diff options
author | Alexander Krotov <ilabdsf@gmail.com> | 2017-11-10 02:23:58 +0300 |
---|---|---|
committer | Alexander Krotov <ilabdsf@gmail.com> | 2017-11-10 02:23:58 +0300 |
commit | f72d7636557b01812dcd13826b7697c3ceb20a76 (patch) | |
tree | b8cade06c632004ada9d309e63a17149046cda5c /src | |
parent | 725500bc582c4f0cd4e7ad3aa921e0bf2b705b02 (diff) | |
download | pandoc-f72d7636557b01812dcd13826b7697c3ceb20a76.tar.gz |
HTML reader: hlint
Diffstat (limited to 'src')
-rw-r--r-- | src/Text/Pandoc/Readers/HTML.hs | 61 |
1 files changed, 30 insertions, 31 deletions
diff --git a/src/Text/Pandoc/Readers/HTML.hs b/src/Text/Pandoc/Readers/HTML.hs index 2b667c63c..7d514e042 100644 --- a/src/Text/Pandoc/Readers/HTML.hs +++ b/src/Text/Pandoc/Readers/HTML.hs @@ -54,8 +54,7 @@ import Data.List (intercalate, isPrefixOf) import Data.List.Split (wordsBy) import qualified Data.Map as M import Data.Maybe (fromMaybe, isJust, isNothing) -import Data.Monoid (First (..)) -import Data.Monoid ((<>)) +import Data.Monoid (First (..), (<>)) import qualified Data.Set as Set import Data.Text (Text) import qualified Data.Text as T @@ -89,7 +88,7 @@ readHtml opts inp = do parseTagsOptions parseOptions{ optTagPosition = True } (crFilter inp) parseDoc = do - blocks <- (fixPlains False) . mconcat <$> manyTill block eof + blocks <- fixPlains False . mconcat <$> manyTill block eof meta <- stateMeta . parserState <$> getState bs' <- replaceNotes (B.toList blocks) reportLogMessages @@ -223,10 +222,10 @@ eSwitch constructor parser = try $ do eCase :: PandocMonad m => TagParser m (Maybe Inlines) eCase = do skipMany pBlank - TagOpen _ attr' <- lookAhead $ pSatisfy $ (matchTagOpen "case" []) + TagOpen _ attr' <- lookAhead $ pSatisfy (matchTagOpen "case" []) let attr = toStringAttr attr' - case (flip lookup namespaces) =<< lookup "required-namespace" attr of - Just p -> Just <$> (pInTags "case" (skipMany pBlank *> p <* skipMany pBlank)) + case flip lookup namespaces =<< lookup "required-namespace" attr of + Just p -> Just <$> pInTags "case" (skipMany pBlank *> p <* skipMany pBlank) Nothing -> Nothing <$ manyTill pAnyTag (pSatisfy (matchTagClose "case")) eFootnote :: PandocMonad m => TagParser m () @@ -235,20 +234,20 @@ eFootnote = try $ do guardEnabled Ext_epub_html_exts (TagOpen tag attr') <- lookAhead pAnyTag let attr = toStringAttr attr' - guard (maybe False (flip elem notes) (lookup "type" attr)) + guard $ maybe False (`elem` notes) (lookup "type" attr) let ident = fromMaybe "" (lookup "id" attr) content <- pInTags tag block addNote ident content addNote :: PandocMonad m => String -> Blocks -> TagParser m () -addNote uid cont = updateState (\s -> s {noteTable = (uid, cont) : (noteTable s)}) +addNote uid cont = updateState (\s -> s {noteTable = (uid, cont) : noteTable s}) eNoteref :: PandocMonad m => TagParser m Inlines eNoteref = try $ do guardEnabled Ext_epub_html_exts - TagOpen tag attr' <- lookAhead $ pAnyTag + TagOpen tag attr' <- lookAhead pAnyTag let attr = toStringAttr attr' - guard (maybe False (== "noteref") (lookup "type" attr)) + guard $ lookup "type" attr == Just "noteref" let ident = maybe "" (dropWhile (== '#')) (lookup "href" attr) guard (not (null ident)) pInTags tag block @@ -258,8 +257,8 @@ eNoteref = try $ do eTOC :: PandocMonad m => TagParser m () eTOC = try $ do guardEnabled Ext_epub_html_exts - (TagOpen tag attr) <- lookAhead $ pAnyTag - guard (maybe False (== "toc") (lookup "type" attr)) + (TagOpen tag attr) <- lookAhead pAnyTag + guard $ lookup "type" attr == Just "toc" void (pInTags tag block) pList :: PandocMonad m => TagParser m Blocks @@ -285,7 +284,7 @@ pListItem nonItem = do (Plain ils:xs) -> B.fromList (Plain [Span (ident, [], []) ils] : xs) _ -> B.divWith (ident, [], []) bs - (maybe id addId (lookup "id" attr)) <$> + maybe id addId (lookup "id" attr) <$> pInTags "li" block <* skipMany nonItem parseListStyleType :: String -> ListNumberStyle @@ -356,14 +355,14 @@ fixPlains :: Bool -> Blocks -> Blocks fixPlains inList bs = if any isParaish bs' then B.fromList $ map plainToPara bs' else bs - where isParaish (Para _) = True - isParaish (CodeBlock _ _) = True - isParaish (Header _ _ _) = True - isParaish (BlockQuote _) = True - isParaish (BulletList _) = not inList - isParaish (OrderedList _ _) = not inList - isParaish (DefinitionList _) = not inList - isParaish _ = False + where isParaish Para{} = True + isParaish CodeBlock{} = True + isParaish Header{} = True + isParaish BlockQuote{} = True + isParaish BulletList{} = not inList + isParaish OrderedList{} = not inList + isParaish DefinitionList{} = not inList + isParaish _ = False plainToPara (Plain xs) = Para xs plainToPara x = x bs' = B.toList bs @@ -427,10 +426,10 @@ eSection = try $ do setInChapter (pInTags tag block) headerLevel :: PandocMonad m => Text -> TagParser m Int -headerLevel tagtype = do +headerLevel tagtype = case safeRead (T.unpack (T.drop 1 tagtype)) of Just level -> - (try $ do + try (do guardEnabled Ext_epub_html_exts asks inChapter >>= guard return (level - 1)) @@ -481,12 +480,12 @@ pTable = try $ do pTBody = pOptInTag "tbody" $ many1 pTr head'' <- pOptInTag "thead" pTh head' <- map snd <$> - (pOptInTag "tbody" $ - if null head'' then pTh else return head'') + pOptInTag "tbody" + (if null head'' then pTh else return head'') rowsLs <- many pTBody rows' <- pOptInTag "tfoot" $ many pTr TagClose _ <- pSatisfy (matchTagClose "table") - let rows'' = (concat rowsLs) <> rows' + let rows'' = concat rowsLs <> rows' let rows''' = map (map snd) rows'' -- let rows''' = map (map snd) rows'' -- fail on empty table @@ -691,7 +690,7 @@ pSubscript :: PandocMonad m => TagParser m Inlines pSubscript = pInlinesInTags "sub" B.subscript pStrikeout :: PandocMonad m => TagParser m Inlines -pStrikeout = do +pStrikeout = pInlinesInTags "s" B.strikeout <|> pInlinesInTags "strike" B.strikeout <|> pInlinesInTags "del" B.strikeout <|> @@ -719,7 +718,7 @@ pLink = try $ do tag <- pSatisfy $ tagOpenLit "a" (const True) let title = T.unpack $ fromAttrib "title" tag -- take id from id attribute if present, otherwise name - let uid = maybe (T.unpack $ fromAttrib "name" tag) id $ + let uid = fromMaybe (T.unpack $ fromAttrib "name" tag) $ maybeFromAttrib "id" tag let cls = words $ T.unpack $ fromAttrib "class" tag lab <- trimInlines . mconcat <$> manyTill inline (pCloses "a") @@ -750,7 +749,7 @@ pImage = do let getAtt k = case fromAttrib k tag of "" -> [] v -> [(T.unpack k, T.unpack v)] - let kvs = concat $ map getAtt ["width", "height", "sizes", "srcset"] + let kvs = concatMap getAtt ["width", "height", "sizes", "srcset"] return $ B.imageWith (uid, cls, kvs) (escapeURI url) title (B.text alt) pCode :: PandocMonad m => TagParser m Inlines @@ -846,7 +845,7 @@ pCloses :: PandocMonad m => Text -> TagParser m () pCloses tagtype = try $ do t <- lookAhead $ pSatisfy $ \tag -> isTagClose tag || isTagOpen tag case t of - (TagClose t') | t' == tagtype -> pAnyTag >> return () + (TagClose t') | t' == tagtype -> void pAnyTag (TagOpen t' _) | t' `closes` tagtype -> return () (TagClose "ul") | tagtype == "li" -> return () (TagClose "ol") | tagtype == "li" -> return () @@ -1197,7 +1196,7 @@ htmlTag f = try $ do mkAttr :: [(String, String)] -> Attr mkAttr attr = (attribsId, attribsClasses, attribsKV) where attribsId = fromMaybe "" $ lookup "id" attr - attribsClasses = (words $ fromMaybe "" $ lookup "class" attr) <> epubTypes + attribsClasses = words (fromMaybe "" $ lookup "class" attr) <> epubTypes attribsKV = filter (\(k,_) -> k /= "class" && k /= "id") attr epubTypes = words $ fromMaybe "" $ lookup "epub:type" attr |