM parseVimwiki def{ stateOptions = opts }
+ (unpack (crFilter s))
case res of
Left e -> throwError e
Right result -> return result
@@ -110,7 +111,7 @@ type VwParser = ParserT [Char] ParserState
specialChars :: [Char]
specialChars = "=*-#[]_~{}`$|:%^,"
-spaceChars :: [Char]
+spaceChars :: [Char]
spaceChars = " \t\n"
-- main parser
@@ -134,7 +135,7 @@ block = do
, mempty <$ comment
, mixedList
, preformatted
- , displayMath
+ , displayMath
, table
, mempty <$ placeholder
, blockQuote
@@ -149,14 +150,14 @@ blockML = choice [preformatted, displayMath, table]
header :: PandocMonad m => VwParser m Blocks
header = try $ do
- sp <- many spaceChar
+ sp <- many spaceChar
eqs <- many1 (char '=')
spaceChar
let lev = length eqs
guard $ lev <= 6
- contents <- trimInlines . mconcat <$> manyTill inline (try $ spaceChar
- >> (string eqs) >> many spaceChar >> newline)
- attr <- registerHeader (makeId contents,
+ contents <- trimInlines . mconcat <$> manyTill inline (try $ spaceChar
+ >> (string eqs) >> many spaceChar >> newline)
+ attr <- registerHeader (makeId contents,
(if sp == "" then [] else ["justcenter"]), []) contents
return $ B.headerWith attr lev contents
@@ -184,7 +185,7 @@ blockQuote = try $ do
else return $ B.blockQuote $ B.plain contents
definitionList :: PandocMonad m => VwParser m Blocks
-definitionList = try $
+definitionList = try $
B.definitionList <$> (many1 (dlItemWithDT <|> dlItemWithoutDT))
dlItemWithDT :: PandocMonad m => VwParser m (Inlines, [Blocks])
@@ -199,15 +200,15 @@ dlItemWithoutDT = do
return $ (mempty, dds)
definitionDef :: PandocMonad m => VwParser m Blocks
-definitionDef = try $
- (notFollowedBy definitionTerm) >> many spaceChar
+definitionDef = try $
+ (notFollowedBy definitionTerm) >> many spaceChar
>> (definitionDef1 <|> definitionDef2)
definitionDef1 :: PandocMonad m => VwParser m Blocks
definitionDef1 = try $ mempty <$ defMarkerE
definitionDef2 :: PandocMonad m => VwParser m Blocks
-definitionDef2 = try $ B.plain <$>
+definitionDef2 = try $ B.plain <$>
(defMarkerM >> (trimInlines . mconcat <$> many inline') <* newline)
@@ -218,11 +219,11 @@ definitionTerm = try $ do
return x
definitionTerm1 :: PandocMonad m => VwParser m Inlines
-definitionTerm1 = try $
+definitionTerm1 = try $
trimInlines . mconcat <$> manyTill inline' (try $ defMarkerE)
definitionTerm2 :: PandocMonad m => VwParser m Inlines
-definitionTerm2 = try $ trimInlines . mconcat <$> manyTill inline'
+definitionTerm2 = try $ trimInlines . mconcat <$> manyTill inline'
(try $ lookAhead $ (defMarkerM >> notFollowedBy hasDefMarkerM))
defMarkerM :: PandocMonad m => VwParser m Char
@@ -236,8 +237,8 @@ hasDefMarkerM = manyTill (noneOf "\n") (try defMarkerM)
preformatted :: PandocMonad m => VwParser m Blocks
preformatted = try $ do
- many spaceChar >> string "{{{"
- attrText <- many (noneOf "\n")
+ many spaceChar >> string "{{{"
+ attrText <- many (noneOf "\n")
lookAhead newline
contents <- manyTill anyChar (try (char '\n' >> many spaceChar >> string "}}}"
>> many spaceChar >> newline))
@@ -246,14 +247,14 @@ preformatted = try $ do
else return $ B.codeBlockWith (makeAttr attrText) contents
makeAttr :: String -> Attr
-makeAttr s =
+makeAttr s =
let xs = splitBy (`elem` " \t") s in
("", [], catMaybes $ map nameValue xs)
nameValue :: String -> Maybe (String, String)
-nameValue s =
+nameValue s =
let t = splitBy (== '=') s in
- if length t /= 2
+ if length t /= 2
then Nothing
else let (a, b) = (head t, last t) in
if ((length b) < 2) || ((head b, last b) /= ('"', '"'))
@@ -269,7 +270,7 @@ displayMath = try $ do
>> many spaceChar >> newline))
let contentsWithTags
| mathTag == "" = "\\[" ++ contents ++ "\n\\]"
- | otherwise = "\\begin{" ++ mathTag ++ "}" ++ contents
+ | otherwise = "\\begin{" ++ mathTag ++ "}" ++ contents
++ "\n\\end{" ++ mathTag ++ "}"
return $ B.plain $ B.str contentsWithTags
@@ -286,7 +287,7 @@ mixedList' prevInd = do
else do
listStart
curLine <- listItemContent
- let listBuilder =
+ let listBuilder =
if builder == "ul" then B.bulletList else B.orderedList
(subList, lowInd) <- (mixedList' curInd)
if lowInd >= curInd
@@ -297,7 +298,7 @@ mixedList' prevInd = do
then return ([listBuilder curList], endInd)
else return (curList, endInd)
else do
- let (curList, endInd) = ((combineList curLine subList),
+ let (curList, endInd) = ((combineList curLine subList),
lowInd)
if curInd > prevInd
then return ([listBuilder curList], endInd)
@@ -328,13 +329,13 @@ blocksThenInline = try $ do
return $ mconcat $ y ++ [x]
listTodoMarker :: PandocMonad m => VwParser m Inlines
-listTodoMarker = try $ do
- x <- between (many spaceChar >> char '[') (char ']' >> spaceChar)
+listTodoMarker = try $ do
+ x <- between (many spaceChar >> char '[') (char ']' >> spaceChar)
(oneOf " .oOX")
return $ makeListMarkerSpan x
makeListMarkerSpan :: Char -> Inlines
-makeListMarkerSpan x =
+makeListMarkerSpan x =
let cl = case x of
' ' -> "done0"
'.' -> "done1"
@@ -347,9 +348,9 @@ makeListMarkerSpan x =
combineList :: Blocks -> [Blocks] -> [Blocks]
combineList x [y] = case toList y of
- [BulletList z] -> [fromList $ (toList x)
+ [BulletList z] -> [fromList $ (toList x)
++ [BulletList z]]
- [OrderedList attr z] -> [fromList $ (toList x)
+ [OrderedList attr z] -> [fromList $ (toList x)
++ [OrderedList attr z]]
_ -> x:[y]
combineList x xs = x:xs
@@ -365,9 +366,9 @@ bulletListMarkers :: PandocMonad m => VwParser m String
bulletListMarkers = "ul" <$ (char '*' <|> char '-')
orderedListMarkers :: PandocMonad m => VwParser m String
-orderedListMarkers =
- ("ol" <$ (choice $ (orderedListMarker Decimal Period):(($OneParen)
- <$> orderedListMarker
+orderedListMarkers =
+ ("ol" <$ (choice $ (orderedListMarker Decimal Period):(($OneParen)
+ <$> orderedListMarker
<$> [Decimal, LowerRoman, UpperRoman, LowerAlpha, UpperAlpha])))
<|> ("ol" <$ char '#')
@@ -397,14 +398,14 @@ table2 = try $ do
tableHeaderSeparator :: PandocMonad m => VwParser m ()
tableHeaderSeparator = try $ do
- many spaceChar >> char '|' >> many1 ((many1 $ char '-') >> char '|')
+ many spaceChar >> char '|' >> many1 ((many1 $ char '-') >> char '|')
>> many spaceChar >> newline
return ()
-
+
tableRow :: PandocMonad m => VwParser m [Blocks]
tableRow = try $ do
many spaceChar >> char '|'
- s <- lookAhead $ manyTill anyChar (try (char '|' >> many spaceChar
+ s <- lookAhead $ manyTill anyChar (try (char '|' >> many spaceChar
>> newline))
guard $ not $ "||" `isInfixOf` ("|" ++ s ++ "|")
tr <- many tableCell
@@ -416,25 +417,25 @@ tableCell = try $
B.plain <$> trimInlines . mconcat <$> (manyTill inline' (char '|'))
placeholder :: PandocMonad m => VwParser m ()
-placeholder = try $
+placeholder = try $
(choice (ph <$> ["title", "date"])) <|> noHtmlPh <|> templatePh
ph :: PandocMonad m => String -> VwParser m ()
ph s = try $ do
many spaceChar >> (string $ '%':s) >> spaceChar
- contents <- (trimInlines . mconcat <$> (manyTill inline (lookAhead newline)))
+ contents <- (trimInlines . mconcat <$> (manyTill inline (lookAhead newline)))
--use lookAhead because of placeholder in the whitespace parser
let meta' = return $ B.setMeta s contents nullMeta :: F Meta
updateState $ \st -> st { stateMeta' = stateMeta' st <> meta' }
noHtmlPh :: PandocMonad m => VwParser m ()
noHtmlPh = try $
- () <$ (many spaceChar >> string "%nohtml" >> many spaceChar
+ () <$ (many spaceChar >> string "%nohtml" >> many spaceChar
>> (lookAhead newline))
templatePh :: PandocMonad m => VwParser m ()
templatePh = try $
- () <$ (many spaceChar >> string "%template" >> (many $ noneOf "\n")
+ () <$ (many spaceChar >> string "%template" >> (many $ noneOf "\n")
>> (lookAhead newline))
-- inline parser
@@ -475,7 +476,7 @@ str :: PandocMonad m => VwParser m Inlines
str = B.str <$> (many1 $ noneOf $ spaceChars ++ specialChars)
whitespace :: PandocMonad m => VwParser m () -> VwParser m Inlines
-whitespace endline = B.space <$ (skipMany1 spaceChar <|>
+whitespace endline = B.space <$ (skipMany1 spaceChar <|>
(try (newline >> (comment <|> placeholder))))
<|> B.softbreak <$ endline
@@ -493,24 +494,24 @@ bareURL = try $ do
strong :: PandocMonad m => VwParser m Inlines
strong = try $ do
s <- lookAhead $ between (char '*') (char '*') (many1 $ noneOf "*")
- guard $ (not $ (head s) `elem` spaceChars)
+ guard $ (not $ (head s) `elem` spaceChars)
&& (not $ (last s) `elem` spaceChars)
char '*'
- contents <- mconcat <$> (manyTill inline' $ char '*'
+ contents <- mconcat <$> (manyTill inline' $ char '*'
>> notFollowedBy alphaNum)
- return $ (B.spanWith ((makeId contents), [], []) mempty)
+ return $ (B.spanWith ((makeId contents), [], []) mempty)
<> (B.strong contents)
-makeId :: Inlines -> String
+makeId :: Inlines -> String
makeId i = concat (stringify <$> (toList i))
emph :: PandocMonad m => VwParser m Inlines
emph = try $ do
s <- lookAhead $ between (char '_') (char '_') (many1 $ noneOf "_")
- guard $ (not $ (head s) `elem` spaceChars)
+ guard $ (not $ (head s) `elem` spaceChars)
&& (not $ (last s) `elem` spaceChars)
char '_'
- contents <- mconcat <$> (manyTill inline' $ char '_'
+ contents <- mconcat <$> (manyTill inline' $ char '_'
>> notFollowedBy alphaNum)
return $ B.emph contents
@@ -532,32 +533,32 @@ superscript = try $
subscript :: PandocMonad m => VwParser m Inlines
subscript = try $
- B.subscript <$> mconcat <$> (string ",,"
+ B.subscript <$> mconcat <$> (string ",,"
>> many1Till inline' (try $ string ",,"))
link :: PandocMonad m => VwParser m Inlines
-link = try $ do
+link = try $ do
string "[["
contents <- lookAhead $ manyTill anyChar (string "]]")
- case '|' `elem` contents of
+ case '|' `elem` contents of
False -> do
- manyTill anyChar (string "]]")
+ manyTill anyChar (string "]]")
-- not using try here because [[hell]o]] is not rendered as a link in vimwiki
return $ B.link (procLink contents) "" (B.str contents)
- True -> do
+ True -> do
url <- manyTill anyChar $ char '|'
lab <- mconcat <$> (manyTill inline $ string "]]")
return $ B.link (procLink url) "" lab
image :: PandocMonad m => VwParser m Inlines
-image = try $ do
+image = try $ do
string "{{"
contentText <- lookAhead $ manyTill (noneOf "\n") (try $ string "}}")
images $ length $ filter (== '|') contentText
images :: PandocMonad m => Int -> VwParser m Inlines
images k
- | k == 0 = do
+ | k == 0 = do
imgurl <- manyTill anyChar (try $ string "}}")
return $ B.image (procImgurl imgurl) "" (B.str "")
| k == 1 = do
@@ -578,15 +579,15 @@ images k
procLink' :: String -> String
procLink' s
- | ((take 6 s) == "local:") = "file" ++ (drop 5 s)
+ | ((take 6 s) == "local:") = "file" ++ (drop 5 s)
| ((take 6 s) == "diary:") = "diary/" ++ (drop 6 s) ++ ".html"
- | or ((`isPrefixOf` s) <$> [ "http:", "https:", "ftp:", "file:", "mailto:",
- "news:", "telnet:" ])
+ | or ((`isPrefixOf` s) <$> [ "http:", "https:", "ftp:", "file:", "mailto:",
+ "news:", "telnet:" ])
= s
| s == "" = ""
| (last s) == '/' = s
| otherwise = s ++ ".html"
-
+
procLink :: String -> String
procLink s = procLink' x ++ y
where (x, y) = break (=='#') s
@@ -606,7 +607,7 @@ tag = try $ do
s <- manyTill (noneOf spaceChars) (try (char ':' >> (lookAhead space)))
guard $ not $ "::" `isInfixOf` (":" ++ s ++ ":")
let ss = splitBy (==':') s
- return $ mconcat $ (makeTagSpan' $ head ss):(makeTagSpan <$> (tail ss))
+ return $ mconcat $ (makeTagSpan' $ head ss):(makeTagSpan <$> (tail ss))
todoMark :: PandocMonad m => VwParser m Inlines
todoMark = try $ do
@@ -623,7 +624,7 @@ endlineBQ = () <$ try (newline <* nFBTTBSB <* string " ")
endlineML :: PandocMonad m => VwParser m ()
endlineML = () <$ try (newline <* nFBTTBSB <* many1 spaceChar)
---- nFBTTBSB is short for notFollowedByThingsThatBreakSoftBreaks
+--- nFBTTBSB is short for notFollowedByThingsThatBreakSoftBreaks
nFBTTBSB :: PandocMonad m => VwParser m ()
nFBTTBSB =
notFollowedBy newline <*
@@ -639,7 +640,7 @@ hasDefMarker :: PandocMonad m => VwParser m ()
hasDefMarker = () <$ (manyTill (noneOf "\n") (string "::" >> oneOf spaceChars))
makeTagSpan' :: String -> Inlines
-makeTagSpan' s = B.spanWith ('-' : s, [], []) (B.str "") <>
+makeTagSpan' s = B.spanWith ('-' : s, [], []) (B.str "") <>
B.spanWith (s, ["tag"], []) (B.str s)
makeTagSpan :: String -> Inlines
@@ -647,7 +648,7 @@ makeTagSpan s = (B.space) <> (makeTagSpan' s)
mathTagParser :: PandocMonad m => VwParser m String
mathTagParser = do
- s <- try $ lookAhead (char '%' >> (manyTill (noneOf spaceChars)
+ s <- try $ lookAhead (char '%' >> (manyTill (noneOf spaceChars)
(try $ char '%' >> many (noneOf $ '%':spaceChars) >> space)))
char '%' >> string s >> char '%'
return s