module MicroCabal.Parse( parseCabal, parseYAML, parseSnapshots, ) where import Control.Applicative import Control.Monad import Data.Char import Data.List import Data.Maybe import Text.ParserComb import MicroCabal.Cabal import MicroCabal.YAML --import Debug.Trace parseCabal :: FilePath -> String -> Cabal parseCabal fn rfile = runP pCabalTop fn $ dropCabalComments rfile parseYAML :: FilePath -> String -> YAMLValue parseYAML fn rfile = runP pYAMLTop fn $ dropYAMLComments rfile parseSnapshots :: FilePath -> String -> [(String, String)] parseSnapshots fn rfile = runP pSnapshotsTop fn rfile runP :: P a -> FilePath -> String -> a runP prsr fn file = case runPrsr prsr (initLexState file) of Left (LastFail n ts msgs) -> error $ "\n" ++ " found: " ++ (map show ts ++ ["EOF"]) !! 0 ++ "\n" ++ " expected: " ++ unwords (nub msgs) ++ "\n" ++ -- " n=" ++ show n ++ "\n" ++ " file: " ++ show fn ++ "\n" ++ " line: " ++ show (1 + length (filter (== '\n') (take (length file - n) file))) ++ "\n" ++ " at: " ++ show (take 100 (drop (length file - n) file)) Right (a:_) -> a Right [] -> undefined -- impossible ------------------------------ type P a = Prsr LexState Char a data LexState = LS Int [Int] [Char] deriving (Show) initLexState :: [Char] -> LexState initLexState cs = LS 0 [] cs end :: Char end = '\EOT' fieldSep :: Char fieldSep = '\FS' instance TokenMachine LexState Char where -- tmNextToken ls | trace ("tmNextToken: " ++ show ls) False = undefined tmNextToken ls@(LS _ [] []) = (end, ls) -- ugly hack tmNextToken (LS i (_:ks) []) = (fieldSep, LS i ks []) tmNextToken (LS i [] (c:cs)) | c == '\n' = (c, LS 0 [] cs) | otherwise = (c, LS (i+1) [] cs) tmNextToken (LS i kks@(k:ks) (c:cs)) | c /= '\n' = (c, LS (i+1) kks cs) | Just cs' <- skipEmpty cs = tmNextToken (LS i kks cs') | otherwise = let lead 0 _ = -- trace ("Just NL kks=" ++ show kks ++ " cs=" ++ show (take 10 cs)) ('\n', LS 0 kks cs) -- There are at least k leading spaces lead j (x:xs) | x == ' ' = lead (j-1) xs -- Count spaces -- | x == '-' = ('\n', LS 0 kks cs) lead _ _ = -- trace ("insert FS kks=" ++ show kks ++ " cs=" ++ show (take 10 cs)) (fieldSep, LS 0 ks (c:cs)) -- Fewer than k spaces. Generate FS, pop, and try again. in lead (k+1) cs tmRawTokens (LS _ _ cs) = cs skipEmpty :: String -> Maybe String skipEmpty s = case dropWhile (== ' ') s of cs@('\n':_) -> Just cs _ -> Nothing pushColumn :: P () pushColumn = mapTokenState (\ (LS i ks cs) -> LS i (i:ks) cs) pushFieldSep :: P () pushFieldSep = mapTokenState (\ (LS i ks cs) -> LS i ks (fieldSep:cs)) lower :: String -> String lower = map toLower -- Change lines with first non-space being '--' into just a newline dropCabalComments :: String -> String dropCabalComments = unlines . map cmt . lines where cmt s | take 2 (dropWhile (== ' ') s) == "--" = "" | otherwise = s satisfySome :: String -> (Char -> Bool) -> P [Char] satisfySome msg p = (:) <$> satisfy msg p <*> satisfyMany p ---------------------------------------------------------------- pCabalTop :: P Cabal pCabalTop = pCabal <* pWhite <* pChar end pCabal :: P Cabal pCabal = Cabal <$> ((:) <$> (Section "global" "" <$> emany pField) <*> emany pSection) pColon :: P () pColon = pWhite <* pChar ':' pWhite :: P () pWhite = () <$ satisfyMany (\ c -> c == ' ' || c == '\n') pChar :: Char -> P () pChar c = () <$ satisfy (show c) (c ==) pFieldSep :: P () pFieldSep = pChar fieldSep pNewLine :: P () pNewLine = pChar '\n' pDot :: P () pDot = pChar '.' pSpaces :: P () pSpaces = () <$ satisfyMany (== ' ') pIdent :: P String pIdent = do c <- satisfy "ident" isAlpha_ cs <- satisfyMany isIdent pure (c:cs) pKeyWordNC :: String -> P String pKeyWordNC s = do pSpaces i <- pIdent guard (s == lower i) pure s isIdent :: Char -> Bool isIdent c = isAlphaNum c || c == '-' || c == '_' || c == '\'' || c == '.' isAlpha_ :: Char -> Bool isAlpha_ c = isAlpha c || c == '_' pNumber :: P Int pNumber = read <$> satisfySome "0..9" isDigit pParens :: P a -> P a pParens p = pStr "(" *> p <* pStr ")" pVersion :: P Version pVersion = pSpaces *> (makeVersion <$> esepBy1 pNumber pDot) pVersionRange :: P VersionRange pVersionRange = pVOr where pVOr = foldr1 VOr <$> esepBy1 pVAnd (pStr "&&") pVAnd = foldr1 VAnd <$> esepBy1 pVOp (pStr "||") pVOp = (pVOper <*> (pSpaces *> pVersion)) <|< pParens pVersionRange <|< (pStr "==" *> pVEq) <|< (pStr "^>=" *> pVGEHat) pVEq = (VEQSet <$> pVSet) <|< do v <- pVersion (VEQWild v <$ pStr ".*") <|< pure (VEQ v) pVSet = (pStr "{" *> pCommaList pVersion <* pStr "}") pVGEHat = (VGEHat <$> pVersion) <|< (VGEHatSet <$> pVSet) pVOper :: P (Version -> VersionRange) pVOper = pSpaces *> choice [ VGT <$ pStr ">", VGT <$ pStr "<", VGT <$ pStr "<=", VGT <$ pStr ">="] pStr :: String -> P () pStr s = pSpaces *> p s where p "" = pure () p (c:cs) = pChar c *> p cs pStrW :: String -> P () pStrW s = pWhite *> pStr s pItem :: P Item pItem = pWhite *> (pString <|< pWord) -- A string in quotation marks. pString :: P String pString = do pChar '"' let achar = satisfy "char" (\ c -> c /= '\n' && c /= end && c /= fieldSep) loop r = do c <- achar if c == '"' then return $ reverse r else if c /= '\\' then loop (c:r) else do e <- achar let e' = fromMaybe e $ lookup e [('n', '\n'), ('t', '\t')] -- could have more loop (e' : r) loop [] pWord :: P Item pWord = satisfySome "word" (\ c -> c `notElem` [' ', '\n', ',', end, fieldSep]) pComma :: P () pComma = pWhite *> pChar ',' pCommaList :: P a -> P [a] pCommaList p = (pStr "," *> esepBy1 p pComma) <|< pCommaList' p pCommaList' :: P a -> P [a] pCommaList' p = esepBy p pComma <* eoptional (pStr ",") pSpaceList :: P a -> P [a] pSpaceList p = esepBy p pWhite pOptCommaList :: P a -> P [a] pOptCommaList p = --pSpaceList p <|> pCommaList p (pStrW "," *> pCommaList' p) -- it starts with a ',', so it must be comma separated <|< do a <- p -- parse one item -- now check if we have a comma or not, and pick the parser for the rest as <- (pStrW "," *> pCommaList' p) <|< pSpaceList p return (a:as) pVComma :: P Value pVComma = VItems <$> pCommaList pItem pVSpace :: P Value pVSpace = VItems <$> pSpaceList pItem pVOptComma :: P Value pVOptComma = VItems <$> pOptCommaList pItem pVLibs :: P Value pVLibs = VPkgs <$> pCommaList pPkg pPkg :: P (Item, [Item], Maybe VersionRange) pPkg = (,,) <$> pNameW <*> (pSpaces *> pLibs) <*> optional pVersionRange where pLibs = do pColon ((:[]) <$> pNameW) <|< (pStr "{" *> pCommaList pName <* pStr "}") <|< pure [] pNameW = pWhite *> pIdent pField :: P Field pField = do pWhite pushColumn fn <- lower <$> pFieldName -- traceM ("pFieldName fn=" ++ show fn) if fn == "if" then do c <- pCond pNewLine t <- emany pField pFieldSep e <- do pWhite pushColumn _ <- pKeyWordNC "else" fs <- emany pField pFieldSep pure fs <|< pure [] pure $ If c t e else do pColon let p = getParser fn v <- p pFieldSep -- traceM ("pField v=" ++ show v) pure $ Field fn v pCond :: P Cond pCond = pCor where pCor = foldr1 Cor <$> esepBy1 pCand (pStr "&&") pCand = foldr1 Cand <$> esepBy1 pCnot (pStr "||") pCnot = (Cnot <$> (pStr "!" *> pCnot)) <|> pCOp pCOp = (CBool <$> pBool) <|< (pKeyWordNC "arch" *> pParens (Carch <$> pName)) <|< (pKeyWordNC "flag" *> pParens (Cflag <$> pName)) <|< (pKeyWordNC "impl" *> pParens (Cimpl <$> pName <*> optional pVersionRange)) <|< (pKeyWordNC "os" *> pParens (Cos <$> pName)) <|< pParens pCond pFreeText :: P Value pFreeText = do txt <- satisfyMany (\ c -> c /= end && c /= fieldSep) let dot "." = "" -- Single '.' used to make an empty line dot s = s pure $ VItem $ unlines . map (dot . dropWhile (== ' ')) . lines $ txt pFieldName :: P FieldName pFieldName = pIdent pName :: P Name pName = pSpaces *> pIdent pFields :: P [Field] pFields = pSpaces *> pNewLine *> emany pField pBool :: P Bool pBool = (False <$ pKeyWordNC "false") <|< (True <$ pKeyWordNC "true") pSection :: P Section pSection = pWhite *> ( Section <$> pKeyWordNC "common" <*> pName <*> pFields <|< Section <$> pKeyWordNC "library" <*> libName <*> pFields <|< Section <$> pKeyWordNC "executable" <*> pName <*> pFields <|< Section <$> pKeyWordNC "source-repository" <*> pName <*> pFields <|< Section <$> pKeyWordNC "flag" <*> pName <*> pFields <|< Section <$> pKeyWordNC "test-suite" <*> pName <*> pFields ) where libName = pName <|< pure "" getParser :: FieldName -> P Value getParser f = if "x-" `isPrefixOf` f then pFreeText else fromMaybe (error $ "Unknown field: " ++ f) $ lookup f parsers parsers :: [(FieldName, P Value)] parsers = [ "asm-options" # pVSpace , "asm-sources" # pVComma , "autogen-includes" # pVOptComma , "autogen-modules" # pVComma , "build-depends" # pVLibs , "build-tool-depends" # pVComma -- XXX , "build-tools" # pVComma -- XXX , "buildable" # (VBool <$> pBool) , "c-sources" # pVComma , "cc-options" # pVComma , "cmm-sources" # pVComma , "cmm-options" # pVComma , "cpp-options" # pVComma , "cxx-options" # pVComma , "default-extensions" # pVOptComma , "default-language" # (VItem <$> pItem) , "exposed-modules" # pVOptComma , "extensions" # pVOptComma , "extra-bundled-libraries" # pVComma , "extra-dynamic-library-flavours" # pVComma , "extra-framework-dirs" # pVComma , "extra-ghci-libraries" # pVComma , "extra-lib-dirs" # pVComma , "extra-lib-dirs-static" # pVComma , "extra-libraries" # pVComma , "frameworks" # pVOptComma , "ghc-options" # pVSpace , "ghc-prof-options" # pVSpace , "ghc-shared-options" # pVSpace , "ghcjs-options" # pVSpace , "ghcjs-prof-options" # pVSpace , "ghcjs-shared-options" # pVSpace , "hs-source-dirs" # pVOptComma , "import" # (VItem <$> pItem) , "include-dirs" # pVOptComma , "includes" # pVOptComma , "install-includes" # pVOptComma , "js-sources" # pVComma , "ld-options" # pVSpace , "mixins" # pFreeText -- XXX , "nhc98-options" # pVSpace , "other-extensions" # pVOptComma , "other-languages" # (VItem <$> pItem) , "other-modules" # pVOptComma , "pkg-config-depends" # pVComma , "virtual-modules" # pVComma --- library fields , "visibility" # (VItem <$> pItem) --- package fields , "author" # pFreeText , "bug-reports" # pFreeText , "build-type" # (VItem <$> pItem) , "cabal-version" # pFreeText -- (VRange <$> pVersionRange) , "category" # pFreeText , "copyright" # pFreeText , "data-dir" # pVSpace , "data-files" # pVComma , "description" # pFreeText , "extra-doc-files" # pVComma , "extra-source-files" # pVOptComma , "extra-tmp-files" # pVComma , "homepage" # pFreeText , "license" # pFreeText , "license-file" # pVOptComma , "license-files" # pVOptComma , "maintainer" # pFreeText , "name" # (VItem <$> pItem) , "package-url" # pFreeText , "stability" # pFreeText , "subdir" # pFreeText , "synopsis" # pFreeText , "tested-with" # pVOptComma , "version" # (VVersion <$> pVersion) -- test suite fields , "main-is" # (VItem <$> pItem) , "test-module" # (VItem <$> pItem) , "type" # (VItem <$> pItem) -- source-repository fields , "location" # pFreeText -- flag fields , "manual" # (VBool <$> pBool) , "default" # (VBool <$> pBool) , "tag" # pFreeText ] where (#) = (,) -- XXX use local fixity ---------------------------------------------------------------------- -- XXX Wrong for strings dropYAMLComments :: String -> String dropYAMLComments [] = [] dropYAMLComments (c:cs) | c == '#' = dropYAMLComments (dropWhile (/= '\n') cs) | otherwise = c : dropYAMLComments cs pYAMLTop :: P YAMLValue pYAMLTop = pYAMLRecord <* pWhite <* pChar end pYAMLValue :: P YAMLValue pYAMLValue = (YBool <$> pBool) <|< (YInt <$> pNumber) -- <|< (YString <$> pString) <|< pYAMLArray <|< pYAMLRecord <|< (YString <$> pYAMLFree) pYAMLArray :: P YAMLValue pYAMLArray = do pWhite let pElem = pChar '-' *> pSpaces *> pYAMLValue pElemFS = pWhite *> pElem <* pFieldSep pElemsFS = esome pElemFS pElemNL = pElem <* pChar '\n' pElemsNL = pFieldSep *> pChar '\n' *> esome pElemNL <* pushFieldSep YArray <$> (pElemsNL <|< pElemsFS) pYAMLRecord :: P YAMLValue pYAMLRecord = YRecord <$> esome pYAMLField pYAMLFree :: P String pYAMLFree = do pSpaces d <- nextToken guard (d /= '-') txt <- satisfyMany (\ c -> c /= end && c /= fieldSep && c /= '\n') pure txt pYAMLField :: P (YAMLFieldName, YAMLValue) pYAMLField = do pWhite pushColumn n <- pFieldName pColon v <- pYAMLValue pFieldSep pure (n, v) ---------------------------------------------------------------------- type Snapshot = (String, String) pSnapshotsTop :: P [Snapshot] pSnapshotsTop = pSnapshots <* pWhite <* pChar end pSnapshots :: P [Snapshot] pSnapshots = pWhite *> pChar '{' *> (esepBy pSnapshot (pWhite *> pChar ',')) <* pWhite <* pChar '}' pSnapshot :: P Snapshot pSnapshot = (,) <$> (pWhite *> pString) <*> (pWhite *> pChar ':' *> pWhite *> pString)