aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
-rw-r--r--src/Text/Pandoc/Shared.hs79
1 files changed, 40 insertions, 39 deletions
diff --git a/src/Text/Pandoc/Shared.hs b/src/Text/Pandoc/Shared.hs
index c2228ffff..12a2fa102 100644
--- a/src/Text/Pandoc/Shared.hs
+++ b/src/Text/Pandoc/Shared.hs
@@ -99,10 +99,10 @@ module Text.Pandoc.Shared (
import Text.Pandoc.Definition
import Text.ParserCombinators.Parsec
-import Text.PrettyPrint.HughesPJ ( Doc (..), fsep )
+import Text.PrettyPrint.HughesPJ ( Doc, fsep )
import Text.Pandoc.CharacterReferences ( characterReference )
-import Data.Char ( toLower, toUpper, ord, chr, isLower, isUpper )
-import Data.List ( find, groupBy, isPrefixOf, isSuffixOf )
+import Data.Char ( toLower, toUpper, ord, isLower, isUpper )
+import Data.List ( find, isPrefixOf )
import Control.Monad ( join )
--
@@ -137,7 +137,7 @@ substitute target replacement lst =
joinWithSep :: [a] -- ^ List to use as separator
-> [[a]] -- ^ Lists to join
-> [a]
-joinWithSep sep [] = []
+joinWithSep _ [] = []
joinWithSep sep lst = foldr1 (\a b -> a ++ sep ++ b) lst
--
@@ -153,7 +153,7 @@ backslashEscapes = map (\ch -> (ch, ['\\',ch]))
-- | Escape a string of characters, using an association list of
-- characters and strings.
escapeStringUsing :: [(Char, String)] -> String -> String
-escapeStringUsing escapeTable [] = ""
+escapeStringUsing _ [] = ""
escapeStringUsing escapeTable (x:xs) =
case (lookup x escapeTable) of
Just str -> str ++ rest
@@ -194,20 +194,20 @@ toRomanNumeral x =
if x >= 4000 || x < 0
then "?"
else case x of
- x | x >= 1000 -> "M" ++ toRomanNumeral (x - 1000)
- x | x >= 900 -> "CM" ++ toRomanNumeral (x - 900)
- x | x >= 500 -> "D" ++ toRomanNumeral (x - 500)
- x | x >= 400 -> "CD" ++ toRomanNumeral (x - 400)
- x | x >= 100 -> "C" ++ toRomanNumeral (x - 100)
- x | x >= 90 -> "XC" ++ toRomanNumeral (x - 90)
- x | x >= 50 -> "L" ++ toRomanNumeral (x - 50)
- x | x >= 40 -> "XL" ++ toRomanNumeral (x - 40)
- x | x >= 10 -> "X" ++ toRomanNumeral (x - 10)
- x | x >= 9 -> "IX" ++ toRomanNumeral (x - 5)
- x | x >= 5 -> "V" ++ toRomanNumeral (x - 5)
- x | x >= 4 -> "IV" ++ toRomanNumeral (x - 4)
- x | x >= 1 -> "I" ++ toRomanNumeral (x - 1)
- 0 -> ""
+ _ | x >= 1000 -> "M" ++ toRomanNumeral (x - 1000)
+ _ | x >= 900 -> "CM" ++ toRomanNumeral (x - 900)
+ _ | x >= 500 -> "D" ++ toRomanNumeral (x - 500)
+ _ | x >= 400 -> "CD" ++ toRomanNumeral (x - 400)
+ _ | x >= 100 -> "C" ++ toRomanNumeral (x - 100)
+ _ | x >= 90 -> "XC" ++ toRomanNumeral (x - 90)
+ _ | x >= 50 -> "L" ++ toRomanNumeral (x - 50)
+ _ | x >= 40 -> "XL" ++ toRomanNumeral (x - 40)
+ _ | x >= 10 -> "X" ++ toRomanNumeral (x - 10)
+ _ | x >= 9 -> "IX" ++ toRomanNumeral (x - 5)
+ _ | x >= 5 -> "V" ++ toRomanNumeral (x - 5)
+ _ | x >= 4 -> "IV" ++ toRomanNumeral (x - 4)
+ _ | x >= 1 -> "I" ++ toRomanNumeral (x - 1)
+ _ -> ""
-- | Wrap inlines to line length.
wrapped :: Monad m => ([Inline] -> m Doc) -> [Inline] -> m Doc
@@ -294,9 +294,9 @@ parseFromString parser str = do
-- | Parse raw line block up to and including blank lines.
lineClump :: GenParser Char st String
lineClump = do
- lines <- many1 (notFollowedBy blankline >> anyLine)
+ lns <- many1 (notFollowedBy blankline >> anyLine)
blanks <- blanklines <|> (eof >> return "\n")
- return $ (unlines lines) ++ blanks
+ return $ (unlines lns) ++ blanks
-- | Parse a string of characters between an open character
-- and a close character, including text between balanced
@@ -327,8 +327,8 @@ charsInBalanced' open close = try $ do
-- | Parses a roman numeral (uppercase or lowercase), returns number.
romanNumeral :: Bool -- ^ Uppercase if true
-> GenParser Char st Int
-romanNumeral upper = do
- let charAnyCase c = char (if upper then toUpper c else c)
+romanNumeral upperCase = do
+ let charAnyCase c = char (if upperCase then toUpper c else c)
let one = charAnyCase 'i'
let five = charAnyCase 'v'
let ten = charAnyCase 'x'
@@ -481,7 +481,7 @@ orderedListMarker style delim = do
Period -> inPeriod
OneParen -> inOneParen
TwoParens -> inTwoParens
- (start, style, delim) <- context num
+ (start, _, _) <- context num
return start
-- | Parses a character reference and returns a Str element.
@@ -626,7 +626,7 @@ indentBy :: Int -- ^ Number of spaces to indent the block
-> Int -- ^ Number of spaces (rel to block) to indent first line
-> String -- ^ Contents of block to indent
-> String
-indentBy num first [] = ""
+indentBy _ _ [] = ""
indentBy num first str =
let (firstLine:restLines) = lines str
firstLineIndent = num + first
@@ -679,21 +679,21 @@ prettyPandoc (Pandoc meta blocks) = "Pandoc " ++ "(" ++ show meta ++
orderedListMarkers :: (Int, ListNumberStyle, ListNumberDelim) -> [String]
orderedListMarkers (start, numstyle, numdelim) =
let singleton c = [c]
- seq = case numstyle of
- DefaultStyle -> map show [start..]
- Decimal -> map show [start..]
- UpperAlpha -> drop (start - 1) $ cycle $
- map singleton ['A'..'Z']
- LowerAlpha -> drop (start - 1) $ cycle $
- map singleton ['a'..'z']
- UpperRoman -> map toRomanNumeral [start..]
- LowerRoman -> map (map toLower . toRomanNumeral) [start..]
+ nums = case numstyle of
+ DefaultStyle -> map show [start..]
+ Decimal -> map show [start..]
+ UpperAlpha -> drop (start - 1) $ cycle $
+ map singleton ['A'..'Z']
+ LowerAlpha -> drop (start - 1) $ cycle $
+ map singleton ['a'..'z']
+ UpperRoman -> map toRomanNumeral [start..]
+ LowerRoman -> map (map toLower . toRomanNumeral) [start..]
inDelim str = case numdelim of
DefaultDelim -> str ++ "."
Period -> str ++ "."
OneParen -> str ++ ")"
TwoParens -> "(" ++ str ++ ")"
- in map inDelim seq
+ in map inDelim nums
-- | Normalize a list of inline elements: remove leading and trailing
-- @Space@ elements, collapse double @Space@s into singles, and
@@ -726,18 +726,18 @@ compactify items =
[Para a] -> if any containsPara others
then items
else others ++ [[Plain a]]
- otherwise -> items
+ _ -> items
containsPara :: [Block] -> Bool
containsPara [] = False
-containsPara ((Para a):rest) = True
+containsPara ((Para _):_) = True
containsPara ((BulletList items):rest) = any containsPara items ||
containsPara rest
containsPara ((OrderedList _ items):rest) = any containsPara items ||
containsPara rest
containsPara ((DefinitionList items):rest) = any containsPara (map snd items) ||
containsPara rest
-containsPara (x:rest) = containsPara rest
+containsPara (_:rest) = containsPara rest
-- | Data structure for defining hierarchical Pandoc documents
data Element = Blk Block
@@ -746,7 +746,7 @@ data Element = Blk Block
-- | Returns @True@ on Header block with at least the specified level
headerAtLeast :: Int -> Block -> Bool
headerAtLeast level (Header x _) = x <= level
-headerAtLeast level _ = False
+headerAtLeast _ _ = False
-- | Convert list of Pandoc blocks into (hierarchical) list of Elements
hierarchicalize :: [Block] -> [Element]
@@ -787,6 +787,7 @@ data WriterOptions = WriterOptions
} deriving Show
-- | Default writer options.
+defaultWriterOptions :: WriterOptions
defaultWriterOptions =
WriterOptions { writerStandalone = False,
writerHeader = "",