{- Copyright (C) 2006-2010 John MacFarlane This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA -} {- | Module : Text.Pandoc.Readers.HTML Copyright : Copyright (C) 2006-2010 John MacFarlane License : GNU GPL, version 2 or above Maintainer : John MacFarlane Stability : alpha Portability : portable Conversion of HTML to 'Pandoc' document. -} module Text.Pandoc.Readers.HTML ( readHtml , htmlTag , htmlInBalanced , isInlineTag , isBlockTag , isTextTag , isCommentTag ) where import Text.ParserCombinators.Parsec import Text.ParserCombinators.Parsec.Pos import Text.HTML.TagSoup import Text.HTML.TagSoup.Match import Text.Pandoc.Definition import Text.Pandoc.Builder (text, toList) import Text.Pandoc.Shared import Text.Pandoc.Parsing import Data.Maybe ( fromMaybe, isJust ) import Data.List ( intercalate ) import Data.Char ( isSpace, isDigit ) import Control.Monad ( liftM, guard ) -- | Convert HTML-formatted string to 'Pandoc' document. readHtml :: ParserState -- ^ Parser state -> String -- ^ String to parse (assumes @'\n'@ line endings) -> Pandoc readHtml st inp = Pandoc meta blocks where blocks = readWith parseBody st rest tags = canonicalizeTags $ parseTagsOptions parseOptions{ optTagPosition = True } inp hasHeader = any (~== TagOpen "head" []) tags (meta, rest) = if hasHeader then parseHeader tags else (Meta [] [] [], tags) type TagParser = GenParser (Tag String) ParserState -- TODO - fix this - not every header has a title tag parseHeader :: [Tag String] -> (Meta, [Tag String]) parseHeader tags = (Meta{docTitle = tit'', docAuthors = [], docDate = []}, rest) where (tit,_) = break (~== TagClose "title") $ drop 1 $ dropWhile (\t -> not $ t ~== TagOpen "title" []) tags tit' = concatMap fromTagText $ filter isTagText tit tit'' = normalizeSpaces $ toList $ text tit' rest = drop 1 $ dropWhile (\t -> not $ t ~== TagClose "head" || t ~== TagOpen "body" []) tags parseBody :: TagParser [Block] parseBody = liftM concat $ manyTill block eof block :: TagParser [Block] block = choice [ pPara , pHeader , pBlockQuote , pCodeBlock , pList , pHrule , pSimpleTable , pPlain , pRawHtmlBlock ] renderTags' :: [Tag String] -> String renderTags' = renderTagsOptions renderOptions{ optMinimize = (`elem` ["hr","br","img"]) } pList :: TagParser [Block] pList = pBulletList <|> pOrderedList <|> pDefinitionList pBulletList :: TagParser [Block] pBulletList = try $ do pSatisfy (~== TagOpen "ul" []) let nonItem = pSatisfy (\t -> not (tagOpen (`elem` ["li","ol","ul","dl"]) (const True) t) && not (t ~== TagClose "ul")) -- note: if they have an
    or
      not in scope of a
    • , -- treat it as a list item, though it's not valid xhtml... skipMany nonItem items <- manyTill (pInTags "li" block >>~ skipMany nonItem) (pCloses "ul") return [BulletList items] pOrderedList :: TagParser [Block] pOrderedList = try $ do TagOpen _ attribs <- pSatisfy (~== TagOpen "ol" []) st <- getState let (start, style) = if stateStrict st then (1, DefaultStyle) else (sta', sty') where sta = fromMaybe "1" $ lookup "start" attribs sta' = if all isDigit sta then read sta else 1 sty = fromMaybe (fromMaybe "" $ lookup "style" attribs) $ lookup "class" attribs sty' = case sty of "lower-roman" -> LowerRoman "upper-roman" -> UpperRoman "lower-alpha" -> LowerAlpha "upper-alpha" -> UpperAlpha "decimal" -> Decimal _ -> DefaultStyle let nonItem = pSatisfy (\t -> not (tagOpen (`elem` ["li","ol","ul","dl"]) (const True) t) && not (t ~== TagClose "ol")) -- note: if they have an
        or
          not in scope of a
        • , -- treat it as a list item, though it's not valid xhtml... skipMany nonItem items <- manyTill (pInTags "li" block >>~ skipMany nonItem) (pCloses "ol") return [OrderedList (start, style, DefaultDelim) items] pDefinitionList :: TagParser [Block] pDefinitionList = try $ do pSatisfy (~== TagOpen "dl" []) items <- manyTill pDefListItem (pCloses "dl") return [DefinitionList items] pDefListItem :: TagParser ([Inline],[[Block]]) pDefListItem = try $ do let nonItem = pSatisfy (\t -> not (t ~== TagOpen "dt" []) && not (t ~== TagOpen "dd" []) && not (t ~== TagClose "dl")) terms <- many1 (try $ skipMany nonItem >> pInTags "dt" inline) defs <- many1 (try $ skipMany nonItem >> pInTags "dd" block) skipMany nonItem let term = intercalate [LineBreak] terms return (term, defs) pRawTag :: TagParser String pRawTag = do tag <- pAnyTag let ignorable x = x `elem` ["html","head","body"] if tagOpen ignorable (const True) tag || tagClose ignorable tag then return [] else return $ renderTags' [tag] pRawHtmlBlock :: TagParser [Block] pRawHtmlBlock = do raw <- pHtmlBlock "script" <|> pHtmlBlock "style" <|> pRawTag state <- getState if stateParseRaw state && not (null raw) then return [RawHtml raw] else return [] pHtmlBlock :: String -> TagParser String pHtmlBlock t = try $ do open <- pSatisfy (~== TagOpen t []) contents <- manyTill pAnyTag (pSatisfy (~== TagClose t)) return $ renderTags' $ [open] ++ contents ++ [TagClose t] pHeader :: TagParser [Block] pHeader = try $ do TagOpen tagtype attr <- pSatisfy $ tagOpen (`elem` ["h1","h2","h3","h4","h5","h6"]) (const True) let bodyTitle = TagOpen tagtype attr ~== TagOpen "h1" [("class","title")] let level = read (drop 1 tagtype) contents <- liftM concat $ manyTill inline (pCloses tagtype <|> eof) return $ if bodyTitle then [] -- skip a representation of the title in the body else [Header level $ normalizeSpaces contents] pHrule :: TagParser [Block] pHrule = do pSelfClosing (=="hr") (const True) return [HorizontalRule] pSimpleTable :: TagParser [Block] pSimpleTable = try $ do TagOpen _ _ <- pSatisfy (~== TagOpen "table" []) skipMany pBlank head' <- option [] $ pInTags "th" pTd rows <- many1 $ try $ skipMany pBlank >> pInTags "tr" pTd skipMany pBlank TagClose _ <- pSatisfy (~== TagClose "table") let cols = maximum $ map length rows let aligns = replicate cols AlignLeft let widths = replicate cols 0 return [Table [] aligns widths head' rows] pTd :: TagParser [TableCell] pTd = try $ do skipMany pBlank res <- pInTags "td" pPlain skipMany pBlank return [res] pBlockQuote :: TagParser [Block] pBlockQuote = do contents <- pInTags "blockquote" block return [BlockQuote contents] pPlain :: TagParser [Block] pPlain = do contents <- liftM (normalizeSpaces . concat) $ many1 inline if null contents then return [] else return [Plain contents] pPara :: TagParser [Block] pPara = do contents <- pInTags "p" inline return [Para $ normalizeSpaces contents] pCodeBlock :: TagParser [Block] pCodeBlock = try $ do TagOpen _ attr <- pSatisfy (~== TagOpen "pre" []) contents <- manyTill pAnyTag (pCloses "pre" <|> eof) let rawText = concatMap fromTagText $ filter isTagText contents -- drop leading newline if any let result' = case rawText of '\n':xs -> xs _ -> rawText -- drop trailing newline if any let result = case reverse result' of '\n':_ -> init result' _ -> result' let attribsId = fromMaybe "" $ lookup "id" attr let attribsClasses = words $ fromMaybe "" $ lookup "class" attr let attribsKV = filter (\(k,_) -> k /= "class" && k /= "id") attr st <- getState let attribs = if stateStrict st then ("",[],[]) else (attribsId, attribsClasses, attribsKV) return [CodeBlock attribs result] inline :: TagParser [Inline] inline = choice [ pTagText , pEmph , pStrong , pSuperscript , pSubscript , pStrikeout , pLineBreak , pLink , pImage , pCode , pRawHtmlInline ] pLocation :: TagParser () pLocation = do (TagPosition r c) <- pSat isTagPosition setPosition $ newPos "input" r c pSat :: (Tag String -> Bool) -> TagParser (Tag String) pSat f = do pos <- getPosition token show (const pos) (\x -> if f x then Just x else Nothing) pSatisfy :: (Tag String -> Bool) -> TagParser (Tag String) pSatisfy f = try $ optional pLocation >> pSat f pAnyTag :: TagParser (Tag String) pAnyTag = pSatisfy (const True) pSelfClosing :: (String -> Bool) -> ([Attribute String] -> Bool) -> TagParser (Tag String) pSelfClosing f g = do open <- pSatisfy (tagOpen f g) optional $ pSatisfy (tagClose f) return open pEmph :: TagParser [Inline] pEmph = pInlinesInTags "em" Emph <|> pInlinesInTags "i" Emph pStrong :: TagParser [Inline] pStrong = pInlinesInTags "strong" Strong <|> pInlinesInTags "b" Strong pSuperscript :: TagParser [Inline] pSuperscript = failIfStrict >> pInlinesInTags "sup" Superscript pSubscript :: TagParser [Inline] pSubscript = failIfStrict >> pInlinesInTags "sub" Subscript pStrikeout :: TagParser [Inline] pStrikeout = do failIfStrict pInlinesInTags "s" Strikeout <|> pInlinesInTags "strike" Strikeout <|> pInlinesInTags "del" Strikeout <|> try (do pSatisfy (~== TagOpen "span" [("class","strikeout")]) contents <- liftM concat $ manyTill inline (pCloses "span") return [Strikeout contents]) pLineBreak :: TagParser [Inline] pLineBreak = do pSelfClosing (=="br") (const True) return [LineBreak] pLink :: TagParser [Inline] pLink = try $ do tag <- pSatisfy (tagOpenLit "a" (isJust . lookup "href")) let url = fromAttrib "href" tag let title = fromAttrib "title" tag lab <- liftM concat $ manyTill inline (pCloses "a") return [Link (normalizeSpaces lab) (escapeURI url, title)] pImage :: TagParser [Inline] pImage = do tag <- pSelfClosing (=="img") (isJust . lookup "src") let url = fromAttrib "src" tag let title = fromAttrib "title" tag let alt = fromAttrib "alt" tag return [Image (toList $ text alt) (escapeURI url, title)] pCode :: TagParser [Inline] pCode = try $ do (TagOpen open _) <- pSatisfy $ tagOpen (`elem` ["code","tt"]) (const True) result <- manyTill pAnyTag (pCloses open) return [Code $ intercalate " " $ lines $ innerText result] pRawHtmlInline :: TagParser [Inline] pRawHtmlInline = do result <- pSatisfy (tagComment (const True)) <|> pSatisfy isInlineTag state <- getState if stateParseRaw state then return [HtmlInline $ renderTags' [result]] else return [] pInlinesInTags :: String -> ([Inline] -> Inline) -> TagParser [Inline] pInlinesInTags tagtype f = do contents <- pInTags tagtype inline return [f contents] pInTags :: String -> TagParser [a] -> TagParser [a] pInTags tagtype parser = try $ do pSatisfy (~== TagOpen tagtype []) liftM concat $ manyTill parser (pCloses tagtype <|> eof) pCloses :: String -> TagParser () pCloses tagtype = try $ do t <- lookAhead $ pSatisfy $ \tag -> isTagClose tag || isTagOpen tag case t of (TagClose t') | t' == tagtype -> pAnyTag >> return () (TagOpen t' _) | t' `closes` tagtype -> return () (TagClose "ul") | tagtype == "li" -> return () (TagClose "ol") | tagtype == "li" -> return () (TagClose "dl") | tagtype == "li" -> return () _ -> pzero pTagText :: TagParser [Inline] pTagText = try $ do (TagText str) <- pSatisfy isTagText st <- getState case runParser (many pTagContents) st "text" str of Left _ -> fail $ "Could not parse `" ++ str ++ "'" Right result -> return result pBlank :: TagParser () pBlank = try $ do (TagText str) <- pSatisfy isTagText guard $ all isSpace str pTagContents :: GenParser Char ParserState Inline pTagContents = pStr <|> pSpace <|> smartPunctuation pTagContents <|> pSymbol pStr :: GenParser Char ParserState Inline pStr = liftM Str $ many1 $ satisfy $ \c -> not (isSpace c) && not (isSpecial c) isSpecial :: Char -> Bool isSpecial '"' = True isSpecial '\'' = True isSpecial '.' = True isSpecial '-' = True isSpecial '\8216' = True isSpecial '\8217' = True isSpecial '\8220' = True isSpecial '\8221' = True isSpecial _ = False pSymbol :: GenParser Char ParserState Inline pSymbol = satisfy isSpecial >>= return . Str . (:[]) pSpace :: GenParser Char ParserState Inline pSpace = many1 (satisfy isSpace) >> return Space -- -- Constants -- eitherBlockOrInline :: [String] eitherBlockOrInline = ["applet", "button", "del", "iframe", "ins", "map", "area", "object"] {- inlineHtmlTags :: [[Char]] inlineHtmlTags = ["a", "abbr", "acronym", "b", "basefont", "bdo", "big", "br", "cite", "code", "dfn", "em", "font", "i", "img", "input", "kbd", "label", "q", "s", "samp", "select", "small", "span", "strike", "strong", "sub", "sup", "textarea", "tt", "u", "var"] -} blockHtmlTags :: [String] blockHtmlTags = ["address", "blockquote", "body", "center", "dir", "div", "dl", "fieldset", "form", "h1", "h2", "h3", "h4", "h5", "h6", "head", "hr", "html", "isindex", "menu", "noframes", "noscript", "ol", "p", "pre", "table", "ul", "dd", "dt", "frameset", "li", "tbody", "td", "tfoot", "th", "thead", "tr", "script", "style"] isInlineTag :: Tag String -> Bool isInlineTag t = tagOpen (`notElem` blockHtmlTags) (const True) t || tagClose (`notElem` blockHtmlTags) t || tagComment (const True) t isBlockTag :: Tag String -> Bool isBlockTag t = tagOpen (`elem` blocktags) (const True) t || tagClose (`elem` blocktags) t || tagComment (const True) t where blocktags = blockHtmlTags ++ eitherBlockOrInline isTextTag :: Tag String -> Bool isTextTag = tagText (const True) isCommentTag :: Tag String -> Bool isCommentTag = tagComment (const True) -- taken from HXT and extended closes :: String -> String -> Bool _ `closes` "body" = False _ `closes` "html" = False "a" `closes` "a" = True "li" `closes` "li" = True "th" `closes` t | t `elem` ["th","td"] = True "tr" `closes` t | t `elem` ["th","td","tr"] = True "dt" `closes` t | t `elem` ["dt","dd"] = True "hr" `closes` "p" = True "p" `closes` "p" = True "meta" `closes` "meta" = True "colgroup" `closes` "colgroup" = True "form" `closes` "form" = True "label" `closes` "label" = True "map" `closes` "map" = True "object" `closes` "object" = True _ `closes` t | t `elem` ["option","style","script","textarea","title"] = True t `closes` "select" | t /= "option" = True "thead" `closes` t | t `elem` ["colgroup"] = True "tfoot" `closes` t | t `elem` ["thead","colgroup"] = True "tbody" `closes` t | t `elem` ["tbody","tfoot","thead","colgroup"] = True t `closes` t2 | t `elem` ["h1","h2","h3","h4","h5","h6","dl","ol","ul","table","div","p"] && t2 `elem` ["h1","h2","h3","h4","h5","h6","p" ] = True -- not "div" t1 `closes` t2 | t1 `elem` blockHtmlTags && t2 `notElem` (blockHtmlTags ++ eitherBlockOrInline) = True _ `closes` _ = False --- parsers for use in markdown, textile readers -- | Matches a stretch of HTML in balanced tags. htmlInBalanced :: (Tag String -> Bool) -> GenParser Char ParserState String htmlInBalanced f = try $ do (TagOpen t _, tag) <- htmlTag f guard $ '/' `notElem` tag -- not a self-closing tag let nonTagChunk = many1 $ satisfy (/= '<') let stopper = htmlTag (~== TagClose t) let anytag = liftM snd $ htmlTag (const True) contents <- many $ notFollowedBy' stopper >> (nonTagChunk <|> htmlInBalanced (const True) <|> anytag) endtag <- liftM snd stopper return $ tag ++ concat contents ++ endtag -- | Matches a tag meeting a certain condition. htmlTag :: (Tag String -> Bool) -> GenParser Char ParserState (Tag String, String) htmlTag f = try $ do lookAhead (char '<') (next : _) <- getInput >>= return . canonicalizeTags . parseTags guard $ f next -- advance the parser case next of TagComment s -> do count (length s + 4) anyChar skipMany (satisfy (/='>')) char '>' return (next, "") _ -> do rendered <- manyTill anyChar (char '>') return (next, rendered ++ ">")