summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorJohn MacFarlane <jgm@berkeley.edu>2018-01-19 21:25:24 -0800
committerJohn MacFarlane <jgm@berkeley.edu>2018-01-19 21:25:24 -0800
commitb8ffd834cff717fe424f22e506351f2ecec4655a (patch)
tree70359c33066bebf2ec4c54c1c2d78f38b49c0fb8
parent8b3707de0402165b5691f626370203fa8982a5dc (diff)
hlint code improvements.
-rw-r--r--data/templates/default.ms2
-rw-r--r--src/Text/Pandoc/Class.hs6
-rw-r--r--src/Text/Pandoc/Filter.hs2
-rw-r--r--src/Text/Pandoc/Lua.hs2
-rw-r--r--src/Text/Pandoc/Lua/Packages.hs1
-rw-r--r--src/Text/Pandoc/Lua/StackInstances.hs1
-rw-r--r--src/Text/Pandoc/Parsing.hs24
-rw-r--r--src/Text/Pandoc/Readers/Docx.hs16
-rw-r--r--src/Text/Pandoc/Readers/Docx/Fields.hs8
-rw-r--r--src/Text/Pandoc/Readers/Docx/Lists.hs8
-rw-r--r--src/Text/Pandoc/Readers/Docx/Parse.hs11
-rw-r--r--src/Text/Pandoc/Readers/HTML.hs8
-rw-r--r--src/Text/Pandoc/Readers/JATS.hs1
-rw-r--r--src/Text/Pandoc/Readers/LaTeX.hs20
-rw-r--r--src/Text/Pandoc/Readers/Markdown.hs58
-rw-r--r--src/Text/Pandoc/Readers/Muse.hs2
-rw-r--r--src/Text/Pandoc/Readers/Odt/Arrows/Utils.hs4
-rw-r--r--src/Text/Pandoc/Readers/Odt/ContentReader.hs19
-rw-r--r--src/Text/Pandoc/Readers/Odt/Namespaces.hs2
-rw-r--r--src/Text/Pandoc/Readers/Odt/StyleReader.hs43
-rw-r--r--src/Text/Pandoc/Readers/Org/Blocks.hs2
-rw-r--r--src/Text/Pandoc/Readers/RST.hs8
-rw-r--r--src/Text/Pandoc/Readers/Textile.hs13
-rw-r--r--src/Text/Pandoc/Readers/TikiWiki.hs14
-rw-r--r--src/Text/Pandoc/Readers/Txt2Tags.hs4
-rw-r--r--src/Text/Pandoc/Readers/Vimwiki.hs4
-rw-r--r--src/Text/Pandoc/Shared.hs2
-rw-r--r--src/Text/Pandoc/Writers/AsciiDoc.hs5
-rw-r--r--src/Text/Pandoc/Writers/Docx.hs13
-rw-r--r--src/Text/Pandoc/Writers/FB2.hs3
-rw-r--r--src/Text/Pandoc/Writers/HTML.hs13
-rw-r--r--src/Text/Pandoc/Writers/Haddock.hs3
-rw-r--r--src/Text/Pandoc/Writers/ICML.hs4
-rw-r--r--src/Text/Pandoc/Writers/LaTeX.hs8
-rw-r--r--src/Text/Pandoc/Writers/Man.hs2
-rw-r--r--src/Text/Pandoc/Writers/Muse.hs2
-rw-r--r--src/Text/Pandoc/Writers/OOXML.hs2
-rw-r--r--src/Text/Pandoc/Writers/OpenDocument.hs2
-rw-r--r--src/Text/Pandoc/Writers/Powerpoint.hs2
-rw-r--r--src/Text/Pandoc/Writers/Powerpoint/Presentation.hs74
-rw-r--r--src/Text/Pandoc/Writers/RST.hs15
-rw-r--r--src/Text/Pandoc/Writers/RTF.hs8
-rw-r--r--src/Text/Pandoc/Writers/Texinfo.hs2
-rw-r--r--test/Tests/Command.hs6
-rw-r--r--test/Tests/Readers/Creole.hs2
-rw-r--r--test/Tests/Readers/Docx.hs11
-rw-r--r--test/Tests/Readers/EPUB.hs2
-rw-r--r--test/Tests/Readers/Muse.hs6
-rw-r--r--test/Tests/Readers/Org/Block/Header.hs2
-rw-r--r--test/Tests/Readers/Org/Block/List.hs18
-rw-r--r--test/Tests/Readers/Org/Directive.hs4
-rw-r--r--test/Tests/Readers/Org/Inline.hs59
-rw-r--r--test/Tests/Readers/Org/Inline/Note.hs1
-rw-r--r--test/Tests/Readers/Org/Inline/Smart.hs4
-rw-r--r--test/Tests/Readers/Org/Meta.hs28
-rw-r--r--test/Tests/Readers/RST.hs29
-rw-r--r--test/Tests/Readers/Txt2Tags.hs8
-rw-r--r--test/Tests/Shared.hs36
-rw-r--r--test/Tests/Writers/ConTeXt.hs7
-rw-r--r--test/Tests/Writers/Docbook.hs2
-rw-r--r--test/Tests/Writers/FB2.hs2
-rw-r--r--test/Tests/Writers/JATS.hs2
-rw-r--r--test/Tests/Writers/Markdown.hs42
-rw-r--r--test/Tests/Writers/Native.hs2
-rw-r--r--test/Tests/Writers/Powerpoint.hs12
-rw-r--r--test/Tests/Writers/TEI.hs2
66 files changed, 349 insertions, 381 deletions
diff --git a/data/templates/default.ms b/data/templates/default.ms
index 6468bf568..f4204338a 100644
--- a/data/templates/default.ms
+++ b/data/templates/default.ms
@@ -40,8 +40,8 @@ $endif$
.nr FL \n[LL]
.\" footnote point size
.nr FPS (\n[PS] - 2000)
-.\" paper size
$if(papersize)$
+.\" paper size
.ds paper $papersize$
$endif$
.\" color used for strikeout
diff --git a/src/Text/Pandoc/Class.hs b/src/Text/Pandoc/Class.hs
index f8d6b6737..ae538046a 100644
--- a/src/Text/Pandoc/Class.hs
+++ b/src/Text/Pandoc/Class.hs
@@ -142,11 +142,11 @@ import qualified System.Environment as IO (lookupEnv)
import System.FilePath.Glob (match, compile)
import System.Directory (createDirectoryIfMissing, getDirectoryContents,
doesDirectoryExist)
-import System.FilePath ((</>), (<.>), takeDirectory,
- takeExtension, dropExtension, isRelative, normalise)
+import System.FilePath
+ ((</>), (<.>), takeDirectory, takeExtension, dropExtension,
+ isRelative, normalise, splitDirectories)
import qualified System.FilePath.Glob as IO (glob)
import qualified System.FilePath.Posix as Posix
-import System.FilePath (splitDirectories)
import qualified System.Directory as IO (getModificationTime)
import Control.Monad as M (fail)
import Control.Monad.State.Strict
diff --git a/src/Text/Pandoc/Filter.hs b/src/Text/Pandoc/Filter.hs
index 67b3a5f2c..e2a3c3e16 100644
--- a/src/Text/Pandoc/Filter.hs
+++ b/src/Text/Pandoc/Filter.hs
@@ -51,7 +51,7 @@ applyFilters :: ReaderOptions
-> [String]
-> Pandoc
-> PandocIO Pandoc
-applyFilters ropts filters args d = do
+applyFilters ropts filters args d =
foldrM ($) d $ map applyFilter filters
where
applyFilter (JSONFilter f) = JSONFilter.apply ropts args f
diff --git a/src/Text/Pandoc/Lua.hs b/src/Text/Pandoc/Lua.hs
index edf803b45..790be47d5 100644
--- a/src/Text/Pandoc/Lua.hs
+++ b/src/Text/Pandoc/Lua.hs
@@ -65,7 +65,7 @@ runLuaFilter' ropts filterPath format pd = do
newtop <- Lua.gettop
-- Use the returned filters, or the implicitly defined global filter if
-- nothing was returned.
- luaFilters <- if (newtop - top >= 1)
+ luaFilters <- if newtop - top >= 1
then peek (-1)
else Lua.getglobal "_G" *> fmap (:[]) popValue
runAll luaFilters pd
diff --git a/src/Text/Pandoc/Lua/Packages.hs b/src/Text/Pandoc/Lua/Packages.hs
index 0169d0045..1e6ff22fe 100644
--- a/src/Text/Pandoc/Lua/Packages.hs
+++ b/src/Text/Pandoc/Lua/Packages.hs
@@ -113,4 +113,3 @@ dataDirScript datadir moduleFile = do
return $ case res of
Left _ -> Nothing
Right s -> Just (unpack s)
-
diff --git a/src/Text/Pandoc/Lua/StackInstances.hs b/src/Text/Pandoc/Lua/StackInstances.hs
index 38404157c..a504e5626 100644
--- a/src/Text/Pandoc/Lua/StackInstances.hs
+++ b/src/Text/Pandoc/Lua/StackInstances.hs
@@ -383,4 +383,3 @@ instance ToLuaStack ReaderOptions where
LuaUtil.addValue "defaultImageExtension" defaultImageExtension
LuaUtil.addValue "trackChanges" trackChanges
LuaUtil.addValue "stripComments" stripComments
-
diff --git a/src/Text/Pandoc/Parsing.hs b/src/Text/Pandoc/Parsing.hs
index f1b823965..e87ea71da 100644
--- a/src/Text/Pandoc/Parsing.hs
+++ b/src/Text/Pandoc/Parsing.hs
@@ -189,12 +189,12 @@ where
import Control.Monad.Identity
import Control.Monad.Reader
-import Data.Char (chr, isAlphaNum, isAscii, isAsciiUpper, isHexDigit, isPunctuation, isSpace,
- ord, toLower, toUpper)
+import Data.Char (chr, isAlphaNum, isAscii, isAsciiUpper, isHexDigit,
+ isPunctuation, isSpace, ord, toLower, toUpper)
import Data.Default
import Data.List (intercalate, isSuffixOf, transpose)
import qualified Data.Map as M
-import Data.Maybe (mapMaybe)
+import Data.Maybe (mapMaybe, fromMaybe)
import Data.Monoid ((<>))
import qualified Data.Set as Set
import Data.Text (Text)
@@ -304,7 +304,7 @@ indentWith :: Stream s m Char
=> Int -> ParserT s st m [Char]
indentWith num = do
tabStop <- getOption readerTabStop
- if (num < tabStop)
+ if num < tabStop
then count num (char ' ')
else choice [ try (count num (char ' '))
, try (char '\t' >> indentWith (num - tabStop)) ]
@@ -573,7 +573,7 @@ uri = try $ do
let uriChunk = skipMany1 wordChar
<|> percentEscaped
<|> entity
- <|> (try $ punct >>
+ <|> try (punct >>
lookAhead (void (satisfy isWordChar) <|> percentEscaped))
str <- snd <$> withRaw (skipMany1 ( () <$
(enclosed (char '(') (char ')') uriChunk
@@ -755,7 +755,7 @@ romanOne = (char 'i' >> return (LowerRoman, 1)) <|>
-- | Parses an ordered list marker and returns list attributes.
anyOrderedListMarker :: Stream s m Char => ParserT s ParserState m ListAttributes
-anyOrderedListMarker = choice $
+anyOrderedListMarker = choice
[delimParser numParser | delimParser <- [inPeriod, inOneParen, inTwoParens],
numParser <- [decimal, exampleNum, defaultNum, romanOne,
lowerAlpha, lowerRoman, upperAlpha, upperRoman]]
@@ -896,7 +896,7 @@ widthsFromIndices numColumns' indices =
quotient = if totLength > numColumns
then fromIntegral totLength
else fromIntegral numColumns
- fracs = map (\l -> (fromIntegral l) / quotient) lengths in
+ fracs = map (\l -> fromIntegral l / quotient) lengths in
tail fracs
---
@@ -977,7 +977,7 @@ gridTableHeader headless blocks = try $ do
then replicate (length underDashes) ""
else map (unlines . map trim) $ transpose
$ map (gridTableSplitLine indices) rawContent
- heads <- fmap sequence $ mapM (parseFromString blocks . trim) rawHeads
+ heads <- sequence <$> mapM (parseFromString blocks . trim) rawHeads
return (heads, aligns, indices)
gridTableRawLine :: Stream s m Char => [Int] -> ParserT s st m [String]
@@ -1323,9 +1323,7 @@ failIfInQuoteContext :: (HasQuoteContext st m, Stream s m t)
-> ParserT s st m ()
failIfInQuoteContext context = do
context' <- getQuoteContext
- if context' == context
- then fail "already inside quotes"
- else return ()
+ when (context' == context) $ fail "already inside quotes"
charOrRef :: Stream s m Char => String -> ParserT s st m Char
charOrRef cs =
@@ -1418,9 +1416,7 @@ a <+?> b = a >>= flip fmap (try b <|> return mempty) . (<>)
extractIdClass :: Attr -> Attr
extractIdClass (ident, cls, kvs) = (ident', cls', kvs')
where
- ident' = case lookup "id" kvs of
- Just v -> v
- Nothing -> ident
+ ident' = fromMaybe ident (lookup "id" kvs)
cls' = case lookup "class" kvs of
Just cl -> words cl
Nothing -> cls
diff --git a/src/Text/Pandoc/Readers/Docx.hs b/src/Text/Pandoc/Readers/Docx.hs
index 21120824f..c24c43901 100644
--- a/src/Text/Pandoc/Readers/Docx.hs
+++ b/src/Text/Pandoc/Readers/Docx.hs
@@ -81,7 +81,7 @@ import qualified Data.ByteString.Lazy as B
import Data.Default (Default)
import Data.List (delete, intersect)
import qualified Data.Map as M
-import Data.Maybe (isJust)
+import Data.Maybe (isJust, fromMaybe)
import Data.Sequence (ViewL (..), viewl)
import qualified Data.Sequence as Seq
import qualified Data.Set as Set
@@ -187,7 +187,7 @@ bodyPartsToMeta' :: PandocMonad m => [BodyPart] -> DocxContext m (M.Map String M
bodyPartsToMeta' [] = return M.empty
bodyPartsToMeta' (bp : bps)
| (Paragraph pPr parParts) <- bp
- , (c : _)<- intersect (pStyle pPr) (M.keys metaStyles)
+ , (c : _)<- (pStyle pPr) `intersect` (M.keys metaStyles)
, (Just metaField) <- M.lookup c metaStyles = do
inlines <- smushInlines <$> mapM parPartToInlines parParts
remaining <- bodyPartsToMeta' bps
@@ -340,7 +340,7 @@ blocksToInlinesWarn cmtId blks = do
notParaOrPlain (Para _) = False
notParaOrPlain (Plain _) = False
notParaOrPlain _ = True
- unless (null $ filter notParaOrPlain blkList) $
+ unless ( not (any notParaOrPlain blkList)) $
lift $ P.report $ DocxParserWarning $
"Docx comment " ++ cmtId ++ " will not retain formatting"
return $ blocksToInlines' blkList
@@ -351,7 +351,7 @@ blocksToInlinesWarn cmtId blks = do
parPartToInlines :: PandocMonad m => ParPart -> DocxContext m Inlines
parPartToInlines parPart =
case parPart of
- (BookMark _ anchor) | not $ anchor `elem` dummyAnchors -> do
+ (BookMark _ anchor) | notElem anchor dummyAnchors -> do
inHdrBool <- asks docxInHeaderBlock
ils <- parPartToInlines' parPart
immedPrevAnchor <- gets docxImmedPrevAnchor
@@ -444,9 +444,9 @@ parPartToInlines' (ExternalHyperLink target runs) = do
return $ link target "" ils
parPartToInlines' (PlainOMath exps) =
return $ math $ writeTeX exps
-parPartToInlines' (SmartTag runs) = do
+parPartToInlines' (SmartTag runs) =
smushInlines <$> mapM runToInlines runs
-parPartToInlines' (Field info runs) = do
+parPartToInlines' (Field info runs) =
case info of
HyperlinkField url -> parPartToInlines' $ ExternalHyperLink url runs
UnknownField -> smushInlines <$> mapM runToInlines runs
@@ -626,9 +626,7 @@ bodyPartToBlocks (ListItem pPr numId lvl (Just levelInfo) parparts) = do
(_, fmt,txt, startFromLevelInfo) = levelInfo
start = case startFromState of
Just n -> n + 1
- Nothing -> case startFromLevelInfo of
- Just n' -> n'
- Nothing -> 1
+ Nothing -> fromMaybe 1 startFromLevelInfo
kvs = [ ("level", lvl)
, ("num-id", numId)
, ("format", fmt)
diff --git a/src/Text/Pandoc/Readers/Docx/Fields.hs b/src/Text/Pandoc/Readers/Docx/Fields.hs
index 69758b431..f0821a751 100644
--- a/src/Text/Pandoc/Readers/Docx/Fields.hs
+++ b/src/Text/Pandoc/Readers/Docx/Fields.hs
@@ -46,7 +46,7 @@ parseFieldInfo = parse fieldInfo ""
fieldInfo :: Parser FieldInfo
fieldInfo =
- (try $ HyperlinkField <$> hyperlink)
+ try (HyperlinkField <$> hyperlink)
<|>
return UnknownField
@@ -54,7 +54,7 @@ escapedQuote :: Parser String
escapedQuote = string "\\\""
inQuotes :: Parser String
-inQuotes = do
+inQuotes =
(try escapedQuote) <|> (anyChar >>= (\c -> return [c]))
quotedString :: Parser String
@@ -63,7 +63,7 @@ quotedString = do
concat <$> manyTill inQuotes (try (char '"'))
unquotedString :: Parser String
-unquotedString = manyTill anyChar (try (space))
+unquotedString = manyTill anyChar (try space)
fieldArgument :: Parser String
fieldArgument = quotedString <|> unquotedString
@@ -82,7 +82,7 @@ hyperlink = do
string "HYPERLINK"
spaces
farg <- fieldArgument
- switches <- (spaces *> many hyperlinkSwitch)
+ switches <- spaces *> many hyperlinkSwitch
let url = case switches of
("\\l", s) : _ -> farg ++ ('#': s)
_ -> farg
diff --git a/src/Text/Pandoc/Readers/Docx/Lists.hs b/src/Text/Pandoc/Readers/Docx/Lists.hs
index fa4870fff..c0f05094a 100644
--- a/src/Text/Pandoc/Readers/Docx/Lists.hs
+++ b/src/Text/Pandoc/Readers/Docx/Lists.hs
@@ -44,14 +44,14 @@ isListItem (Div (_, classes, _) _) | "list-item" `elem` classes = True
isListItem _ = False
getLevel :: Block -> Maybe Integer
-getLevel (Div (_, _, kvs) _) = fmap read $ lookup "level" kvs
+getLevel (Div (_, _, kvs) _) = read <$> lookup "level" kvs
getLevel _ = Nothing
getLevelN :: Block -> Integer
getLevelN b = fromMaybe (-1) (getLevel b)
getNumId :: Block -> Maybe Integer
-getNumId (Div (_, _, kvs) _) = fmap read $ lookup "num-id" kvs
+getNumId (Div (_, _, kvs) _) = read <$> lookup "num-id" kvs
getNumId _ = Nothing
getNumIdN :: Block -> Integer
@@ -140,8 +140,8 @@ flatToBullets' num xs@(b : elems)
(children, remaining) =
span
(\b' ->
- (getLevelN b') > bLevel ||
- ((getLevelN b') == bLevel && (getNumIdN b') == bNumId))
+ getLevelN b' > bLevel ||
+ (getLevelN b' == bLevel && getNumIdN b' == bNumId))
xs
in
case getListType b of
diff --git a/src/Text/Pandoc/Readers/Docx/Parse.hs b/src/Text/Pandoc/Readers/Docx/Parse.hs
index 5f648666f..c123a0018 100644
--- a/src/Text/Pandoc/Readers/Docx/Parse.hs
+++ b/src/Text/Pandoc/Readers/Docx/Parse.hs
@@ -358,9 +358,7 @@ archiveToDocument zf = do
docElem <- maybeToD $ (parseXMLDoc . UTF8.toStringLazy . fromEntry) entry
let namespaces = elemToNameSpaces docElem
bodyElem <- maybeToD $ findChildByName namespaces "w" "body" docElem
- let bodyElem' = case walkDocument namespaces bodyElem of
- Just e -> e
- Nothing -> bodyElem
+ let bodyElem' = fromMaybe bodyElem (walkDocument namespaces bodyElem)
body <- elemToBody namespaces bodyElem'
return $ Document namespaces body
@@ -603,7 +601,7 @@ elemToTblLook ns element | isElem ns "w" "tblLook" element =
Just bitMask -> testBitMask bitMask 0x020
Nothing -> False
in
- return $ TblLook{firstRowFormatting = firstRowFmt}
+ return TblLook{firstRowFormatting = firstRowFmt}
elemToTblLook _ _ = throwError WrongElem
elemToRow :: NameSpaces -> Element -> D Row
@@ -623,7 +621,7 @@ elemToCell _ _ = throwError WrongElem
elemToParIndentation :: NameSpaces -> Element -> Maybe ParIndentation
elemToParIndentation ns element | isElem ns "w" "ind" element =
- Just $ ParIndentation {
+ Just ParIndentation {
leftParIndent =
findAttrByName ns "w" "left" element >>=
stringToInteger
@@ -1173,8 +1171,7 @@ elemToRunElems ns element
let font = do
fontElem <- findElement (qualName "rFonts") element
stringToFont =<<
- foldr (<|>) Nothing (
- map (flip findAttr fontElem . qualName) ["ascii", "hAnsi"])
+ foldr ((<|>) . (flip findAttr fontElem . qualName)) Nothing ["ascii", "hAnsi"]
local (setFont font) (mapD (elemToRunElem ns) (elChildren element))
elemToRunElems _ _ = throwError WrongElem
diff --git a/src/Text/Pandoc/Readers/HTML.hs b/src/Text/Pandoc/Readers/HTML.hs
index f15bf1c96..0e79f9ec3 100644
--- a/src/Text/Pandoc/Readers/HTML.hs
+++ b/src/Text/Pandoc/Readers/HTML.hs
@@ -43,14 +43,14 @@ module Text.Pandoc.Readers.HTML ( readHtml
) where
import Control.Applicative ((<|>))
-import Control.Arrow ((***))
+import Control.Arrow (first)
import Control.Monad (guard, mplus, msum, mzero, unless, void)
import Control.Monad.Except (throwError)
import Control.Monad.Reader (ReaderT, ask, asks, lift, local, runReaderT)
import Data.Char (isAlphaNum, isDigit, isLetter)
import Data.Default (Default (..), def)
import Data.Foldable (for_)
-import Data.List (intercalate, isPrefixOf)
+import Data.List (isPrefixOf)
import Data.List.Split (wordsBy, splitWhen)
import qualified Data.Map as M
import Data.Maybe (fromMaybe, isJust, isNothing)
@@ -777,7 +777,7 @@ pCode = try $ do
(TagOpen open attr') <- pSatisfy $ tagOpen (`elem` ["code","tt"]) (const True)
let attr = toStringAttr attr'
result <- manyTill pAnyTag (pCloses open)
- return $ B.codeWith (mkAttr attr) $ intercalate " " $ lines $ T.unpack $
+ return $ B.codeWith (mkAttr attr) $ unwords $ lines $ T.unpack $
innerText result
pSpan :: PandocMonad m => TagParser m Inlines
@@ -1227,7 +1227,7 @@ stripPrefixes = map stripPrefix
stripPrefix :: Tag Text -> Tag Text
stripPrefix (TagOpen s as) =
- TagOpen (stripPrefix' s) (map (stripPrefix' *** id) as)
+ TagOpen (stripPrefix' s) (map (first stripPrefix') as)
stripPrefix (TagClose s) = TagClose (stripPrefix' s)
stripPrefix x = x
diff --git a/src/Text/Pandoc/Readers/JATS.hs b/src/Text/Pandoc/Readers/JATS.hs
index 9223db68c..8158a4511 100644
--- a/src/Text/Pandoc/Readers/JATS.hs
+++ b/src/Text/Pandoc/Readers/JATS.hs
@@ -494,4 +494,3 @@ parseInline (Elem e) =
"" -> []
l -> [l]
return $ codeWith (attrValue "id" e,classes',[]) $ strContentRecursive e
-
diff --git a/src/Text/Pandoc/Readers/LaTeX.hs b/src/Text/Pandoc/Readers/LaTeX.hs
index 3408201eb..1ce3d18e5 100644
--- a/src/Text/Pandoc/Readers/LaTeX.hs
+++ b/src/Text/Pandoc/Readers/LaTeX.hs
@@ -272,7 +272,7 @@ rawLaTeXBlock = do
lookAhead (try (char '\\' >> letter))
-- we don't want to apply newly defined latex macros to their own
-- definitions:
- (snd <$> rawLaTeXParser macroDef)
+ snd <$> rawLaTeXParser macroDef
<|> ((snd <$> rawLaTeXParser (environment <|> blockCommand)) >>= applyMacros)
rawLaTeXInline :: (PandocMonad m, HasMacros s, HasReaderOptions s)
@@ -351,7 +351,7 @@ totoks pos t =
Tok pos (Arg i) ("#" <> t1)
: totoks (incSourceColumn pos (1 + T.length t1)) t2
Nothing ->
- Tok pos Symbol ("#")
+ Tok pos Symbol "#"
: totoks (incSourceColumn pos 1) t2
| c == '^' ->
case T.uncons rest of
@@ -369,10 +369,10 @@ totoks pos t =
| d < '\128' ->
Tok pos Esc1 (T.pack ['^','^',d])
: totoks (incSourceColumn pos 3) rest''
- _ -> Tok pos Symbol ("^") :
- Tok (incSourceColumn pos 1) Symbol ("^") :
+ _ -> Tok pos Symbol "^" :
+ Tok (incSourceColumn pos 1) Symbol "^" :
totoks (incSourceColumn pos 2) rest'
- _ -> Tok pos Symbol ("^")
+ _ -> Tok pos Symbol "^"
: totoks (incSourceColumn pos 1) rest
| otherwise ->
Tok pos Symbol (T.singleton c) : totoks (incSourceColumn pos 1) rest
@@ -454,7 +454,7 @@ doMacros n = do
addTok _ (Tok _ (CtrlSeq x) txt)
acc@(Tok _ Word _ : _)
| not (T.null txt) &&
- (isLetter (T.last txt)) =
+ isLetter (T.last txt) =
Tok spos (CtrlSeq x) (txt <> " ") : acc
addTok _ t acc = setpos spos t : acc
ts' <- getInput
@@ -1244,7 +1244,7 @@ inlineEnvironments = M.fromList [
]
inlineCommands :: PandocMonad m => M.Map Text (LP m Inlines)
-inlineCommands = M.union inlineLanguageCommands $ M.fromList $
+inlineCommands = M.union inlineLanguageCommands $ M.fromList
[ ("emph", extractSpaces emph <$> tok)
, ("textit", extractSpaces emph <$> tok)
, ("textsl", extractSpaces emph <$> tok)
@@ -1501,7 +1501,7 @@ foreignlanguage :: PandocMonad m => LP m Inlines
foreignlanguage = do
babelLang <- T.unpack . untokenize <$> braced
case babelLangToBCP47 babelLang of
- Just lang -> spanWith ("", [], [("lang", renderLang $ lang)]) <$> tok
+ Just lang -> spanWith ("", [], [("lang", renderLang lang)]) <$> tok
_ -> tok
inlineLanguageCommands :: PandocMonad m => M.Map Text (LP m Inlines)
@@ -2021,7 +2021,7 @@ closing = do
return $ para (trimInlines contents) <> sigs
blockCommands :: PandocMonad m => M.Map Text (LP m Blocks)
-blockCommands = M.fromList $
+blockCommands = M.fromList
[ ("par", mempty <$ skipopts)
, ("parbox", skipopts >> braced >> grouped blocks)
, ("title", mempty <$ (skipopts *>
@@ -2444,7 +2444,7 @@ parseAligns = try $ do
spaces
spec <- braced
case safeRead ds of
- Just n -> do
+ Just n ->
getInput >>= setInput . (mconcat (replicate n spec) ++)
Nothing -> fail $ "Could not parse " ++ ds ++ " as number"
bgroup
diff --git a/src/Text/Pandoc/Readers/Markdown.hs b/src/Text/Pandoc/Readers/Markdown.hs
index 94f04eee7..92e9098bd 100644
--- a/src/Text/Pandoc/Readers/Markdown.hs
+++ b/src/Text/Pandoc/Readers/Markdown.hs
@@ -36,7 +36,7 @@ import Control.Monad
import Control.Monad.Except (throwError)
import Data.Char (isAlphaNum, isPunctuation, isSpace, toLower)
import qualified Data.HashMap.Strict as H
-import Data.List (findIndex, intercalate, sortBy, transpose)
+import Data.List (intercalate, sortBy, transpose, elemIndex)
import qualified Data.Map as M
import Data.Maybe
import Data.Monoid ((<>))
@@ -162,16 +162,14 @@ inlinesInBalancedBrackets =
stripBracket xs = if last xs == ']' then init xs else xs
go :: PandocMonad m => Int -> MarkdownParser m ()
go 0 = return ()
- go openBrackets = do
+ go openBrackets =
(() <$ (escapedChar <|>
- code <|>
- rawHtmlInline <|>
- rawLaTeXInline') >> go openBrackets)
+ code <|>
+ rawHtmlInline <|>
+ rawLaTeXInline') >> go openBrackets)
<|>
(do char ']'
- if openBrackets > 1
- then go (openBrackets - 1)
- else return ())
+ Control.Monad.when (openBrackets > 1) $ go (openBrackets - 1))
<|>
(char '[' >> go (openBrackets + 1))
<|>
@@ -257,13 +255,13 @@ yamlMetaBlock = try $ do
v' <- yamlToMeta v
let k' = T.unpack k
updateState $ \st -> st{ stateMeta' =
- (do m <- stateMeta' st
- -- if there's already a value, leave it unchanged
- case lookupMeta k' m of
- Just _ -> return m
- Nothing -> do
- v'' <- v'
- return $ B.setMeta (T.unpack k) v'' m)}
+ do m <- stateMeta' st
+ -- if there's already a value, leave it unchanged
+ case lookupMeta k' m of
+ Just _ -> return m
+ Nothing -> do
+ v'' <- v'
+ return $ B.setMeta (T.unpack k) v'' m}
) alist
Right Yaml.Null -> return ()
Right _ -> do
@@ -596,7 +594,7 @@ setextHeader = try $ do
underlineChar <- oneOf setextHChars
many (char underlineChar)
blanklines
- let level = fromMaybe 0 (findIndex (== underlineChar) setextHChars) + 1
+ let level = fromMaybe 0 (elemIndex underlineChar setextHChars) + 1
attr' <- registerHeader attr (runF text defaultParserState)
guardDisabled Ext_implicit_header_references
<|> registerImplicitHeader raw attr'
@@ -851,7 +849,7 @@ orderedListStart mbstydelim = try $ do
return (num, style, delim))
listStart :: PandocMonad m => MarkdownParser m ()
-listStart = bulletListStart <|> (Control.Monad.void (orderedListStart Nothing))
+listStart = bulletListStart <|> Control.Monad.void (orderedListStart Nothing)
listLine :: PandocMonad m => Int -> MarkdownParser m String
listLine continuationIndent = try $ do
@@ -881,7 +879,7 @@ rawListItem fourSpaceRule start = try $ do
pos2 <- getPosition
let continuationIndent = if fourSpaceRule
then 4
- else (sourceColumn pos2 - sourceColumn pos1)
+ else sourceColumn pos2 - sourceColumn pos1
first <- listLineCommon
rest <- many (do notFollowedBy listStart
notFollowedBy (() <$ codeBlockFenced)
@@ -912,10 +910,10 @@ listContinuation continuationIndent = try $ do
return $ concat (x:xs) ++ blanks
notFollowedByDivCloser :: PandocMonad m => MarkdownParser m ()
-notFollowedByDivCloser = do
+notFollowedByDivCloser =
guardDisabled Ext_fenced_divs <|>
- do divLevel <- stateFencedDivLevel <$> getState
- guard (divLevel < 1) <|> notFollowedBy divFenceEnd
+ do divLevel <- stateFencedDivLevel <$> getState
+ guard (divLevel < 1) <|> notFollowedBy divFenceEnd
notFollowedByHtmlCloser :: PandocMonad m => MarkdownParser m ()
notFollowedByHtmlCloser = do
@@ -1222,7 +1220,7 @@ simpleTableHeader headless = try $ do
if headless
then lookAhead anyLine
else return rawContent
- let aligns = zipWith alignType (map ((: [])) rawHeads) lengths
+ let aligns = zipWith alignType (map (: []) rawHeads) lengths
let rawHeads' = if headless
then replicate (length dashes) ""
else rawHeads
@@ -1418,11 +1416,11 @@ pipeTableHeaderPart = try $ do
skipMany spaceChar
let len = length pipe + maybe 0 (const 1) left + maybe 0 (const 1) right
return
- ((case (left,right) of
- (Nothing,Nothing) -> AlignDefault
- (Just _,Nothing) -> AlignLeft
- (Nothing,Just _) -> AlignRight
- (Just _,Just _) -> AlignCenter), len)
+ (case (left,right) of
+ (Nothing,Nothing) -> AlignDefault
+ (Just _,Nothing) -> AlignLeft
+ (Nothing,Just _) -> AlignRight
+ (Just _,Just _) -> AlignCenter, len)
-- Succeed only if current line contains a pipe.
scanForPipe :: PandocMonad m => ParserT [Char] st m ()
@@ -1929,7 +1927,7 @@ rawConTeXtEnvironment = try $ do
(try $ string "\\stop" >> string completion)
return $ "\\start" ++ completion ++ concat contents ++ "\\stop" ++ completion
-inBrackets :: PandocMonad m => (ParserT [Char] st m Char) -> ParserT [Char] st m String
+inBrackets :: PandocMonad m => ParserT [Char] st m Char -> ParserT [Char] st m String
inBrackets parser = do
char '['
contents <- many parser
@@ -2150,6 +2148,6 @@ doubleQuoted :: PandocMonad m => MarkdownParser m (F Inlines)
doubleQuoted = try $ do
doubleQuoteStart
contents <- mconcat <$> many (try $ notFollowedBy doubleQuoteEnd >> inline)
- (withQuoteContext InDoubleQuote $ doubleQuoteEnd >> return
+ withQuoteContext InDoubleQuote (doubleQuoteEnd >> return
(fmap B.doubleQuoted . trimInlinesF $ contents))
- <|> (return $ return (B.str "\8220") <> contents)
+ <|> return (return (B.str "\8220") <> contents)
diff --git a/src/Text/Pandoc/Readers/Muse.hs b/src/Text/Pandoc/Readers/Muse.hs
index 4c6d1278e..973dfa15c 100644
--- a/src/Text/Pandoc/Readers/Muse.hs
+++ b/src/Text/Pandoc/Readers/Muse.hs
@@ -482,7 +482,7 @@ definitionList :: PandocMonad m => MuseParser m (F Blocks)
definitionList = try $ do
many spaceChar
pos <- getPosition
- (guardDisabled Ext_amuse) <|> (guard (sourceColumn pos /= 1)) -- Initial space is required by Amusewiki, but not Emacs Muse
+ guardDisabled Ext_amuse <|> guard (sourceColumn pos /= 1) -- Initial space is required by Amusewiki, but not Emacs Muse
first <- definitionListItem 0
rest <- many $ try (optionMaybe blankline >> definitionListItem (sourceColumn pos - 1))
return $ B.definitionList <$> sequence (first : rest)
diff --git a/src/Text/Pandoc/Readers/Odt/Arrows/Utils.hs b/src/Text/Pandoc/Readers/Odt/Arrows/Utils.hs
index cdfa8f8df..ef8b2d18a 100644
--- a/src/Text/Pandoc/Readers/Odt/Arrows/Utils.hs
+++ b/src/Text/Pandoc/Readers/Odt/Arrows/Utils.hs
@@ -211,9 +211,9 @@ a ^>>?% f = arr a >>?^ (uncurry f)
---
(>>?%?) :: (ArrowChoice a)
=> FallibleArrow a x f (b,b')
- -> (b -> b' -> (Either f c))
+ -> (b -> b' -> Either f c)
-> FallibleArrow a x f c
-a >>?%? f = a >>?^? (uncurry f)
+a >>?%? f = a >>?^? uncurry f
infixr 1 >>?, >>?^, >>?^?
infixr 1 ^>>?, >>?!
diff --git a/src/Text/Pandoc/Readers/Odt/ContentReader.hs b/src/Text/Pandoc/Readers/Odt/ContentReader.hs
index cc9b798b3..380f16c66 100644
--- a/src/Text/Pandoc/Readers/Odt/ContentReader.hs
+++ b/src/Text/Pandoc/Readers/Odt/ContentReader.hs
@@ -322,7 +322,7 @@ type InlineModifier = Inlines -> Inlines
modifierFromStyleDiff :: PropertyTriple -> InlineModifier
modifierFromStyleDiff propertyTriple =
composition $
- (getVPosModifier propertyTriple)
+ getVPosModifier propertyTriple
: map (first ($ propertyTriple) >>> ifThen_else ignore)
[ (hasEmphChanged , emph )
, (hasChanged isStrong , strong )
@@ -352,7 +352,7 @@ modifierFromStyleDiff propertyTriple =
]
hasChanged property triple@(_, property -> newProperty, _) =
- maybe True (/=newProperty) (lookupPreviousValue property triple)
+ (/= Just newProperty) (lookupPreviousValue property triple)
hasChangedM property triple@(_, textProps,_) =
fromMaybe False $ (/=) <$> property textProps <*> lookupPreviousValueM property triple
@@ -362,7 +362,7 @@ modifierFromStyleDiff propertyTriple =
lookupPreviousValueM f = lookupPreviousStyleValue ((f =<<).textProperties)
lookupPreviousStyleValue f (ReaderState{..},_,mFamily)
- = ( findBy f $ extendedStylePropertyChain styleTrace styleSet )
+ = findBy f (extendedStylePropertyChain styleTrace styleSet)
<|> ( f =<< fmap (lookupDefaultStyle' styleSet) mFamily )
@@ -793,8 +793,7 @@ read_image_src = matchingElement NsDraw "image"
Left _ -> returnV "" -< ()
read_frame_title :: InlineMatcher
-read_frame_title = matchingElement NsSVG "title"
- $ (matchChildContent [] read_plain_text)
+read_frame_title = matchingElement NsSVG "title" (matchChildContent [] read_plain_text)
read_frame_text_box :: InlineMatcher
read_frame_text_box = matchingElement NsDraw "text-box"
@@ -803,12 +802,12 @@ read_frame_text_box = matchingElement NsDraw "text-box"
arr read_img_with_caption -< toList paragraphs
read_img_with_caption :: [Block] -> Inlines
-read_img_with_caption ((Para [Image attr alt (src,title)]) : _) =
+read_img_with_caption (Para [Image attr alt (src,title)] : _) =
singleton (Image attr alt (src, 'f':'i':'g':':':title)) -- no text, default caption
read_img_with_caption (Para (Image attr _ (src,title) : txt) : _) =
singleton (Image attr txt (src, 'f':'i':'g':':':title) ) -- override caption with the text that follows
-read_img_with_caption ( (Para (_ : xs)) : ys) =
- read_img_with_caption ((Para xs) : ys)
+read_img_with_caption ( Para (_ : xs) : ys) =
+ read_img_with_caption (Para xs : ys)
read_img_with_caption _ =
mempty
@@ -909,8 +908,8 @@ post_process (Pandoc m blocks) =
Pandoc m (post_process' blocks)
post_process' :: [Block] -> [Block]
-post_process' ((Table _ a w h r) : (Div ("", ["caption"], _) [Para inlines] ) : xs) =
- (Table inlines a w h r) : ( post_process' xs )
+post_process' (Table _ a w h r : Div ("", ["caption"], _) [Para inlines] : xs) =
+ Table inlines a w h r : post_process' xs
post_process' bs = bs
read_body :: OdtReader _x (Pandoc, MediaBag)
diff --git a/src/Text/Pandoc/Readers/Odt/Namespaces.hs b/src/Text/Pandoc/Readers/Odt/Namespaces.hs
index 3c11aeb8e..92e12931d 100644
--- a/src/Text/Pandoc/Readers/Odt/Namespaces.hs
+++ b/src/Text/Pandoc/Readers/Odt/Namespaces.hs
@@ -48,7 +48,7 @@ instance NameSpaceID Namespace where
findID :: NameSpaceIRI -> Maybe Namespace
-findID iri = listToMaybe [nsID | (iri',~nsID) <- nsIDs, iri' `isPrefixOf` iri]
+findID iri = listToMaybe [nsID | (iri',nsID) <- nsIDs, iri' `isPrefixOf` iri]
nsIDmap :: NameSpaceIRIs Namespace
nsIDmap = foldr (uncurry $ flip M.insert) M.empty nsIDs
diff --git a/src/Text/Pandoc/Readers/Odt/StyleReader.hs b/src/Text/Pandoc/Readers/Odt/StyleReader.hs
index 6129c1664..58be8e4a3 100644
--- a/src/Text/Pandoc/Readers/Odt/StyleReader.hs
+++ b/src/Text/Pandoc/Readers/Odt/StyleReader.hs
@@ -131,13 +131,12 @@ type StyleReaderSafe a b = XMLReaderSafe FontPitches a b
-- | A reader for font pitches
fontPitchReader :: XMLReader _s _x FontPitches
fontPitchReader = executeIn NsOffice "font-face-decls" (
- ( withEveryL NsStyle "font-face" $ liftAsSuccess (
+ withEveryL NsStyle "font-face" (liftAsSuccess (
findAttr' NsStyle "name"
&&&
lookupDefaultingAttr NsStyle "font-pitch"
- )
- )
- >>?^ ( M.fromList . (foldl accumLegalPitches []) )
+ ))
+ >>?^ ( M.fromList . foldl accumLegalPitches [] )
)
where accumLegalPitches ls (Nothing,_) = ls
accumLegalPitches ls (Just n,p) = (n,p):ls
@@ -383,11 +382,11 @@ data ListLevelStyle = ListLevelStyle { listLevelType :: ListLevelType
instance Show ListLevelStyle where
show ListLevelStyle{..} = "<LLS|"
- ++ (show listLevelType)
+ ++ show listLevelType
++ "|"
- ++ (maybeToString listItemPrefix)
- ++ (show listItemFormat)
- ++ (maybeToString listItemSuffix)
+ ++ maybeToString listItemPrefix
+ ++ show listItemFormat
+ ++ maybeToString listItemSuffix
++ ">"
where maybeToString = fromMaybe ""
@@ -483,14 +482,14 @@ readTextProperties =
( liftA6 PropT
( searchAttr NsXSL_FO "font-style" False isFontEmphasised )
( searchAttr NsXSL_FO "font-weight" False isFontBold )
- ( findPitch )
+ findPitch
( getAttr NsStyle "text-position" )
- ( readUnderlineMode )
- ( readStrikeThroughMode )
+ readUnderlineMode
+ readStrikeThroughMode
)
where isFontEmphasised = [("normal",False),("italic",True),("oblique",True)]
isFontBold = ("normal",False):("bold",True)
- :(map ((,True).show) ([100,200..900]::[Int]))
+ :map ((,True).show) ([100,200..900]::[Int])
readUnderlineMode :: StyleReaderSafe _x (Maybe UnderlineMode)
readUnderlineMode = readLineMode "text-underline-mode"
@@ -510,7 +509,7 @@ readLineMode modeAttr styleAttr = proc x -> do
Nothing -> returnA -< Just UnderlineModeNormal
else returnA -< Nothing
where
- isLinePresent = [("none",False)] ++ map (,True)
+ isLinePresent = ("none",False) : map (,True)
[ "dash" , "dot-dash" , "dot-dot-dash" , "dotted"
, "long-dash" , "solid" , "wave"
]
@@ -547,20 +546,18 @@ readListStyle =
findAttr NsStyle "name"
>>?! keepingTheValue
( liftA ListStyle
- $ ( liftA3 SM.union3
+ $ liftA3 SM.union3
( readListLevelStyles NsText "list-level-style-number" LltNumbered )
( readListLevelStyles NsText "list-level-style-bullet" LltBullet )
- ( readListLevelStyles NsText "list-level-style-image" LltImage )
- ) >>^ M.mapMaybe chooseMostSpecificListLevelStyle
+ ( readListLevelStyles NsText "list-level-style-image" LltImage ) >>^ M.mapMaybe chooseMostSpecificListLevelStyle
)
--
readListLevelStyles :: Namespace -> ElementName
-> ListLevelType
-> StyleReaderSafe _x (SM.SetMap Int ListLevelStyle)
readListLevelStyles namespace elementName levelType =
- ( tryAll namespace elementName (readListLevelStyle levelType)
+ tryAll namespace elementName (readListLevelStyle levelType)
>>^ SM.fromList
- )
--
readListLevelStyle :: ListLevelType -> StyleReader _x (Int, ListLevelStyle)
@@ -632,7 +629,7 @@ parents style styles = unfoldr findNextParent style -- Ha!
getStyleFamily :: Style -> Styles -> Maybe StyleFamily
getStyleFamily style@Style{..} styles
= styleFamily
- <|> (F.asum $ map (`getStyleFamily` styles) $ parents style styles)
+ <|> F.asum (map (`getStyleFamily` styles) $ parents style styles)
-- | Each 'Style' has certain 'StyleProperties'. But sometimes not all property
-- values are specified. Instead, a value might be inherited from a
@@ -654,7 +651,7 @@ stylePropertyChain style styles
--
extendedStylePropertyChain :: [Style] -> Styles -> [StyleProperties]
extendedStylePropertyChain [] _ = []
-extendedStylePropertyChain [style] styles = (stylePropertyChain style styles)
- ++ (maybeToList (fmap (lookupDefaultStyle' styles) (getStyleFamily style styles)))
-extendedStylePropertyChain (style:trace) styles = (stylePropertyChain style styles)
- ++ (extendedStylePropertyChain trace styles)
+extendedStylePropertyChain [style] styles = stylePropertyChain style styles
+ ++ maybeToList (fmap (lookupDefaultStyle' styles) (getStyleFamily style styles))
+extendedStylePropertyChain (style:trace) styles = stylePropertyChain style styles
+ ++ extendedStylePropertyChain trace styles
diff --git a/src/Text/Pandoc/Readers/Org/Blocks.hs b/src/Text/Pandoc/Readers/Org/Blocks.hs
index c5a7d8e10..fa016283c 100644
--- a/src/Text/Pandoc/Readers/Org/Blocks.hs
+++ b/src/Text/Pandoc/Readers/Org/Blocks.hs
@@ -516,7 +516,7 @@ include = try $ do
blocksParser <- case includeArgs of
("example" : _) -> return $ pure . B.codeBlock <$> parseRaw
["export"] -> return . returnF $ B.fromList []
- ("export" : format : []) -> return $ pure . B.rawBlock format <$> parseRaw
+ ["export", format] -> return $ pure . B.rawBlock format <$> parseRaw
("src" : rest) -> do
let attr = case rest of
[lang] -> (mempty, [lang], mempty)
diff --git a/src/Text/Pandoc/Readers/RST.hs b/src/Text/Pandoc/Readers/RST.hs
index 49cc3018c..0e90fe945 100644
--- a/src/Text/Pandoc/Readers/RST.hs
+++ b/src/Text/Pandoc/Readers/RST.hs
@@ -1263,7 +1263,7 @@ simpleTableHeader headless = try $ do
let rawHeads = if headless
then replicate (length dashes) ""
else simpleTableSplitLine indices rawContent
- heads <- mapM ( (parseFromString' (mconcat <$> many plain)) . trim) rawHeads
+ heads <- mapM ( parseFromString' (mconcat <$> many plain) . trim) rawHeads
return (heads, aligns, indices)
-- Parse a simple table.
@@ -1414,7 +1414,7 @@ renderRole contents fmt role attr = case role of
pepUrl = "http://www.python.org/dev/peps/pep-" ++ padNo ++ "/"
addClass :: String -> Attr -> Attr
-addClass c (ident, classes, keyValues) = (ident, union classes [c], keyValues)
+addClass c (ident, classes, keyValues) = (ident, classes `union` [c], keyValues)
roleName :: PandocMonad m => RSTParser m String
roleName = many1 (letter <|> char '-')
@@ -1454,7 +1454,7 @@ endline = try $ do
notFollowedBy blankline
-- parse potential list-starts at beginning of line differently in a list:
st <- getState
- when ((stateParserContext st) == ListItemState) $ notFollowedBy (anyOrderedListMarker >> spaceChar) >>
+ when (stateParserContext st == ListItemState) $ notFollowedBy (anyOrderedListMarker >> spaceChar) >>
notFollowedBy' bulletListStart
return B.softbreak
@@ -1577,7 +1577,7 @@ note = try $ do
-- not yet in this implementation.
updateState $ \st -> st{ stateNotes = [] }
contents <- parseFromString' parseBlocks raw
- let newnotes = if (ref == "*" || ref == "#") -- auto-numbered
+ let newnotes = if ref == "*" || ref == "#" -- auto-numbered
-- delete the note so the next auto-numbered note
-- doesn't get the same contents:
then deleteFirstsBy (==) notes [(ref,raw)]
diff --git a/src/Text/Pandoc/Readers/Textile.hs b/src/Text/Pandoc/Readers/Textile.hs
index 46d6301e4..30bb6a715 100644
--- a/src/Text/Pandoc/Readers/Textile.hs
+++ b/src/Text/Pandoc/Readers/Textile.hs
@@ -110,7 +110,7 @@ noteBlock = try $ do
startPos <- getPosition
ref <- noteMarker
optional blankline
- contents <- fmap unlines $ many1Till anyLine (blanklines <|> noteBlock)
+ contents <- unlines <$> many1Till anyLine (blanklines <|> noteBlock)
endPos <- getPosition
let newnote = (ref, contents ++ "\n")
st <- getState
@@ -360,7 +360,7 @@ cellAttributes = try $ do
tableCell :: PandocMonad m => ParserT [Char] ParserState m ((Bool, Alignment), Blocks)
tableCell = try $ do
char '|'
- (isHeader, alignment) <- option (False, AlignDefault) $ cellAttributes
+ (isHeader, alignment) <- option (False, AlignDefault) cellAttributes
notFollowedBy blankline
raw <- trim <$>
many (noneOf "|\n" <|> try (char '\n' <* notFollowedBy blankline))
@@ -499,7 +499,7 @@ copy = do
note :: PandocMonad m => ParserT [Char] ParserState m Inlines
note = try $ do
- ref <- (char '[' *> many1 digit <* char ']')
+ ref <- char '[' *> many1 digit <* char ']'
notes <- stateNotes <$> getState
case lookup ref notes of
Nothing -> fail "note not found"
@@ -530,7 +530,7 @@ hyphenedWords = do
wordChunk :: PandocMonad m => ParserT [Char] ParserState m String
wordChunk = try $ do
hd <- noneOf wordBoundaries
- tl <- many ( (noneOf wordBoundaries) <|>
+ tl <- many ( noneOf wordBoundaries <|>
try (notFollowedBy' note *> oneOf markupChars
<* lookAhead (noneOf wordBoundaries) ) )
return $ hd:tl
@@ -614,7 +614,7 @@ escapedEqs = B.str <$>
-- | literal text escaped btw <notextile> tags
escapedTag :: PandocMonad m => ParserT [Char] ParserState m Inlines
escapedTag = B.str <$>
- (try $ string "<notextile>" *>
+ try (string "<notextile>" *>
manyTill anyChar' (try $ string "</notextile>"))
-- | Any special symbol defined in wordBoundaries
@@ -630,7 +630,8 @@ code = code1 <|> code2
-- any character except a newline before a blank line
anyChar' :: PandocMonad m => ParserT [Char] ParserState m Char
anyChar' =
- satisfy (/='\n') <|> (try $ char '\n' <* notFollowedBy blankline)
+ satisfy (/='\n') <|>
+ try (char '\n' <* notFollowedBy blankline)
code1 :: PandocMonad m => ParserT [Char] ParserState m Inlines
code1 = B.code <$> surrounded (char '@') anyChar'
diff --git a/src/Text/Pandoc/Readers/TikiWiki.hs b/src/Text/Pandoc/Readers/TikiWiki.hs
index 4a66cc13d..a92f7bed2 100644
--- a/src/Text/Pandoc/Readers/TikiWiki.hs
+++ b/src/Text/Pandoc/Readers/TikiWiki.hs
@@ -168,7 +168,7 @@ table = try $ do
where
-- The headers are as many empty srings as the number of columns
-- in the first row
- headers rows = map (B.plain . B.str) $replicate (length $ rows !! 0) ""
+ headers rows = map (B.plain . B.str) $replicate (length $ head rows) ""
para :: PandocMonad m => TikiWikiParser m B.Blocks
para = fmap (result . mconcat) ( many1Till inline endOfParaElement)
@@ -238,8 +238,8 @@ fixListNesting [first] = [recurseOnList first]
fixListNesting (first:second:rest) =
let secondBlock = head $ B.toList second in
case secondBlock of
- BulletList _ -> fixListNesting $ (mappend (recurseOnList first) (recurseOnList second)) : rest
- OrderedList _ _ -> fixListNesting $ (mappend (recurseOnList first) (recurseOnList second)) : rest
+ BulletList _ -> fixListNesting $ mappend (recurseOnList first) (recurseOnList second) : rest
+ OrderedList _ _ -> fixListNesting $ mappend (recurseOnList first) (recurseOnList second) : rest
_ -> recurseOnList first : fixListNesting (second:rest)
-- This function walks the Block structure for fixListNesting,
@@ -285,7 +285,7 @@ spanFoldUpList ln (first:rest) =
-- level and of the same type.
splitListNesting :: ListNesting -> (ListNesting, B.Blocks) -> Bool
splitListNesting ln1 (ln2, _)
- | (lnnest ln1) < (lnnest ln2) =
+ | lnnest ln1 < lnnest ln2 =
True
| ln1 == ln2 =
True
@@ -341,7 +341,7 @@ listItemLine nest = lineContent >>= parseContent
lineContent = do
content <- anyLine
continuation <- optionMaybe listContinuation
- return $ filterSpaces content ++ "\n" ++ maybe "" id continuation
+ return $ filterSpaces content ++ "\n" ++ Data.Maybe.fromMaybe "" continuation
filterSpaces = reverse . dropWhile (== ' ') . reverse
listContinuation = string (replicate nest '+') >> lineContent
parseContent x = do
@@ -410,7 +410,7 @@ inline = choice [ whitespace
] <?> "inline"
whitespace :: PandocMonad m => TikiWikiParser m B.Inlines
-whitespace = (lb <|> regsp)
+whitespace = lb <|> regsp
where lb = try $ skipMany spaceChar >> linebreak >> return B.space
regsp = try $ skipMany1 spaceChar >> return B.space
@@ -501,7 +501,7 @@ escapedChar = try $ do
string "~"
inner <- many1 $ oneOf "0123456789"
string "~"
- return $B.str [(toEnum (read inner :: Int)) :: Char]
+ return $B.str [toEnum (read inner :: Int) :: Char]
-- UNSUPPORTED, as there doesn't seem to be any facility in calibre
-- for this
diff --git a/src/Text/Pandoc/Readers/Txt2Tags.hs b/src/Text/Pandoc/Readers/Txt2Tags.hs
index 68399afc9..b4f4bc564 100644
--- a/src/Text/Pandoc/Readers/Txt2Tags.hs
+++ b/src/Text/Pandoc/Readers/Txt2Tags.hs
@@ -36,7 +36,7 @@ import Control.Monad.Except (catchError, throwError)
import Control.Monad.Reader (Reader, asks, runReader)
import Data.Char (toLower)
import Data.Default
-import Data.List (intercalate, intersperse, transpose)
+import Data.List (intercalate, transpose)
import Data.Maybe (fromMaybe)
import Data.Monoid ((<>))
import Data.Text (Text)
@@ -463,7 +463,7 @@ titleLink = try $ do
char ']'
let link' = last tokens
guard $ not $ null link'
- let tit = concat (intersperse " " (init tokens))
+ let tit = unwords (init tokens)
return $ B.link link' "" (B.text tit)
-- Link with image
diff --git a/src/Text/Pandoc/Readers/Vimwiki.hs b/src/Text/Pandoc/Readers/Vimwiki.hs
index 162fb371e..d717a1ba8 100644
--- a/src/Text/Pandoc/Readers/Vimwiki.hs
+++ b/src/Text/Pandoc/Readers/Vimwiki.hs
@@ -388,9 +388,7 @@ bulletListMarkers = "ul" <$ (char '*' <|> char '-')
orderedListMarkers :: PandocMonad m => VwParser m String
orderedListMarkers =
- ("ol" <$choice (orderedListMarker Decimal Period:(($OneParen)
- <$> orderedListMarker
- <$> [Decimal, LowerRoman, UpperRoman, LowerAlpha, UpperAlpha])))
+ ("ol" <$choice (orderedListMarker Decimal Period:(($OneParen) . orderedListMarker <$> [Decimal, LowerRoman, UpperRoman, LowerAlpha, UpperAlpha])))
<|> ("ol" <$ char '#')
--many need trimInlines
diff --git a/src/Text/Pandoc/Shared.hs b/src/Text/Pandoc/Shared.hs
index 583c7a63f..52e1447db 100644
--- a/src/Text/Pandoc/Shared.hs
+++ b/src/Text/Pandoc/Shared.hs
@@ -494,7 +494,7 @@ hierarchicalizeWithIds (Header level attr@(_,classes,_) title':xs) = do
return $ Sec level newnum attr title' sectionContents' : rest'
hierarchicalizeWithIds (Div ("",["references"],[])
(Header level (ident,classes,kvs) title' : xs):ys) =
- hierarchicalizeWithIds (Header level (ident,("references":classes),kvs)
+ hierarchicalizeWithIds (Header level (ident,"references":classes,kvs)
title' : (xs ++ ys))
hierarchicalizeWithIds (x:rest) = do
rest' <- hierarchicalizeWithIds rest
diff --git a/src/Text/Pandoc/Writers/AsciiDoc.hs b/src/Text/Pandoc/Writers/AsciiDoc.hs
index a6906eb68..b8f647b66 100644
--- a/src/Text/Pandoc/Writers/AsciiDoc.hs
+++ b/src/Text/Pandoc/Writers/AsciiDoc.hs
@@ -265,8 +265,7 @@ blockToAsciiDoc opts (OrderedList (_start, sty, _delim) items) = do
let markers' = map (\m -> if length m < 3
then m ++ replicate (3 - length m) ' '
else m) markers
- contents <- mapM (uncurry (orderedListItemToAsciiDoc opts)) $
- zip markers' items
+ contents <- zipWithM (orderedListItemToAsciiDoc opts) markers' items
return $ cat contents <> blankline
blockToAsciiDoc opts (DefinitionList items) = do
contents <- mapM (definitionListItemToAsciiDoc opts) items
@@ -452,7 +451,7 @@ inlineToAsciiDoc opts (Link _ txt (src, _tit)) = do
else prefix <> text src <> "[" <> linktext <> "]"
inlineToAsciiDoc opts (Image attr alternate (src, tit)) = do
-- image:images/logo.png[Company logo, title="blah"]
- let txt = if (null alternate) || (alternate == [Str ""])
+ let txt = if null alternate || (alternate == [Str ""])
then [Str "image"]
else alternate
linktext <- inlineListToAsciiDoc opts txt
diff --git a/src/Text/Pandoc/Writers/Docx.hs b/src/Text/Pandoc/Writers/Docx.hs
index adf5f232a..928eaa712 100644
--- a/src/Text/Pandoc/Writers/Docx.hs
+++ b/src/Text/Pandoc/Writers/Docx.hs
@@ -1057,12 +1057,9 @@ getParaProps displayMathPara = do
props <- asks envParaProperties
listLevel <- asks envListLevel
numid <- asks envListNumId
- let listPr = if listLevel >= 0 && not displayMathPara
- then [ mknode "w:numPr" []
- [ mknode "w:numId" [("w:val",show numid)] ()
- , mknode "w:ilvl" [("w:val",show listLevel)] () ]
- ]
- else []
+ let listPr = [mknode "w:numPr" []
+ [ mknode "w:numId" [("w:val",show numid)] ()
+ , mknode "w:ilvl" [("w:val",show listLevel)] () ] | listLevel >= 0 && not displayMathPara]
return $ case props ++ listPr of
[] -> []
ps -> [mknode "w:pPr" [] ps]
@@ -1145,7 +1142,7 @@ inlineToOpenXML' opts (Span (ident,classes,kvs) ils) = do
return $ \f -> do
x <- f
return [ mknode "w:ins"
- [("w:id", (show insId)),
+ [("w:id", show insId),
("w:author", author),
("w:date", date)] x ]
else return id
@@ -1272,7 +1269,7 @@ inlineToOpenXML' opts (Image attr alt (src, title)) = do
Nothing ->
catchError
(do (img, mt) <- P.fetchItem src
- ident <- ("rId"++) `fmap` ((lift . lift) getUniqueId)
+ ident <- ("rId"++) `fmap` (lift . lift) getUniqueId
let (xpt,ypt) = desiredSizeInPoints opts attr
(either (const def) id (imageSize opts img))
-- 12700 emu = 1 pt
diff --git a/src/Text/Pandoc/Writers/FB2.hs b/src/Text/Pandoc/Writers/FB2.hs
index b1e8c8575..e322c7d98 100644
--- a/src/Text/Pandoc/Writers/FB2.hs
+++ b/src/Text/Pandoc/Writers/FB2.hs
@@ -131,8 +131,7 @@ description meta' = do
_ -> return []
return $ el "description"
[ el "title-info" (genre : (bt ++ as ++ dd ++ lang))
- , el "document-info" ([ el "program-used" "pandoc" ] -- FIXME: +version
- ++ coverpage)
+ , el "document-info" (el "program-used" "pandoc" : coverpage)
]
booktitle :: PandocMonad m => Meta -> FBM m [Content]
diff --git a/src/Text/Pandoc/Writers/HTML.hs b/src/Text/Pandoc/Writers/HTML.hs
index 5d5c88dd9..9e2347798 100644
--- a/src/Text/Pandoc/Writers/HTML.hs
+++ b/src/Text/Pandoc/Writers/HTML.hs
@@ -56,7 +56,8 @@ import qualified Data.Text.Lazy as TL
import Network.HTTP (urlEncode)
import Network.URI (URI (..), parseURIReference, unEscapeString)
import Numeric (showHex)
-import Text.Blaze.Internal (customLeaf, MarkupM(Empty))
+import Text.Blaze.Internal
+ (customLeaf, MarkupM(Empty), preEscapedString, preEscapedText)
import Text.Blaze.Html hiding (contents)
import Text.Pandoc.Definition
import Text.Pandoc.Highlighting (formatHtmlBlock, formatHtmlInline, highlight,
@@ -424,7 +425,7 @@ elementToHtml slideLevel opts (Sec level num (id',classes,keyvals) title' elemen
modify (\st -> st{ stElement = False})
return res
- let isSec (Sec{}) = True
+ let isSec Sec{} = True
isSec (Blk _) = False
let isPause (Blk x) = x == Para [Str ".",Space,Str ".",Space,Str "."]
isPause _ = False
@@ -618,7 +619,7 @@ imageExts = [ "art", "bmp", "cdr", "cdt", "cpt", "cr2", "crw", "djvu", "erf",
treatAsImage :: FilePath -> Bool
treatAsImage fp =
- let path = fromMaybe fp (uriPath `fmap` parseURIReference fp)
+ let path = maybe fp uriPath (parseURIReference fp)
ext = map toLower $ drop 1 $ takeExtension path
in null ext || ext `elem` imageExts
@@ -797,8 +798,8 @@ blockToHtml opts (OrderedList (startnum, numstyle, _) lst) = do
let numstyle' = case numstyle of
Example -> "decimal"
_ -> camelCaseToHyphenated $ show numstyle
- let attribs = ([A.start $ toValue startnum | startnum /= 1]) ++
- ([A.class_ "example" | numstyle == Example]) ++
+ let attribs = [A.start $ toValue startnum | startnum /= 1] ++
+ [A.class_ "example" | numstyle == Example] ++
(if numstyle /= DefaultStyle
then if html5
then [A.type_ $
@@ -819,7 +820,7 @@ blockToHtml opts (DefinitionList lst) = do
do term' <- if null term
then return mempty
else liftM H.dt $ inlineListToHtml opts term
- defs' <- mapM (liftM (\x -> H.dd $ (x >> nl opts)) .
+ defs' <- mapM (liftM (\x -> H.dd (x >> nl opts)) .
blockListToHtml opts) defs
return $ mconcat $ nl opts : term' : nl opts :
intersperse (nl opts) defs') lst
diff --git a/src/Text/Pandoc/Writers/Haddock.hs b/src/Text/Pandoc/Writers/Haddock.hs
index 9ed3be6cf..688c1f390 100644
--- a/src/Text/Pandoc/Writers/Haddock.hs
+++ b/src/Text/Pandoc/Writers/Haddock.hs
@@ -168,8 +168,7 @@ blockToHaddock opts (OrderedList (start,_,delim) items) = do
let markers' = map (\m -> if length m < 3
then m ++ replicate (3 - length m) ' '
else m) markers
- contents <- mapM (uncurry (orderedListItemToHaddock opts)) $
- zip markers' items
+ contents <- zipWithM (orderedListItemToHaddock opts) markers' items
return $ cat contents <> blankline
blockToHaddock opts (DefinitionList items) = do
contents <- mapM (definitionListItemToHaddock opts) items
diff --git a/src/Text/Pandoc/Writers/ICML.hs b/src/Text/Pandoc/Writers/ICML.hs
index 80d2fcbef..a5d851e40 100644
--- a/src/Text/Pandoc/Writers/ICML.hs
+++ b/src/Text/Pandoc/Writers/ICML.hs
@@ -154,7 +154,7 @@ writeICML opts (Pandoc meta blocks) = do
-- | Auxilary functions for parStylesToDoc and charStylesToDoc.
contains :: String -> (String, (String, String)) -> [(String, String)]
contains s rule =
- [snd rule | isInfixOf (fst rule) s]
+ [snd rule | (fst rule) `isInfixOf` s]
-- | The monospaced font to use as default.
monospacedFont :: Doc
@@ -282,7 +282,7 @@ hyperlinksToDoc (x:xs) = hyp x $$ hyperlinksToDoc xs
("Source","htss-"++show ident), ("Visible","true"), ("DestinationUniqueKey","1")]
$ inTags True "Properties" []
$ inTags False "BorderColor" [("type","enumeration")] (text "Black")
- $$ inTags False "Destination" [("type","object")] (text $ "HyperlinkURLDestination/"++(escapeColons (escapeStringForXML url))) -- HyperlinkURLDestination with more than one colon crashes CS6
+ $$ inTags False "Destination" [("type","object")] (text $ "HyperlinkURLDestination/"++escapeColons (escapeStringForXML url)) -- HyperlinkURLDestination with more than one colon crashes CS6
-- | Convert a list of Pandoc blocks to ICML.
diff --git a/src/Text/Pandoc/Writers/LaTeX.hs b/src/Text/Pandoc/Writers/LaTeX.hs
index de2cc3480..fa72f0f1a 100644
--- a/src/Text/Pandoc/Writers/LaTeX.hs
+++ b/src/Text/Pandoc/Writers/LaTeX.hs
@@ -41,7 +41,7 @@ import Data.Char (isAlphaNum, isAscii, isDigit, isLetter, isPunctuation, ord,
toLower)
import Data.List (foldl', intercalate, intersperse, isInfixOf, nubBy,
stripPrefix, (\\))
-import Data.Maybe (catMaybes, fromMaybe, isJust, mapMaybe)
+import Data.Maybe (catMaybes, fromMaybe, isJust, mapMaybe, isNothing)
import Data.Text (Text)
import qualified Data.Text as T
import Network.URI (unEscapeString)
@@ -401,7 +401,7 @@ elementToBeamer slideLevel (Sec lvl _num (ident,classes,kvs) tit elts)
let frameoptions = ["allowdisplaybreaks", "allowframebreaks", "fragile",
"b", "c", "t", "environment",
"label", "plain", "shrink", "standout"]
- let optionslist = ["fragile" | fragile && lookup "fragile" kvs == Nothing] ++
+ let optionslist = ["fragile" | fragile && isNothing (lookup "fragile" kvs)] ++
[k | k <- classes, k `elem` frameoptions] ++
[k ++ "=" ++ v | (k,v) <- kvs, k `elem` frameoptions]
let options = if null optionslist
@@ -819,7 +819,7 @@ listItemToLaTeX lst
-- we need to put some text before a header if it's the first
-- element in an item. This will look ugly in LaTeX regardless, but
-- this will keep the typesetter from throwing an error.
- | (Header _ _ _ :_) <- lst =
+ | (Header{} :_) <- lst =
blockListToLaTeX lst >>= return . (text "\\item ~" $$) . nest 2
| otherwise = blockListToLaTeX lst >>= return . (text "\\item" $$) .
nest 2
@@ -856,7 +856,7 @@ sectionHeader unnumbered ident level lst = do
plain <- stringToLaTeX TextString $ concatMap stringify lst
let removeInvalidInline (Note _) = []
removeInvalidInline (Span (id', _, _) _) | not (null id') = []
- removeInvalidInline (Image{}) = []
+ removeInvalidInline Image{} = []
removeInvalidInline x = [x]
let lstNoNotes = foldr (mappend . (\x -> walkM removeInvalidInline x)) mempty lst
txtNoNotes <- inlineListToLaTeX lstNoNotes
diff --git a/src/Text/Pandoc/Writers/Man.hs b/src/Text/Pandoc/Writers/Man.hs
index c1427b15c..1be955fe3 100644
--- a/src/Text/Pandoc/Writers/Man.hs
+++ b/src/Text/Pandoc/Writers/Man.hs
@@ -114,7 +114,7 @@ notesToMan :: PandocMonad m => WriterOptions -> [[Block]] -> StateT WriterState
notesToMan opts notes =
if null notes
then return empty
- else mapM (uncurry (noteToMan opts)) (zip [1..] notes) >>=
+ else zipWithM (noteToMan opts) [1..] notes >>=
return . (text ".SH NOTES" $$) . vcat
-- | Return man representation of a note.
diff --git a/src/Text/Pandoc/Writers/Muse.hs b/src/Text/Pandoc/Writers/Muse.hs
index 7c4865da8..fbebe5c20 100644
--- a/src/Text/Pandoc/Writers/Muse.hs
+++ b/src/Text/Pandoc/Writers/Muse.hs
@@ -218,7 +218,7 @@ blockToMuse (DefinitionList items) = do
descriptionToMuse :: PandocMonad m
=> [Block]
-> StateT WriterState m Doc
- descriptionToMuse desc = (hang 4 " :: ") <$> blockListToMuse desc
+ descriptionToMuse desc = hang 4 " :: " <$> blockListToMuse desc
blockToMuse (Header level (ident,_,_) inlines) = do
opts <- gets stOptions
contents <- inlineListToMuse inlines
diff --git a/src/Text/Pandoc/Writers/OOXML.hs b/src/Text/Pandoc/Writers/OOXML.hs
index 2a9b9bc84..30d8d72dd 100644
--- a/src/Text/Pandoc/Writers/OOXML.hs
+++ b/src/Text/Pandoc/Writers/OOXML.hs
@@ -104,5 +104,5 @@ fitToPage :: (Double, Double) -> Integer -> (Integer, Integer)
fitToPage (x, y) pageWidth
-- Fixes width to the page width and scales the height
| x > fromIntegral pageWidth =
- (pageWidth, floor $ ((fromIntegral pageWidth) / x) * y)
+ (pageWidth, floor $ (fromIntegral pageWidth / x) * y)
| otherwise = (floor x, floor y)
diff --git a/src/Text/Pandoc/Writers/OpenDocument.hs b/src/Text/Pandoc/Writers/OpenDocument.hs
index e0097f507..17edc0cbd 100644
--- a/src/Text/Pandoc/Writers/OpenDocument.hs
+++ b/src/Text/Pandoc/Writers/OpenDocument.hs
@@ -594,7 +594,7 @@ paraStyle attrs = do
tight = if t then [ ("fo:margin-top" , "0in" )
, ("fo:margin-bottom" , "0in" )]
else []
- indent = if (i /= 0 || b)
+ indent = if i /= 0 || b
then [ ("fo:margin-left" , indentVal)
, ("fo:margin-right" , "0in" )
, ("fo:text-indent" , "0in" )
diff --git a/src/Text/Pandoc/Writers/Powerpoint.hs b/src/Text/Pandoc/Writers/Powerpoint.hs
index acb33f582..645a4cb86 100644
--- a/src/Text/Pandoc/Writers/Powerpoint.hs
+++ b/src/Text/Pandoc/Writers/Powerpoint.hs
@@ -1,4 +1,4 @@
-{-# LANGUAGE PatternGuards #-}
+
{-
Copyright (C) 2017-2018 Jesse Rosenthal <jrosenthal@jhu.edu>
diff --git a/src/Text/Pandoc/Writers/Powerpoint/Presentation.hs b/src/Text/Pandoc/Writers/Powerpoint/Presentation.hs
index f5f7d850f..0cf01ee01 100644
--- a/src/Text/Pandoc/Writers/Powerpoint/Presentation.hs
+++ b/src/Text/Pandoc/Writers/Powerpoint/Presentation.hs
@@ -72,7 +72,7 @@ import qualified Text.Pandoc.Shared as Shared -- so we don't overlap "Element"
import Text.Pandoc.Writers.Shared (metaValueToInlines)
import qualified Data.Map as M
import qualified Data.Set as S
-import Data.Maybe (maybeToList)
+import Data.Maybe (maybeToList, fromMaybe)
import Text.Pandoc.Highlighting
import qualified Data.Text as T
import Control.Applicative ((<|>))
@@ -136,7 +136,7 @@ reservedSlideIds = S.fromList [ metadataSlideId
uniqueSlideId' :: Integer -> S.Set SlideId -> String -> SlideId
uniqueSlideId' n idSet s =
- let s' = if n == 0 then s else (s ++ "-" ++ show n)
+ let s' = if n == 0 then s else s ++ "-" ++ show n
in if SlideId s' `S.member` idSet
then uniqueSlideId' (n+1) idSet s
else SlideId s'
@@ -152,7 +152,7 @@ runUniqueSlideId s = do
return sldId
addLogMessage :: LogMessage -> Pres ()
-addLogMessage msg = modify $ \st -> st{stLog = msg : (stLog st)}
+addLogMessage msg = modify $ \st -> st{stLog = msg : stLog st}
type Pres = ReaderT WriterEnv (State WriterState)
@@ -180,7 +180,7 @@ data DocProps = DocProps { dcTitle :: Maybe String
data Slide = Slide { slideId :: SlideId
, slideLayout :: Layout
- , slideNotes :: (Maybe Notes)
+ , slideNotes :: Maybe Notes
} deriving (Show, Eq)
newtype SlideId = SlideId String
@@ -345,12 +345,12 @@ inlineToParElems (SmallCaps ils) =
inlineToParElems Space = inlineToParElems (Str " ")
inlineToParElems SoftBreak = inlineToParElems (Str " ")
inlineToParElems LineBreak = return [Break]
-inlineToParElems (Link _ ils (url, title)) = do
+inlineToParElems (Link _ ils (url, title)) =
local (\r ->r{envRunProps = (envRunProps r){rLink = Just $ ExternalTarget (url, title)}}) $
- inlinesToParElems ils
-inlineToParElems (Code _ str) = do
+ inlinesToParElems ils
+inlineToParElems (Code _ str) =
local (\r ->r{envRunProps = (envRunProps r){rPropCode = True}}) $
- inlineToParElems $ Str str
+ inlineToParElems $ Str str
inlineToParElems (Math mathtype str) =
return [MathElem mathtype (TeXString str)]
inlineToParElems (Note blks) = do
@@ -409,7 +409,7 @@ blockToParagraphs (CodeBlock attr str) =
Just sty ->
case highlight synMap (formatSourceLines sty) attr str of
Right pElems -> do pProps <- asks envParaProps
- return $ [Paragraph pProps pElems]
+ return [Paragraph pProps pElems]
Left _ -> blockToParagraphs $ Para [Str str]
Nothing -> blockToParagraphs $ Para [Str str]
-- We can't yet do incremental lists, but we should render a
@@ -463,7 +463,7 @@ blockToParagraphs (DefinitionList entries) = do
definition <- concatMapM (blockToParagraphs . BlockQuote) blksLst
return $ term ++ definition
concatMapM go entries
-blockToParagraphs (Div (_, ("notes" : []), _) _) = return []
+blockToParagraphs (Div (_, "notes" : [], _) _) = return []
blockToParagraphs (Div _ blks) = concatMapM blockToParagraphs blks
blockToParagraphs blk = do
addLogMessage $ BlockNotRendered blk
@@ -481,7 +481,7 @@ multiParBullet (b:bs) = do
cellToParagraphs :: Alignment -> TableCell -> Pres [Paragraph]
cellToParagraphs algn tblCell = do
- paras <- mapM (blockToParagraphs) tblCell
+ paras <- mapM blockToParagraphs tblCell
let alignment = case algn of
AlignLeft -> Just AlgnLeft
AlignRight -> Just AlgnRight
@@ -494,7 +494,7 @@ rowToParagraphs :: [Alignment] -> [TableCell] -> Pres [[Paragraph]]
rowToParagraphs algns tblCells = do
-- We have to make sure we have the right number of alignments
let pairs = zip (algns ++ repeat AlignDefault) tblCells
- mapM (\(a, tc) -> cellToParagraphs a tc) pairs
+ mapM (uncurry cellToParagraphs) pairs
withAttr :: Attr -> Shape -> Shape
withAttr attr (Pic picPr url caption) =
@@ -507,17 +507,17 @@ withAttr _ sp = sp
blockToShape :: Block -> Pres Shape
blockToShape (Plain (il:_)) | Image attr ils (url, _) <- il =
- (withAttr attr . Pic def url) <$> (inlinesToParElems ils)
+ (withAttr attr . Pic def url) <$> inlinesToParElems ils
blockToShape (Para (il:_)) | Image attr ils (url, _) <- il =
- (withAttr attr . Pic def url) <$> (inlinesToParElems ils)
+ (withAttr attr . Pic def url) <$> inlinesToParElems ils
blockToShape (Plain (il:_)) | Link _ (il':_) target <- il
, Image attr ils (url, _) <- il' =
(withAttr attr . Pic def {picPropLink = Just $ ExternalTarget target} url) <$>
- (inlinesToParElems ils)
+ inlinesToParElems ils
blockToShape (Para (il:_)) | Link _ (il':_) target <- il
, Image attr ils (url, _) <- il' =
(withAttr attr . Pic def{picPropLink = Just $ ExternalTarget target} url) <$>
- (inlinesToParElems ils)
+ inlinesToParElems ils
blockToShape (Table caption algn _ hdrCells rows) = do
caption' <- inlinesToParElems caption
hdrCells' <- rowToParagraphs algn hdrCells
@@ -537,11 +537,11 @@ blockToShape blk = do paras <- blockToParagraphs blk
combineShapes :: [Shape] -> [Shape]
combineShapes [] = []
-combineShapes (s : []) = [s]
-combineShapes (pic@(Pic _ _ _) : ss) = pic : combineShapes ss
-combineShapes ((TextBox []) : ss) = combineShapes ss
+combineShapes[s] = [s]
+combineShapes (pic@(Pic{}) : ss) = pic : combineShapes ss
+combineShapes (TextBox [] : ss) = combineShapes ss
combineShapes (s : TextBox [] : ss) = combineShapes (s : ss)
-combineShapes ((TextBox (p:ps)) : (TextBox (p':ps')) : ss) =
+combineShapes (TextBox (p:ps) : TextBox (p':ps') : ss) =
combineShapes $ TextBox ((p:ps) ++ (p':ps')) : ss
combineShapes (s:ss) = s : combineShapes ss
@@ -549,8 +549,8 @@ blocksToShapes :: [Block] -> Pres [Shape]
blocksToShapes blks = combineShapes <$> mapM blockToShape blks
isImage :: Inline -> Bool
-isImage (Image _ _ _) = True
-isImage (Link _ ((Image _ _ _) : _) _) = True
+isImage (Image{}) = True
+isImage (Link _ (Image _ _ _ : _) _) = True
isImage _ = False
splitBlocks' :: [Block] -> [[Block]] -> [Block] -> Pres [[Block]]
@@ -565,27 +565,27 @@ splitBlocks' cur acc (h@(Header n _ _) : blks) = do
GT -> splitBlocks' (cur ++ [h]) acc blks
-- `blockToParagraphs` treats Plain and Para the same, so we can save
-- some code duplication by treating them the same here.
-splitBlocks' cur acc ((Plain ils) : blks) = splitBlocks' cur acc ((Para ils) : blks)
-splitBlocks' cur acc ((Para (il:ils)) : blks) | isImage il = do
+splitBlocks' cur acc (Plain ils : blks) = splitBlocks' cur acc (Para ils : blks)
+splitBlocks' cur acc (Para (il:ils) : blks) | isImage il = do
slideLevel <- asks envSlideLevel
case cur of
- (Header n _ _) : [] | n == slideLevel ->
+ [(Header n _ _)] | n == slideLevel ->
splitBlocks' []
(acc ++ [cur ++ [Para [il]]])
- (if null ils then blks else (Para ils) : blks)
+ (if null ils then blks else Para ils : blks)
_ -> splitBlocks' []
(acc ++ (if null cur then [] else [cur]) ++ [[Para [il]]])
- (if null ils then blks else (Para ils) : blks)
-splitBlocks' cur acc (tbl@(Table _ _ _ _ _) : blks) = do
+ (if null ils then blks else Para ils : blks)
+splitBlocks' cur acc (tbl@(Table{}) : blks) = do
slideLevel <- asks envSlideLevel
case cur of
- (Header n _ _) : [] | n == slideLevel ->
+ [(Header n _ _)] | n == slideLevel ->
splitBlocks' [] (acc ++ [cur ++ [tbl]]) blks
_ -> splitBlocks' [] (acc ++ (if null cur then [] else [cur]) ++ [[tbl]]) blks
splitBlocks' cur acc (d@(Div (_, classes, _) _): blks) | "columns" `elem` classes = do
slideLevel <- asks envSlideLevel
case cur of
- (Header n _ _) : [] | n == slideLevel ->
+ [(Header n _ _)] | n == slideLevel ->
splitBlocks' [] (acc ++ [cur ++ [d]]) blks
_ -> splitBlocks' [] (acc ++ (if null cur then [] else [cur]) ++ [[d]]) blks
splitBlocks' cur acc (blk : blks) = splitBlocks' (cur ++ [blk]) acc blks
@@ -594,12 +594,12 @@ splitBlocks :: [Block] -> Pres [[Block]]
splitBlocks = splitBlocks' [] []
blocksToSlide' :: Int -> [Block] -> Pres Slide
-blocksToSlide' lvl ((Header n (ident, _, _) ils) : blks)
+blocksToSlide' lvl (Header n (ident, _, _) ils : blks)
| n < lvl = do
registerAnchorId ident
sldId <- asks envCurSlideId
hdr <- inlinesToParElems ils
- return $ Slide sldId (TitleSlide {titleSlideHeader = hdr}) Nothing
+ return $ Slide sldId TitleSlide {titleSlideHeader = hdr} Nothing
| n == lvl = do
registerAnchorId ident
hdr <- inlinesToParElems ils
@@ -614,7 +614,7 @@ blocksToSlide' lvl ((Header n (ident, _, _) ils) : blks)
blocksToSlide' _ (blk : blks)
| Div (_, classes, _) divBlks <- blk
, "columns" `elem` classes
- , (Div (_, clsL, _) blksL) : (Div (_, clsR, _) blksR) : remaining <- divBlks
+ , Div (_, clsL, _) blksL : Div (_, clsR, _) blksR : remaining <- divBlks
, "column" `elem` clsL, "column" `elem` clsR = do
unless (null blks)
(mapM (addLogMessage . BlockNotRendered) blks >> return ())
@@ -672,7 +672,7 @@ makeNoteEntry n blks =
in
case blks of
(Para ils : blks') -> (Para $ enum : Space : ils) : blks'
- _ -> (Para [enum]) : blks
+ _ -> Para [enum] : blks
forceFontSize :: Pixels -> Pres a -> Pres a
forceFontSize px x = do
@@ -860,7 +860,7 @@ blocksToPresentationSlides blks = do
(\env -> env { envCurSlideId = endNotesSlideId
, envInNoteSlide = True
})
- (blocksToSlide $ endNotesSlideBlocks)
+ (blocksToSlide endNotesSlideBlocks)
return [endNotesSlide]
let slides = metadataslides ++ tocSlides ++ bodyslides ++ endNotesSlides
@@ -889,9 +889,7 @@ documentToPresentation :: WriterOptions
documentToPresentation opts (Pandoc meta blks) =
let env = def { envOpts = opts
, envMetadata = meta
- , envSlideLevel = case writerSlideLevel opts of
- Just lvl -> lvl
- Nothing -> getSlideLevel blks
+ , envSlideLevel = fromMaybe (getSlideLevel blks) (writerSlideLevel opts)
}
(presSlides, msgs) = runPres env def $ blocksToPresentationSlides blks
docProps = metaToDocProps meta
diff --git a/src/Text/Pandoc/Writers/RST.hs b/src/Text/Pandoc/Writers/RST.hs
index a57527aa8..95cb46643 100644
--- a/src/Text/Pandoc/Writers/RST.hs
+++ b/src/Text/Pandoc/Writers/RST.hs
@@ -132,7 +132,7 @@ keyToRST (label, (src, _)) = do
-- | Return RST representation of notes.
notesToRST :: PandocMonad m => [[Block]] -> RST m Doc
notesToRST notes =
- mapM (uncurry noteToRST) (zip [1..] notes) >>=
+ zipWithM noteToRST [1..] notes >>=
return . vsep
-- | Return RST representation of a note.
@@ -306,8 +306,7 @@ blockToRST (OrderedList (start, style', delim) items) = do
let maxMarkerLength = maximum $ map length markers
let markers' = map (\m -> let s = maxMarkerLength - length m
in m ++ replicate s ' ') markers
- contents <- mapM (uncurry orderedListItemToRST) $
- zip markers' items
+ contents <- zipWithM orderedListItemToRST markers' items
-- ensure that sublists have preceding blank line
return $ blankline $$ chomp (vcat contents) $$ blankline
blockToRST (DefinitionList items) = do
@@ -356,12 +355,12 @@ blockListToRST' topLevel blocks = do
let fixBlocks (b1:b2@(BlockQuote _):bs)
| toClose b1 = b1 : commentSep : b2 : fixBlocks bs
where
- toClose (Plain{}) = False
- toClose (Header{}) = False
- toClose (LineBlock{}) = False
- toClose (HorizontalRule) = False
+ toClose Plain{} = False
+ toClose Header{} = False
+ toClose LineBlock{} = False
+ toClose HorizontalRule = False
toClose (Para [Image _ _ (_,'f':'i':'g':':':_)]) = True
- toClose (Para{}) = False
+ toClose Para{} = False
toClose _ = True
commentSep = RawBlock "rst" "..\n\n"
fixBlocks (b:bs) = b : fixBlocks bs
diff --git a/src/Text/Pandoc/Writers/RTF.hs b/src/Text/Pandoc/Writers/RTF.hs
index 790bebc01..7006b58d1 100644
--- a/src/Text/Pandoc/Writers/RTF.hs
+++ b/src/Text/Pandoc/Writers/RTF.hs
@@ -31,6 +31,7 @@ Conversion of 'Pandoc' documents to RTF (rich text format).
module Text.Pandoc.Writers.RTF ( writeRTF
) where
import Control.Monad.Except (catchError, throwError)
+import Control.Monad
import qualified Data.ByteString as B
import Data.Char (chr, isDigit, ord)
import Data.List (intercalate, isSuffixOf)
@@ -278,8 +279,7 @@ blockToRTF indent alignment (BulletList lst) = (spaceAtEnd . concat) <$>
mapM (listItemToRTF alignment indent (bulletMarker indent)) lst
blockToRTF indent alignment (OrderedList attribs lst) =
(spaceAtEnd . concat) <$>
- mapM (uncurry (listItemToRTF alignment indent))
- (zip (orderedMarkers indent attribs) lst)
+ zipWithM (listItemToRTF alignment indent) (orderedMarkers indent attribs) lst
blockToRTF indent alignment (DefinitionList lst) = (spaceAtEnd . concat) <$>
mapM (definitionListItemToRTF alignment indent) lst
blockToRTF indent _ HorizontalRule = return $
@@ -303,8 +303,8 @@ tableRowToRTF header indent aligns sizes' cols = do
let sizes = if all (== 0) sizes'
then replicate (length cols) (1.0 / fromIntegral (length cols))
else sizes'
- columns <- concat <$> mapM (uncurry (tableItemToRTF indent))
- (zip aligns cols)
+ columns <- concat <$>
+ zipWithM (tableItemToRTF indent) aligns cols
let rightEdges = tail $ scanl (\sofar new -> sofar + floor (new * totalTwips))
(0 :: Integer) sizes
let cellDefs = map (\edge -> (if header
diff --git a/src/Text/Pandoc/Writers/Texinfo.hs b/src/Text/Pandoc/Writers/Texinfo.hs
index b5d72aa56..bf434642e 100644
--- a/src/Text/Pandoc/Writers/Texinfo.hs
+++ b/src/Text/Pandoc/Writers/Texinfo.hs
@@ -475,7 +475,7 @@ inlineToTexinfo (Link _ txt (src@('#':_), _)) = do
inlineToTexinfo (Link _ txt (src, _)) =
case txt of
[Str x] | escapeURI x == src -> -- autolink
- do return $ text $ "@url{" ++ x ++ "}"
+ return $ text $ "@url{" ++ x ++ "}"
_ -> do contents <- escapeCommas $ inlineListToTexinfo txt
let src1 = stringToTexinfo src
return $ text ("@uref{" ++ src1 ++ ",") <> contents <>
diff --git a/test/Tests/Command.hs b/test/Tests/Command.hs
index 4999ff45a..de83d0639 100644
--- a/test/Tests/Command.hs
+++ b/test/Tests/Command.hs
@@ -40,7 +40,7 @@ runTest testname pandocpath cmd inp norm = testCase testname $ do
-- filter \r so the tests will work on Windows machines
let out = filter (/= '\r') $ err' ++ out'
result <- if ec == ExitSuccess
- then do
+ then
if out == norm
then return TestPassed
else return
@@ -52,6 +52,7 @@ runTest testname pandocpath cmd inp norm = testCase testname $ do
assertBool (show result) (result == TestPassed)
tests :: TestTree
+{-# NOINLINE tests #-}
tests = unsafePerformIO $ do
pandocpath <- findPandoc
files <- filter (".md" `isSuffixOf`) <$>
@@ -89,7 +90,6 @@ extractCommandTest pandocpath fp = unsafePerformIO $ do
contents <- UTF8.toText <$> BS.readFile ("command" </> fp)
Pandoc _ blocks <- runIOorExplode (readMarkdown
def{ readerExtensions = pandocExtensions } contents)
- let codeblocks = map extractCode $ filter isCodeBlock $ blocks
+ let codeblocks = map extractCode $ filter isCodeBlock blocks
let cases = map (runCommandTest pandocpath) $ zip [1..] codeblocks
return $ testGroup fp cases
-
diff --git a/test/Tests/Readers/Creole.hs b/test/Tests/Readers/Creole.hs
index 3a21df738..3f60a523d 100644
--- a/test/Tests/Readers/Creole.hs
+++ b/test/Tests/Readers/Creole.hs
@@ -224,7 +224,7 @@ tests = [
<> " bar")
, "escaped auto link" =:
"foo ~http://foo.example.com/bar/baz.html bar"
- =?> para ("foo http://foo.example.com/bar/baz.html bar")
+ =?> para "foo http://foo.example.com/bar/baz.html bar"
, "wiki link simple" =:
"foo [[http://foo.example.com/foo.png]] bar"
=?> para ("foo "
diff --git a/test/Tests/Readers/Docx.hs b/test/Tests/Readers/Docx.hs
index d58e219de..89a605bf7 100644
--- a/test/Tests/Readers/Docx.hs
+++ b/test/Tests/Readers/Docx.hs
@@ -5,6 +5,7 @@ import qualified Data.ByteString as BS
import qualified Data.ByteString.Lazy as B
import qualified Data.Map as M
import qualified Data.Text as T
+import Data.Maybe
import System.IO.Unsafe
import Test.Tasty
import Test.Tasty.HUnit
@@ -46,7 +47,7 @@ compareOutput opts docxFile nativeFile = do
nf <- UTF8.toText <$> BS.readFile nativeFile
p <- runIOorExplode $ readDocx opts df
df' <- runIOorExplode $ readNative def nf
- return $ (noNorm p, noNorm df')
+ return (noNorm p, noNorm df')
testCompareWithOptsIO :: ReaderOptions -> String -> FilePath -> FilePath -> IO TestTree
testCompareWithOptsIO opts name docxFile nativeFile = do
@@ -87,11 +88,9 @@ compareMediaPathIO mediaPath mediaBag docxPath = do
Nothing -> error ("couldn't find " ++
mediaPath ++
" in media bag")
- docxBS = case docxMedia of
- Just bs -> bs
- Nothing -> error ("couldn't find " ++
- mediaPath ++
- " in media bag")
+ docxBS = fromMaybe (error ("couldn't find " ++
+ mediaPath ++
+ " in media bag")) docxMedia
return $ mbBS == docxBS
compareMediaBagIO :: FilePath -> IO Bool
diff --git a/test/Tests/Readers/EPUB.hs b/test/Tests/Readers/EPUB.hs
index 201fd10a5..1337a9c11 100644
--- a/test/Tests/Readers/EPUB.hs
+++ b/test/Tests/Readers/EPUB.hs
@@ -17,7 +17,7 @@ getMediaBag fp = do
testMediaBag :: FilePath -> [(String, String, Int)] -> IO ()
testMediaBag fp bag = do
- actBag <- (mediaDirectory <$> getMediaBag fp)
+ actBag <- mediaDirectory <$> getMediaBag fp
assertBool (show "MediaBag did not match:\nExpected: "
++ show bag
++ "\nActual: "
diff --git a/test/Tests/Readers/Muse.hs b/test/Tests/Readers/Muse.hs
index c92b395ff..76c18135e 100644
--- a/test/Tests/Readers/Muse.hs
+++ b/test/Tests/Readers/Muse.hs
@@ -30,8 +30,8 @@ spcSep = mconcat . intersperse space
-- Tables and definition lists don't round-trip yet
makeRoundTrip :: Block -> Block
-makeRoundTrip (Table{}) = Para [Str "table was here"]
-makeRoundTrip (DefinitionList{}) = Para [Str "deflist was here"]
+makeRoundTrip Table{} = Para [Str "table was here"]
+makeRoundTrip DefinitionList{} = Para [Str "deflist was here"]
makeRoundTrip x = x
-- Demand that any AST produced by Muse reader and written by Muse writer can be read back exactly the same way.
@@ -44,7 +44,7 @@ roundTrip b = d'' == d'''
d'' = rewrite d'
d''' = rewrite d''
rewrite = amuse . T.pack . (++ "\n") . T.unpack .
- (purely $ writeMuse def { writerExtensions = extensionsFromList [Ext_amuse]
+ purely (writeMuse def { writerExtensions = extensionsFromList [Ext_amuse]
, writerWrapText = WrapPreserve
})
diff --git a/test/Tests/Readers/Org/Block/Header.hs b/test/Tests/Readers/Org/Block/Header.hs
index d895c86e2..e8ad88558 100644
--- a/test/Tests/Readers/Org/Block/Header.hs
+++ b/test/Tests/Readers/Org/Block/Header.hs
@@ -130,7 +130,7 @@ tests =
mconcat [ para "foo"
, headerWith ("thing-other-thing", [], [])
1
- ((strikeout "thing") <> " other thing")
+ (strikeout "thing" <> " other thing")
]
, "Comment Trees" =:
diff --git a/test/Tests/Readers/Org/Block/List.hs b/test/Tests/Readers/Org/Block/List.hs
index 32bb13294..343682a80 100644
--- a/test/Tests/Readers/Org/Block/List.hs
+++ b/test/Tests/Readers/Org/Block/List.hs
@@ -75,16 +75,16 @@ tests =
]
, "Bullet List with Decreasing Indent" =:
- (" - Discovery\n\
- \ - Human After All\n") =?>
+ " - Discovery\n\
+ \ - Human After All\n" =?>
mconcat [ bulletList [ plain "Discovery" ]
, bulletList [ plain ("Human" <> space <> "After" <> space <> "All")]
]
, "Header follows Bullet List" =:
- (" - Discovery\n\
+ " - Discovery\n\
\ - Human After All\n\
- \* Homework") =?>
+ \* Homework" =?>
mconcat [ bulletList [ plain "Discovery"
, plain ("Human" <> space <> "After" <> space <> "All")
]
@@ -92,9 +92,9 @@ tests =
]
, "Bullet List Unindented with trailing Header" =:
- ("- Discovery\n\
+ "- Discovery\n\
\- Homework\n\
- \* NotValidListItem") =?>
+ \* NotValidListItem" =?>
mconcat [ bulletList [ plain "Discovery"
, plain "Homework"
]
@@ -166,14 +166,14 @@ tests =
, "Ordered List in Bullet List" =:
("- Emacs\n" <>
" 1. Org\n") =?>
- bulletList [ (plain "Emacs") <>
- (orderedList [ plain "Org"])
+ bulletList [ plain "Emacs" <>
+ orderedList [ plain "Org"]
]
, "Bullet List in Ordered List" =:
("1. GNU\n" <>
" - Freedom\n") =?>
- orderedList [ (plain "GNU") <> bulletList [ (plain "Freedom") ] ]
+ orderedList [ plain "GNU" <> bulletList [ plain "Freedom" ] ]
, "Definition List" =:
T.unlines [ "- PLL :: phase-locked loop"
diff --git a/test/Tests/Readers/Org/Directive.hs b/test/Tests/Readers/Org/Directive.hs
index 862315ef3..7e2c0fb8d 100644
--- a/test/Tests/Readers/Org/Directive.hs
+++ b/test/Tests/Readers/Org/Directive.hs
@@ -107,8 +107,8 @@ tests =
] =?>
mconcat [ para "first block"
, orderedList
- [ (para "top-level section 1" <>
- orderedList [ para "subsection" ])
+ [ para "top-level section 1" <>
+ orderedList [ para "subsection" ]
, para "top-level section 2" ]
]
diff --git a/test/Tests/Readers/Org/Inline.hs b/test/Tests/Readers/Org/Inline.hs
index cb50ba630..da0d1db0b 100644
--- a/test/Tests/Readers/Org/Inline.hs
+++ b/test/Tests/Readers/Org/Inline.hs
@@ -36,7 +36,7 @@ tests =
, "Underline" =:
"_underline_" =?>
- para (underlineSpan $ "underline")
+ para (underlineSpan "underline")
, "Strikeout" =:
"+Kill Bill+" =?>
@@ -127,11 +127,12 @@ tests =
, "Markup should work properly after a blank line" =:
T.unlines ["foo", "", "/bar/"] =?>
- (para $ text "foo") <> (para $ emph $ text "bar")
+ para (text "foo") <>
+ para (emph $ text "bar")
, "Inline math must stay within three lines" =:
T.unlines [ "$a", "b", "c$", "$d", "e", "f", "g$" ] =?>
- para ((math "a\nb\nc") <> softbreak <>
+ para (math "a\nb\nc" <> softbreak <>
"$d" <> softbreak <> "e" <> softbreak <>
"f" <> softbreak <> "g$")
@@ -139,7 +140,7 @@ tests =
"$a$ $b$! $c$?" =?>
para (spcSep [ math "a"
, "$b$!"
- , (math "c") <> "?"
+ , math "c" <> "?"
])
, "Markup may not span more than two lines" =:
@@ -166,12 +167,12 @@ tests =
para (mconcat $ intersperse softbreak
[ "a" <> subscript "(a(b)(c)d)"
, "e" <> superscript "(f(g)h)"
- , "i" <> (subscript "(jk)") <> "l)"
- , "m" <> (superscript "()") <> "n"
+ , "i" <> subscript "(jk)" <> "l)"
+ , "m" <> superscript "()" <> "n"
, "o" <> subscript "p{q{}r}"
, "s" <> superscript "t{u}v"
- , "w" <> (subscript "xy") <> "z}"
- , "1" <> (superscript "") <> "2"
+ , "w" <> subscript "xy" <> "z}"
+ , "1" <> superscript "" <> "2"
, "3" <> subscript "{}"
, "4" <> superscript ("(a(" <> strong "b(c" <> ")d))")
])
@@ -182,17 +183,17 @@ tests =
, testGroup "Images"
[ "Image" =:
"[[./sunset.jpg]]" =?>
- (para $ image "./sunset.jpg" "" "")
+ para (image "./sunset.jpg" "" "")
, "Image with explicit file: prefix" =:
"[[file:sunrise.jpg]]" =?>
- (para $ image "sunrise.jpg" "" "")
+ para (image "sunrise.jpg" "" "")
, "Multiple images within a paragraph" =:
T.unlines [ "[[file:sunrise.jpg]]"
, "[[file:sunset.jpg]]"
] =?>
- (para $ (image "sunrise.jpg" "" "")
+ para ((image "sunrise.jpg" "" "")
<> softbreak
<> (image "sunset.jpg" "" ""))
@@ -200,75 +201,75 @@ tests =
T.unlines [ "#+ATTR_HTML: :width 50%"
, "[[file:guinea-pig.gif]]"
] =?>
- (para $ imageWith ("", [], [("width", "50%")]) "guinea-pig.gif" "" "")
+ para (imageWith ("", [], [("width", "50%")]) "guinea-pig.gif" "" "")
]
, "Explicit link" =:
"[[http://zeitlens.com/][pseudo-random /nonsense/]]" =?>
- (para $ link "http://zeitlens.com/" ""
+ para (link "http://zeitlens.com/" ""
("pseudo-random" <> space <> emph "nonsense"))
, "Self-link" =:
"[[http://zeitlens.com/]]" =?>
- (para $ link "http://zeitlens.com/" "" "http://zeitlens.com/")
+ para (link "http://zeitlens.com/" "" "http://zeitlens.com/")
, "Absolute file link" =:
"[[/url][hi]]" =?>
- (para $ link "file:///url" "" "hi")
+ para (link "file:///url" "" "hi")
, "Link to file in parent directory" =:
"[[../file.txt][moin]]" =?>
- (para $ link "../file.txt" "" "moin")
+ para (link "../file.txt" "" "moin")
, "Empty link (for gitit interop)" =:
"[[][New Link]]" =?>
- (para $ link "" "" "New Link")
+ para (link "" "" "New Link")
, "Image link" =:
"[[sunset.png][file:dusk.svg]]" =?>
- (para $ link "sunset.png" "" (image "dusk.svg" "" ""))
+ para (link "sunset.png" "" (image "dusk.svg" "" ""))
, "Image link with non-image target" =:
"[[http://example.com][./logo.png]]" =?>
- (para $ link "http://example.com" "" (image "./logo.png" "" ""))
+ para (link "http://example.com" "" (image "./logo.png" "" ""))
, "Plain link" =:
"Posts on http://zeitlens.com/ can be funny at times." =?>
- (para $ spcSep [ "Posts", "on"
+ para (spcSep [ "Posts", "on"
, link "http://zeitlens.com/" "" "http://zeitlens.com/"
, "can", "be", "funny", "at", "times."
])
, "Angle link" =:
"Look at <http://moltkeplatz.de> for fnords." =?>
- (para $ spcSep [ "Look", "at"
+ para (spcSep [ "Look", "at"
, link "http://moltkeplatz.de" "" "http://moltkeplatz.de"
, "for", "fnords."
])
, "Absolute file link" =:
"[[file:///etc/passwd][passwd]]" =?>
- (para $ link "file:///etc/passwd" "" "passwd")
+ para (link "file:///etc/passwd" "" "passwd")
, "File link" =:
"[[file:target][title]]" =?>
- (para $ link "target" "" "title")
+ para (link "target" "" "title")
, "Anchor" =:
"<<anchor>> Link here later." =?>
- (para $ spanWith ("anchor", [], []) mempty <>
+ para (spanWith ("anchor", [], []) mempty <>
"Link" <> space <> "here" <> space <> "later.")
, "Inline code block" =:
"src_emacs-lisp{(message \"Hello\")}" =?>
- (para $ codeWith ( ""
+ para (codeWith ( ""
, [ "commonlisp" ]
, [ ("org-language", "emacs-lisp") ])
"(message \"Hello\")")
, "Inline code block with arguments" =:
"src_sh[:export both :results output]{echo 'Hello, World'}" =?>
- (para $ codeWith ( ""
+ para (codeWith ( ""
, [ "bash" ]
, [ ("org-language", "sh")
, ("export", "both")
@@ -279,7 +280,7 @@ tests =
, "Inline code block with toggle" =:
"src_sh[:toggle]{echo $HOME}" =?>
- (para $ codeWith ( ""
+ para (codeWith ( ""
, [ "bash" ]
, [ ("org-language", "sh")
, ("toggle", "yes")
@@ -415,7 +416,7 @@ tests =
in [
"Berkeley-style in-text citation" =:
"See @Dominik201408." =?>
- (para $ "See "
+ para ("See "
<> cite [dominikInText] "@Dominik201408"
<> ".")
@@ -468,7 +469,7 @@ tests =
, "MathML symbol in LaTeX-style" =:
"There is a hackerspace in Lübeck, Germany, called nbsp (unicode symbol: '\\nbsp')." =?>
- para ("There is a hackerspace in Lübeck, Germany, called nbsp (unicode symbol: ' ').")
+ para "There is a hackerspace in Lübeck, Germany, called nbsp (unicode symbol: ' ')."
, "MathML symbol in LaTeX-style, including braces" =:
"\\Aacute{}stor" =?>
diff --git a/test/Tests/Readers/Org/Inline/Note.hs b/test/Tests/Readers/Org/Inline/Note.hs
index 46416d7d8..9eb1d02d6 100644
--- a/test/Tests/Readers/Org/Inline/Note.hs
+++ b/test/Tests/Readers/Org/Inline/Note.hs
@@ -84,4 +84,3 @@ tests =
, para "next"
]
]
-
diff --git a/test/Tests/Readers/Org/Inline/Smart.hs b/test/Tests/Readers/Org/Inline/Smart.hs
index 7a5e653cf..77f10699d 100644
--- a/test/Tests/Readers/Org/Inline/Smart.hs
+++ b/test/Tests/Readers/Org/Inline/Smart.hs
@@ -38,9 +38,9 @@ tests =
, test orgSmart "Single quotes can be followed by emphasized text"
("Singles on the '/meat market/'" =?>
- para ("Singles on the " <> (singleQuoted $ emph "meat market")))
+ para ("Singles on the " <> singleQuoted (emph "meat market")))
, test orgSmart "Double quotes can be followed by emphasized text"
("Double income, no kids: \"/DINK/\"" =?>
- para ("Double income, no kids: " <> (doubleQuoted $ emph "DINK")))
+ para ("Double income, no kids: " <> doubleQuoted (emph "DINK")))
]
diff --git a/test/Tests/Readers/Org/Meta.hs b/test/Tests/Readers/Org/Meta.hs
index 3ad6f5d8b..409cd00ae 100644
--- a/test/Tests/Readers/Org/Meta.hs
+++ b/test/Tests/Readers/Org/Meta.hs
@@ -30,32 +30,32 @@ tests =
, "Title" =:
"#+TITLE: Hello, World" =?>
let titleInline = toList $ "Hello," <> space <> "World"
- meta = setMeta "title" (MetaInlines titleInline) $ nullMeta
+ meta = setMeta "title" (MetaInlines titleInline) nullMeta
in Pandoc meta mempty
, "Author" =:
"#+author: John /Emacs-Fanboy/ Doe" =?>
let author = toList . spcSep $ [ "John", emph "Emacs-Fanboy", "Doe" ]
- meta = setMeta "author" (MetaList [MetaInlines author]) $ nullMeta
+ meta = setMeta "author" (MetaList [MetaInlines author]) nullMeta
in Pandoc meta mempty
, "Multiple authors" =:
"#+author: James Dewey Watson, Francis Harry Compton Crick " =?>
let watson = MetaInlines $ toList "James Dewey Watson"
crick = MetaInlines $ toList "Francis Harry Compton Crick"
- meta = setMeta "author" (MetaList [watson, crick]) $ nullMeta
+ meta = setMeta "author" (MetaList [watson, crick]) nullMeta
in Pandoc meta mempty
, "Date" =:
"#+Date: Feb. *28*, 2014" =?>
- let date = toList . spcSep $ [ "Feb.", (strong "28") <> ",", "2014" ]
- meta = setMeta "date" (MetaInlines date) $ nullMeta
+ let date = toList . spcSep $ [ "Feb.", strong "28" <> ",", "2014" ]
+ meta = setMeta "date" (MetaInlines date) nullMeta
in Pandoc meta mempty
, "Description" =:
"#+DESCRIPTION: Explanatory text" =?>
let description = "Explanatory text"
- meta = setMeta "description" (MetaString description) $ nullMeta
+ meta = setMeta "description" (MetaString description) nullMeta
in Pandoc meta mempty
, "Properties drawer" =:
@@ -94,7 +94,7 @@ tests =
, "#+author: Max"
] =?>
let author = MetaInlines [Str "Max"]
- meta = setMeta "author" (MetaList [author]) $ nullMeta
+ meta = setMeta "author" (MetaList [author]) nullMeta
in Pandoc meta mempty
, "Logbook drawer" =:
@@ -135,7 +135,7 @@ tests =
, "Search links are read as emph" =:
"[[Wally][Where's Wally?]]" =?>
- (para (emph $ "Where's" <> space <> "Wally?"))
+ para (emph $ "Where's" <> space <> "Wally?")
, "Link to nonexistent anchor" =:
T.unlines [ "<<link-here>> Target."
@@ -149,25 +149,25 @@ tests =
T.unlines [ "#+LINK: wp https://en.wikipedia.org/wiki/%s"
, "[[wp:Org_mode][Wikipedia on Org-mode]]"
] =?>
- (para (link "https://en.wikipedia.org/wiki/Org_mode" ""
- ("Wikipedia" <> space <> "on" <> space <> "Org-mode")))
+ para (link "https://en.wikipedia.org/wiki/Org_mode" ""
+ ("Wikipedia" <> space <> "on" <> space <> "Org-mode"))
, "Link abbreviation, defined after first use" =:
T.unlines [ "[[zl:non-sense][Non-sense articles]]"
, "#+LINK: zl http://zeitlens.com/tags/%s.html"
] =?>
- (para (link "http://zeitlens.com/tags/non-sense.html" ""
- ("Non-sense" <> space <> "articles")))
+ para (link "http://zeitlens.com/tags/non-sense.html" ""
+ ("Non-sense" <> space <> "articles"))
, "Link abbreviation, URL encoded arguments" =:
T.unlines [ "#+link: expl http://example.com/%h/foo"
, "[[expl:Hello, World!][Moin!]]"
] =?>
- (para (link "http://example.com/Hello%2C%20World%21/foo" "" "Moin!"))
+ para (link "http://example.com/Hello%2C%20World%21/foo" "" "Moin!")
, "Link abbreviation, append arguments" =:
T.unlines [ "#+link: expl http://example.com/"
, "[[expl:foo][bar]]"
] =?>
- (para (link "http://example.com/foo" "" "bar"))
+ para (link "http://example.com/foo" "" "bar")
]
diff --git a/test/Tests/Readers/RST.hs b/test/Tests/Readers/RST.hs
index 928fc1a99..3753fbf12 100644
--- a/test/Tests/Readers/RST.hs
+++ b/test/Tests/Readers/RST.hs
@@ -36,8 +36,8 @@ tests = [ "line block with blank line" =:
, ":Parameter i: integer"
, ":Final: item"
, " on two lines" ]
- =?> ( doc
- $ para "para" <>
+ =?>
+ doc (para "para" <>
definitionList [ (str "Hostname", [para "media08"])
, (text "IP address", [para "10.0.0.19"])
, (str "Size", [para "3ru"])
@@ -56,10 +56,10 @@ tests = [ "line block with blank line" =:
, ""
, ":Version: 1"
]
- =?> ( setMeta "version" (para "1")
- $ setMeta "title" ("Title" :: Inlines)
+ =?>
+ setMeta "version" (para "1") (setMeta "title" ("Title" :: Inlines)
$ setMeta "subtitle" ("Subtitle" :: Inlines)
- $ doc mempty )
+ $ doc mempty)
, "with inline markup" =: T.unlines
[ ":*Date*: today"
, ""
@@ -73,8 +73,8 @@ tests = [ "line block with blank line" =:
, ".. _two: http://example.com"
, ".. _three: http://example.org"
]
- =?> ( setMeta "date" (str "today")
- $ doc
+ =?>
+ setMeta "date" (str "today") (doc
$ definitionList [ (emph "one", [para "emphasis"])
, (link "http://example.com" "" "two", [para "reference"])
, (link "http://example.org" "" "three", [para "another one"])
@@ -102,13 +102,12 @@ tests = [ "line block with blank line" =:
, " def func(x):"
, " return y"
] =?>
- ( doc $ codeBlockWith
+ doc (codeBlockWith
( ""
, ["sourceCode", "python", "numberLines", "class1", "class2", "class3"]
, [ ("startFrom", "34") ]
)
- "def func(x):\n return y"
- )
+ "def func(x):\n return y")
, "Code directive with number-lines, no line specified" =: T.unlines
[ ".. code::python"
, " :number-lines: "
@@ -116,13 +115,12 @@ tests = [ "line block with blank line" =:
, " def func(x):"
, " return y"
] =?>
- ( doc $ codeBlockWith
+ doc (codeBlockWith
( ""
, ["sourceCode", "python", "numberLines"]
, [ ("startFrom", "") ]
)
- "def func(x):\n return y"
- )
+ "def func(x):\n return y")
, testGroup "literal / line / code blocks"
[ "indented literal block" =: T.unlines
[ "::"
@@ -131,7 +129,8 @@ tests = [ "line block with blank line" =:
, ""
, " can go on for many lines"
, "but must stop here"]
- =?> (doc $
+ =?>
+ doc (
codeBlock "block quotes\n\ncan go on for many lines" <>
para "but must stop here")
, "line block with 3 lines" =: "| a\n| b\n| c"
@@ -185,6 +184,6 @@ tests = [ "line block with blank line" =:
, ".. [1]"
, " bar"
] =?>
- (para $ "foo" <> (note $ para "bar"))
+ para ("foo" <> (note $ para "bar"))
]
]
diff --git a/test/Tests/Readers/Txt2Tags.hs b/test/Tests/Readers/Txt2Tags.hs
index 28f647de4..9c5053af7 100644
--- a/test/Tests/Readers/Txt2Tags.hs
+++ b/test/Tests/Readers/Txt2Tags.hs
@@ -30,11 +30,11 @@ simpleTable' :: Int
-> [Blocks]
-> [[Blocks]]
-> Blocks
-simpleTable' n = table "" (take n $ repeat (AlignCenter, 0.0))
+simpleTable' n = table "" (replicate n (AlignCenter, 0.0))
tests :: [TestTree]
tests =
- [ testGroup "Inlines" $
+ [ testGroup "Inlines"
[ "Plain String" =:
"Hello, World" =?>
para (spcSep [ "Hello,", "World" ])
@@ -114,7 +114,7 @@ tests =
]
- , testGroup "Basic Blocks" $
+ , testGroup "Basic Blocks"
["Paragraph, lines grouped together" =:
"A paragraph\n A blank line ends the \n current paragraph\n"
=?> para "A paragraph\n A blank line ends the\n current paragraph"
@@ -197,7 +197,7 @@ tests =
]
- , testGroup "Lists" $
+ , testGroup "Lists"
[ "Simple Bullet Lists" =:
("- Item1\n" <>
"- Item2\n") =?>
diff --git a/test/Tests/Shared.hs b/test/Tests/Shared.hs
index 6fdbda3dd..cc448419c 100644
--- a/test/Tests/Shared.hs
+++ b/test/Tests/Shared.hs
@@ -19,21 +19,21 @@ tests = [ testGroup "compactifyDL"
testCollapse :: [TestTree]
testCollapse = map (testCase "collapse")
- [ (collapseFilePath (joinPath [ ""]) @?= (joinPath [ ""]))
- , (collapseFilePath (joinPath [ ".","foo"]) @?= (joinPath [ "foo"]))
- , (collapseFilePath (joinPath [ ".",".","..","foo"]) @?= (joinPath [ joinPath ["..", "foo"]]))
- , (collapseFilePath (joinPath [ "..","foo"]) @?= (joinPath [ "..","foo"]))
- , (collapseFilePath (joinPath [ "","bar","..","baz"]) @?= (joinPath [ "","baz"]))
- , (collapseFilePath (joinPath [ "","..","baz"]) @?= (joinPath [ "","..","baz"]))
- , (collapseFilePath (joinPath [ ".","foo","..",".","bar","..",".",".","baz"]) @?= (joinPath [ "baz"]))
- , (collapseFilePath (joinPath [ ".",""]) @?= (joinPath [ ""]))
- , (collapseFilePath (joinPath [ ".",".",""]) @?= (joinPath [ ""]))
- , (collapseFilePath (joinPath [ "..",""]) @?= (joinPath [ ".."]))
- , (collapseFilePath (joinPath [ "..",".",""]) @?= (joinPath [ ".."]))
- , (collapseFilePath (joinPath [ ".","..",""]) @?= (joinPath [ ".."]))
- , (collapseFilePath (joinPath [ "..","..",""]) @?= (joinPath [ "..",".."]))
- , (collapseFilePath (joinPath [ "parent","foo","baz","..","bar"]) @?= (joinPath [ "parent","foo","bar"]))
- , (collapseFilePath (joinPath [ "parent","foo","baz","..","..","bar"]) @?= (joinPath [ "parent","bar"]))
- , (collapseFilePath (joinPath [ "parent","foo",".."]) @?= (joinPath [ "parent"]))
- , (collapseFilePath (joinPath [ "","parent","foo","..","..","bar"]) @?= (joinPath [ "","bar"]))
- , (collapseFilePath (joinPath [ "",".","parent","foo"]) @?= (joinPath [ "","parent","foo"]))]
+ [ collapseFilePath (joinPath [ ""]) @?= (joinPath [ ""])
+ , collapseFilePath (joinPath [ ".","foo"]) @?= (joinPath [ "foo"])
+ , collapseFilePath (joinPath [ ".",".","..","foo"]) @?= (joinPath [ joinPath ["..", "foo"]])
+ , collapseFilePath (joinPath [ "..","foo"]) @?= (joinPath [ "..","foo"])
+ , collapseFilePath (joinPath [ "","bar","..","baz"]) @?= (joinPath [ "","baz"])
+ , collapseFilePath (joinPath [ "","..","baz"]) @?= (joinPath [ "","..","baz"])
+ , collapseFilePath (joinPath [ ".","foo","..",".","bar","..",".",".","baz"]) @?= (joinPath [ "baz"])
+ , collapseFilePath (joinPath [ ".",""]) @?= (joinPath [ ""])
+ , collapseFilePath (joinPath [ ".",".",""]) @?= (joinPath [ ""])
+ , collapseFilePath (joinPath [ "..",""]) @?= (joinPath [ ".."])
+ , collapseFilePath (joinPath [ "..",".",""]) @?= (joinPath [ ".."])
+ , collapseFilePath (joinPath [ ".","..",""]) @?= (joinPath [ ".."])
+ , collapseFilePath (joinPath [ "..","..",""]) @?= (joinPath [ "..",".."])
+ , collapseFilePath (joinPath [ "parent","foo","baz","..","bar"]) @?= (joinPath [ "parent","foo","bar"])
+ , collapseFilePath (joinPath [ "parent","foo","baz","..","..","bar"]) @?= (joinPath [ "parent","bar"])
+ , collapseFilePath (joinPath [ "parent","foo",".."]) @?= (joinPath [ "parent"])
+ , collapseFilePath (joinPath [ "","parent","foo","..","..","bar"]) @?= (joinPath [ "","bar"])
+ , collapseFilePath (joinPath [ "",".","parent","foo"]) @?= (joinPath [ "","parent","foo"])]
diff --git a/test/Tests/Writers/ConTeXt.hs b/test/Tests/Writers/ConTeXt.hs
index 7145240e3..5ce629674 100644
--- a/test/Tests/Writers/ConTeXt.hs
+++ b/test/Tests/Writers/ConTeXt.hs
@@ -41,9 +41,9 @@ tests = [ testGroup "inline code"
, "without '}'" =: code "]" =?> "\\type{]}"
, testProperty "code property" $ \s -> null s ||
if '{' `elem` s || '}' `elem` s
- then (context' $ code s) == "\\mono{" ++
- (context' $ str s) ++ "}"
- else (context' $ code s) == "\\type{" ++ s ++ "}"
+ then context' (code s) == "\\mono{" ++
+ context' (str s) ++ "}"
+ else context' (code s) == "\\type{" ++ s ++ "}"
]
, testGroup "headers"
[ "level 1" =:
@@ -124,4 +124,3 @@ tests = [ testGroup "inline code"
, "\\stopplacetable" ]
]
]
-
diff --git a/test/Tests/Writers/Docbook.hs b/test/Tests/Writers/Docbook.hs
index 90ae073fa..89ea76586 100644
--- a/test/Tests/Writers/Docbook.hs
+++ b/test/Tests/Writers/Docbook.hs
@@ -230,7 +230,7 @@ tests = [ testGroup "line blocks"
]
]
]
- , testGroup "writer options" $
+ , testGroup "writer options"
[ testGroup "top-level division" $
let
headers = header 1 (text "header1")
diff --git a/test/Tests/Writers/FB2.hs b/test/Tests/Writers/FB2.hs
index b4d11abf4..6663c42f8 100644
--- a/test/Tests/Writers/FB2.hs
+++ b/test/Tests/Writers/FB2.hs
@@ -23,7 +23,7 @@ tests = [ testGroup "block elements"
]
, testGroup "inlines"
[
- "Emphasis" =: emph ("emphasized")
+ "Emphasis" =: emph "emphasized"
=?> fb2 "<emphasis>emphasized</emphasis>"
]
, "bullet list" =: bulletList [ plain $ text "first"
diff --git a/test/Tests/Writers/JATS.hs b/test/Tests/Writers/JATS.hs
index f14f1c229..572b16451 100644
--- a/test/Tests/Writers/JATS.hs
+++ b/test/Tests/Writers/JATS.hs
@@ -120,5 +120,3 @@ tests = [ testGroup "inline code"
\</sec>"
]
]
-
-
diff --git a/test/Tests/Writers/Markdown.hs b/test/Tests/Writers/Markdown.hs
index 012e0888c..7f9ac3627 100644
--- a/test/Tests/Writers/Markdown.hs
+++ b/test/Tests/Writers/Markdown.hs
@@ -80,7 +80,7 @@ noteTestDoc =
".") <>
blockQuote (para ("A note inside a block quote." <>
note (para "The second note.")) <>
- para ("A second paragraph.")) <>
+ para "A second paragraph.") <>
header 1 "Second Header" <>
para "Some more text."
@@ -91,7 +91,7 @@ noteTests = testGroup "note and reference location"
[ test (markdownWithOpts defopts)
"footnotes at the end of a document" $
noteTestDoc =?>
- (unlines $ [ "First Header"
+ (unlines [ "First Header"
, "============"
, ""
, "This is a footnote.[^1] And this is a [link](https://www.google.com)."
@@ -112,7 +112,7 @@ noteTests = testGroup "note and reference location"
, test (markdownWithOpts defopts{writerReferenceLocation=EndOfBlock})
"footnotes at the end of blocks" $
noteTestDoc =?>
- (unlines $ [ "First Header"
+ (unlines [ "First Header"
, "============"
, ""
, "This is a footnote.[^1] And this is a [link](https://www.google.com)."
@@ -133,7 +133,7 @@ noteTests = testGroup "note and reference location"
, test (markdownWithOpts defopts{writerReferenceLocation=EndOfBlock, writerReferenceLinks=True})
"footnotes and reference links at the end of blocks" $
noteTestDoc =?>
- (unlines $ [ "First Header"
+ (unlines [ "First Header"
, "============"
, ""
, "This is a footnote.[^1] And this is a [link]."
@@ -156,7 +156,7 @@ noteTests = testGroup "note and reference location"
, test (markdownWithOpts defopts{writerReferenceLocation=EndOfSection})
"footnotes at the end of section" $
noteTestDoc =?>
- (unlines $ [ "First Header"
+ (unlines [ "First Header"
, "============"
, ""
, "This is a footnote.[^1] And this is a [link](https://www.google.com)."
@@ -186,27 +186,27 @@ shortcutLinkRefsTests =
(=:) = test (purely (writeMarkdown defopts{writerReferenceLinks = True}) . toPandoc)
in testGroup "Shortcut reference links"
[ "Simple link (shortcutable)"
- =: (para (link "/url" "title" "foo"))
+ =: para (link "/url" "title" "foo")
=?> "[foo]\n\n [foo]: /url \"title\""
, "Followed by another link (unshortcutable)"
- =: (para ((link "/url1" "title1" "first")
- <> (link "/url2" "title2" "second")))
+ =: para ((link "/url1" "title1" "first")
+ <> (link "/url2" "title2" "second"))
=?> unlines [ "[first][][second]"
, ""
, " [first]: /url1 \"title1\""
, " [second]: /url2 \"title2\""
]
, "Followed by space and another link (unshortcutable)"
- =: (para ((link "/url1" "title1" "first") <> " "
- <> (link "/url2" "title2" "second")))
+ =: para ((link "/url1" "title1" "first") <> " "
+ <> (link "/url2" "title2" "second"))
=?> unlines [ "[first][] [second]"
, ""
, " [first]: /url1 \"title1\""
, " [second]: /url2 \"title2\""
]
, "Reference link is used multiple times (unshortcutable)"
- =: (para ((link "/url1" "" "foo") <> (link "/url2" "" "foo")
- <> (link "/url3" "" "foo")))
+ =: para ((link "/url1" "" "foo") <> (link "/url2" "" "foo")
+ <> (link "/url3" "" "foo"))
=?> unlines [ "[foo][][foo][1][foo][2]"
, ""
, " [foo]: /url1"
@@ -214,8 +214,8 @@ shortcutLinkRefsTests =
, " [2]: /url3"
]
, "Reference link is used multiple times (unshortcutable)"
- =: (para ((link "/url1" "" "foo") <> " " <> (link "/url2" "" "foo")
- <> " " <> (link "/url3" "" "foo")))
+ =: para ((link "/url1" "" "foo") <> " " <> (link "/url2" "" "foo")
+ <> " " <> (link "/url3" "" "foo"))
=?> unlines [ "[foo][] [foo][1] [foo][2]"
, ""
, " [foo]: /url1"
@@ -223,43 +223,43 @@ shortcutLinkRefsTests =
, " [2]: /url3"
]
, "Reference link is followed by text in brackets"
- =: (para ((link "/url" "" "link") <> "[text in brackets]"))
+ =: para ((link "/url" "" "link") <> "[text in brackets]")
=?> unlines [ "[link][]\\[text in brackets\\]"
, ""
, " [link]: /url"
]
, "Reference link is followed by space and text in brackets"
- =: (para ((link "/url" "" "link") <> " [text in brackets]"))
+ =: para ((link "/url" "" "link") <> " [text in brackets]")
=?> unlines [ "[link][] \\[text in brackets\\]"
, ""
, " [link]: /url"
]
, "Reference link is followed by RawInline"
- =: (para ((link "/url" "" "link") <> rawInline "markdown" "[rawText]"))
+ =: para ((link "/url" "" "link") <> rawInline "markdown" "[rawText]")
=?> unlines [ "[link][][rawText]"
, ""
, " [link]: /url"
]
, "Reference link is followed by space and RawInline"
- =: (para ((link "/url" "" "link") <> space <> rawInline "markdown" "[rawText]"))
+ =: para ((link "/url" "" "link") <> space <> rawInline "markdown" "[rawText]")
=?> unlines [ "[link][] [rawText]"
, ""
, " [link]: /url"
]
, "Reference link is followed by RawInline with space"
- =: (para ((link "/url" "" "link") <> rawInline "markdown" " [rawText]"))
+ =: para ((link "/url" "" "link") <> rawInline "markdown" " [rawText]")
=?> unlines [ "[link][] [rawText]"
, ""
, " [link]: /url"
]
, "Reference link is followed by citation"
- =: (para ((link "/url" "" "link") <> cite [Citation "author" [] [] NormalCitation 0 0] (str "[@author]")))
+ =: para ((link "/url" "" "link") <> cite [Citation "author" [] [] NormalCitation 0 0] (str "[@author]"))
=?> unlines [ "[link][][@author]"
, ""
, " [link]: /url"
]
, "Reference link is followed by space and citation"
- =: (para ((link "/url" "" "link") <> space <> cite [Citation "author" [] [] NormalCitation 0 0] (str "[@author]")))
+ =: para ((link "/url" "" "link") <> space <> cite [Citation "author" [] [] NormalCitation 0 0] (str "[@author]"))
=?> unlines [ "[link][] [@author]"
, ""
, " [link]: /url"
diff --git a/test/Tests/Writers/Native.hs b/test/Tests/Writers/Native.hs
index c22185968..0c4bf7623 100644
--- a/test/Tests/Writers/Native.hs
+++ b/test/Tests/Writers/Native.hs
@@ -18,5 +18,5 @@ p_write_blocks_rt bs =
tests :: [TestTree]
tests = [ testProperty "p_write_rt" p_write_rt
, testProperty "p_write_blocks_rt" $ mapSize
- (\x -> if x > 3 then 3 else x) $ p_write_blocks_rt
+ (\x -> if x > 3 then 3 else x) p_write_blocks_rt
]
diff --git a/test/Tests/Writers/Powerpoint.hs b/test/Tests/Writers/Powerpoint.hs
index cc94f822d..e179742ed 100644
--- a/test/Tests/Writers/Powerpoint.hs
+++ b/test/Tests/Writers/Powerpoint.hs
@@ -72,7 +72,7 @@ numSlideTests = testGroup "Number of slides in output"
def
(doc $
para "first slide" <>
- (para $ image "lalune.jpg" "" "") <>
+ para (image "lalune.jpg" "" "") <>
para "foo")
, testNumberOfSlides
"With image slide, header" 3
@@ -80,14 +80,14 @@ numSlideTests = testGroup "Number of slides in output"
(doc $
para "first slide" <>
header 2 "image header" <>
- (para $ image "lalune.jpg" "" "") <>
+ para (image "lalune.jpg" "" "") <>
para "foo")
, testNumberOfSlides
"With table, no header" 3
def
(doc $
para "first slide" <>
- (simpleTable [para "foo" <> para "bar"] [[para "this" <> para "that"]]) <>
+ simpleTable [para "foo" <> para "bar"] [[para "this" <> para "that"]] <>
para "foo")
, testNumberOfSlides
"With table, header" 3
@@ -95,7 +95,7 @@ numSlideTests = testGroup "Number of slides in output"
(doc $
para "first slide" <>
header 2 "table header" <>
- (simpleTable [para "foo" <> para "bar"] [[para "this" <> para "that"]]) <>
+ simpleTable [para "foo" <> para "bar"] [[para "this" <> para "that"]] <>
para "foo")
, testNumberOfSlides
"hrule" 2
@@ -117,7 +117,7 @@ contentTypesFileExists opts pd =
testCase "Existence of [Content_Types].xml file" $
do archive <- getPptxArchive opts pd
assertBool "Missing [Content_Types].xml file" $
- "[Content_Types].xml" `elem` (filesInArchive archive)
+ "[Content_Types].xml" `elem` filesInArchive archive
@@ -138,7 +138,7 @@ prop_ContentOverrides pd = do
Nothing -> throwIO $
PandocSomeError "Missing [Content_Types].xml file"
typesElem <- case parseXMLDoc contentTypes of
- Just element -> return $ element
+ Just element -> return element
Nothing -> throwIO $
PandocSomeError "[Content_Types].xml cannot be parsed"
let ns = findAttr (QName "xmlns" Nothing Nothing) typesElem
diff --git a/test/Tests/Writers/TEI.hs b/test/Tests/Writers/TEI.hs
index f0a034bbd..fa372909f 100644
--- a/test/Tests/Writers/TEI.hs
+++ b/test/Tests/Writers/TEI.hs
@@ -31,7 +31,7 @@ tests = [ testGroup "block elements"
]
, testGroup "inlines"
[
- "Emphasis" =: emph ("emphasized")
+ "Emphasis" =: emph "emphasized"
=?> "<p><hi rendition=\"simple:italic\">emphasized</hi></p>"
,"SingleQuoted" =: singleQuoted (text "quoted material")
=?> "<p><quote>quoted material</quote></p>"