diff --git a/src/Text/Show/Unicode.hs b/src/Text/Show/Unicode.hs --- a/src/Text/Show/Unicode.hs +++ b/src/Text/Show/Unicode.hs @@ -50,9 +50,10 @@ module Text.Show.Unicode (ushow, uprint, ushowWith, uprintWith) where import Control.Applicative ((<|>)) -import Data.Char (isPrint) +import Data.Char (isAscii, isPrint) import Text.ParserCombinators.ReadP import Text.Read.Lex (lexChar) +import qualified Data.List as L -- Represents a replaced character using its literal form and its escaped form. type Replacement = (String, String) @@ -63,18 +64,38 @@ -- * Otherwise, leave the string as it was. -- * Note that special delimiter sequence "\&" may appear in a string. c.f. . recoverChar :: (Char -> Bool) -> ReadP Replacement -recoverChar p = (represent <$> gather lexChar) <|> (("\\&","\&") <$ string "\\&") +recoverChar p = represent <$> gather lexCharAndConsumeEmpties where represent :: (String, Char) -> Replacement represent (o,lc) - | p lc = (o, [lc]) + -- This is too dirty a hack. + -- However, I couldn't think of any other way to recover the & consumed by lexChar while not needlessly increasing the number of & by mis-detecting the escape sequence. + | p lc = + if head o /= '\\' && + "\\&" `L.isSuffixOf` o + then (o, lc : "\\&") + else (o, [lc]) | otherwise = (o, o) +-- | The base library lexChar has been handling & by itself since 4.9.1.0, +-- so consumeEmpties is a meaningless action, +-- but it makes sense for older versions of lexChar. +lexCharAndConsumeEmpties :: ReadP Char +lexCharAndConsumeEmpties = lexChar <* consumeEmpties + where + -- Consumes the string "\&" repeatedly and greedily (will only produce one match) + consumeEmpties :: ReadP () + consumeEmpties = do + rest <- look + case rest of + ('\\':'&':_) -> string "\\&" >> consumeEmpties + _ -> return () + -- | Show the input, and then replace Haskell character literals -- with the character it represents, for any Unicode printable characters except backslash, single and double quotation marks. -- If something fails, fallback to standard 'show'. ushow :: Show a => a -> String -ushow = ushowWith (\c -> isPrint c && not (c `elem` ['\\', '\'','\"'] )) +ushow = ushowWith (\c -> isPrint c && not (isAscii c)) -- | A version of 'print' that uses 'ushow'. uprint :: Show a => a -> IO () @@ -89,7 +110,7 @@ go _ [] = "" go _ (([],""):_) = "" go _ ((rs,""):_) = snd $ last rs - go _ ((_,o):[]) = o + go _ [(_,o)] = o go pr (([],_):rest) = go pr rest go _ ((rs,_):rest) = let r = last rs in snd r ++ go r rest diff --git a/test/Spec.hs b/test/Spec.hs --- a/test/Spec.hs +++ b/test/Spec.hs @@ -3,17 +3,31 @@ import Test.Hspec import Test.Hspec.QuickCheck +import Text.Read (readMaybe) import Text.Show.Unicode -data T試6験 = Å4 { すけろく :: String} deriving (Eq, Ord, Show, Read) -data T試7験 = String :@\& String deriving (Eq, Ord, Show, Read) -data T試8験 = String :＠\& String deriving (Eq, Ord, Show, Read) - +newtype T試6験 = Å4 { すけろく :: String } deriving (Eq, Ord, Show, Read) +data T試7験 = String :@\& String deriving (Eq, Ord, Show, Read) +data T試8験 = String :＠\& String deriving (Eq, Ord, Show, Read) +data T試9験 = String :\&＠\& String deriving (Eq, Ord, Show, Read) +data T試10験 = String :\&\& String deriving (Eq, Ord, Show, Read) ushowTo :: Show a => a -> String -> Spec ushowTo f t = it ("ushow " ++ show f ++ " == " ++ t) $ t `shouldBe` ushow f +-- | check `read . ushow == id` when `read . show == id`. +-- The reason why we don't test if the show fails is that older versions may fail to read the result of the show, +-- which cannot be handled at the library level, so we exclude it. +-- ==> is not used because it will cause an error if there is no test case that can be executed. +readUShowIsIdWhenOkPrelude :: (Eq a, Show a, Read a) => a -> Expectation +readUShowIsIdWhenOkPrelude v = + if preludeOk + then ushowOk + else pure () + where preludeOk = readMaybe (show v) == Just v + ushowOk = read (ushow v) `shouldBe` v + spec :: Spec spec = describe "individual representations test" $ do @@ -37,16 +51,22 @@ \x -> read (ushow x) `shouldBe` (x :: [(Char,())]) prop "read . read . ushow . ushow == id, for String" $ - \str -> (read $ read $ ushow $ ushow str) `shouldBe` (str :: String) + \str -> read (read $ ushow $ ushow str) `shouldBe` (str :: String) - prop "read . ushow == id, for some crazy Unicode type" $ - \str -> let v = Å4 str in read (ushow v) `shouldBe` v + prop "read . ushow == id, for some crazy Unicode type: T試6験" $ + \str -> readUShowIsIdWhenOkPrelude $ Å4 str - prop "read . ushow == id, for some crazy Unicode type" $ - \a b -> let v = a :@\& b in read (ushow v) `shouldBe` v + prop "read . ushow == id, for some crazy Unicode type: T試7験" $ + \a b -> readUShowIsIdWhenOkPrelude $ a :@\& b - prop "read . ushow == id, for some crazy Unicode type" $ - \a b -> let v = a :＠\& b in read (ushow v) `shouldBe` v + prop "read . ushow == id, for some crazy Unicode type: T試8験" $ + \a b -> readUShowIsIdWhenOkPrelude $ a :＠\& b + + prop "read . ushow == id, for some crazy Unicode type: T試9験" $ + \a b -> readUShowIsIdWhenOkPrelude $ a :\&＠\& b + + prop "read . ushow == id, for some crazy Unicode type: T試10験" $ + \a b -> readUShowIsIdWhenOkPrelude $ a :\&\& b prop "read . ushow == id, for compound type" $ \str -> read (ushow str) `shouldBe` (str :: Either [String] (String,String)) diff --git a/unicode-show.cabal b/unicode-show.cabal --- a/unicode-show.cabal +++ b/unicode-show.cabal @@ -1,5 +1,5 @@ name: unicode-show -version: 0.1.0.4 +version: 0.1.0.5 synopsis: print and show in unicode description: This package provides variants of 'show' and 'print' functions that does not escape non-ascii characters. @@ -25,7 +25,7 @@ library hs-source-dirs: src exposed-modules: Text.Show.Unicode - build-depends: base >= 4.7 && < 5 + build-depends: base >= 4.8.1.0 && < 5 default-language: Haskell2010 test-suite unicode-show-test