{-# LANGUAGE CPP #-}
{-# LANGUAGE PatternGuards #-}
{-# LANGUAGE DeriveFunctor #-}
{-# LANGUAGE TypeSynonymInstances #-}
{-# LANGUAGE MultiParamTypeClasses #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE GeneralizedNewtypeDeriving #-}
{-# LANGUAGE FlexibleContexts #-}
{-# LANGUAGE UndecidableInstances #-}
{-# LANGUAGE ScopedTypeVariables #-}
#if MIN_VERSION_base(4,8,0)
#else
{-# LANGUAGE OverlappingInstances #-}
#endif

{-
Copyright (C) 2016-17 Jesse Rosenthal <jrosenthal@jhu.edu>
and John MacFarlane.

This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
-}

{- |
   Module      : Text.Pandoc.Class
   Copyright   : Copyright (C) 2016-17 Jesse Rosenthal, John MacFarlane
   License     : GNU GPL, version 2 or above

   Maintainer  : Jesse Rosenthal <jrosenthal@jhu.edu>
   Stability   : alpha
   Portability : portable

This module defines a type class, 'PandocMonad', for pandoc readers
and writers. A pure instance 'PandocPure' and an impure instance
'PandocIO' are provided.  This allows users of the library to choose
whether they want conversions to perform IO operations (such as
reading include files or images).
-}

module Text.Pandoc.Class ( PandocMonad(..)
                         , CommonState(..)
                         , PureState(..)
                         , getPureState
                         , getsPureState
                         , putPureState
                         , modifyPureState
                         , getPOSIXTime
                         , getZonedTime
                         , readFileFromDirs
                         , report
                         , setTrace
                         , setRequestHeader
                         , getLog
                         , setVerbosity
                         , getVerbosity
                         , getMediaBag
                         , setMediaBag
                         , insertMedia
                         , setUserDataDir
                         , getUserDataDir
                         , fetchItem
                         , getInputFiles
                         , setInputFiles
                         , getOutputFile
                         , setOutputFile
                         , setResourcePath
                         , getResourcePath
                         , PandocIO(..)
                         , PandocPure(..)
                         , FileTree(..)
                         , FileInfo(..)
                         , addToFileTree
                         , runIO
                         , runIOorExplode
                         , runPure
                         , readDefaultDataFile
                         , readDataFile
                         , fetchMediaResource
                         , fillMediaBag
                         , extractMedia
                         , toLang
                         , setTranslations
                         , translateTerm
                         , Translations
                         ) where

import Prelude hiding (readFile)
import System.Random (StdGen, next, mkStdGen)
import qualified System.Random as IO (newStdGen)
import Codec.Archive.Zip
import qualified Data.CaseInsensitive as CI
import Data.Unique (hashUnique)
import Data.List (stripPrefix)
import qualified Data.Unique as IO (newUnique)
import qualified Text.Pandoc.UTF8 as UTF8
import qualified System.Directory as Directory
import Text.Pandoc.Compat.Time (UTCTime)
import Text.Pandoc.Logging
import Text.Parsec (ParsecT, getPosition, sourceLine, sourceName)
import qualified Text.Pandoc.Compat.Time as IO (getCurrentTime)
import Text.Pandoc.MIME (MimeType, getMimeType, extensionFromMimeType)
import Text.Pandoc.Definition
import Data.Digest.Pure.SHA (sha1, showDigest)
import Data.Maybe (fromMaybe)
import Data.Time.Clock.POSIX ( utcTimeToPOSIXSeconds
                             , posixSecondsToUTCTime
                             , POSIXTime )
import Data.Time.LocalTime (TimeZone, ZonedTime, utcToZonedTime, utc)
import Data.ByteString.Base64 (decodeLenient)
import Network.URI ( escapeURIString, nonStrictRelativeTo,
                     unEscapeString, parseURIReference, isAllowedInURI,
                     parseURI, URI(..) )
import Network.HTTP.Client
       (httpLbs, responseBody, responseHeaders,
        Request(port, host, requestHeaders), parseRequest, newManager)
import Network.HTTP.Client.Internal (addProxy)
import Network.HTTP.Client.TLS (tlsManagerSettings)
import System.Environment (getEnv)
import Network.HTTP.Types.Header ( hContentType )
import Network (withSocketsDo)
import Data.ByteString.Lazy (toChunks)
import qualified Control.Exception as E
import qualified Data.Time.LocalTime as IO (getCurrentTimeZone)
import Text.Pandoc.MediaBag (MediaBag, lookupMedia, mediaDirectory)
import Text.Pandoc.Walk (walkM, walk)
import qualified Text.Pandoc.MediaBag as MB
import qualified Data.ByteString as B
import qualified Data.ByteString.Lazy as BL
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 qualified System.FilePath.Glob as IO (glob)
import qualified System.Directory as IO (getModificationTime)
import Control.Monad as M (fail)
import Control.Monad.State.Strict
import Control.Monad.Except
import Data.Word (Word8)
import Data.Default
import System.IO.Error
import System.IO (stderr)
import qualified Data.Map as M
import Text.Pandoc.Error
import Text.Pandoc.BCP47 (Lang(..), parseBCP47, renderLang)
import Text.Pandoc.Translations (Term(..), Translations, lookupTerm,
                                 readTranslations)
import qualified Debug.Trace
#ifdef EMBED_DATA_FILES
import Text.Pandoc.Data (dataFiles)
import qualified System.FilePath.Posix as Posix
import System.FilePath (splitDirectories)
#else
import qualified Paths_pandoc as Paths
#endif

-- | The PandocMonad typeclass contains all the potentially
-- IO-related functions used in pandoc's readers and writers.
-- Instances of this typeclass may implement these functions
-- in IO (as in 'PandocIO') or using an internal state that
-- represents a file system, time, and so on (as in 'PandocPure').
class (Functor m, Applicative m, Monad m, MonadError PandocError m)
      => PandocMonad m where
  -- | Lookup an environment variable.
  lookupEnv :: String -> m (Maybe String)
  -- | Get the current (UTC) time.
  getCurrentTime :: m UTCTime
  -- | Get the locale's time zone.
  getCurrentTimeZone :: m TimeZone
  -- | Return a new generator for random numbers.
  newStdGen :: m StdGen
  -- | Return a new unique integer.
  newUniqueHash :: m Int
  -- | Retrieve contents and mime type from a URL, raising
  -- an error on failure.
  openURL :: String -> m (B.ByteString, Maybe MimeType)
  -- | Read the lazy ByteString contents from a file path,
  -- raising an error on failure.
  readFileLazy :: FilePath -> m BL.ByteString
  -- | Read the strict ByteString contents from a file path,
  -- raising an error on failure.
  readFileStrict :: FilePath -> m B.ByteString
  -- | Return a list of paths that match a glob, relative to
  -- the working directory.  See 'System.FilePath.Glob' for
  -- the glob syntax.
  glob :: String -> m [FilePath]
  -- | Returns True if file exists.
  fileExists :: FilePath -> m Bool
  -- | Returns the path of data file.
  getDataFileName :: FilePath -> m FilePath
  -- | Return the modification time of a file.
  getModificationTime :: FilePath -> m UTCTime
  -- | Get the value of the 'CommonState' used by all instances
  -- of 'PandocMonad'.
  getCommonState :: m CommonState
  -- | Set the value of the 'CommonState' used by all instances
  -- of 'PandocMonad'.
  -- | Get the value of a specific field of 'CommonState'.
  putCommonState :: CommonState -> m ()
  -- | Get the value of a specific field of 'CommonState'.
  getsCommonState :: (CommonState -> a) -> m a
  getsCommonState f = f <$> getCommonState
  -- | Modify the 'CommonState'.
  modifyCommonState :: (CommonState -> CommonState) -> m ()
  modifyCommonState f = getCommonState >>= putCommonState . f
  -- Output a log message.
  logOutput :: LogMessage -> m ()
  -- Output a debug message to sterr, using 'Debug.Trace.trace',
  -- if tracing is enabled.  Note: this writes to stderr even in
  -- pure instances.
  trace :: String -> m ()
  trace msg = do
    tracing <- getsCommonState stTrace
    when tracing $ Debug.Trace.trace ("[trace] " ++ msg) (return ())

-- * Functions defined for all PandocMonad instances

-- | Set the verbosity level.
setVerbosity :: PandocMonad m => Verbosity -> m ()
setVerbosity verbosity =
  modifyCommonState $ \st -> st{ stVerbosity = verbosity }

-- | Get the verbosity level.
getVerbosity :: PandocMonad m => m Verbosity
getVerbosity = getsCommonState stVerbosity

-- Get the accomulated log messages (in temporal order).
getLog :: PandocMonad m => m [LogMessage]
getLog = reverse <$> getsCommonState stLog

-- | Log a message using 'logOutput'.  Note that 'logOutput' is
-- called only if the verbosity level exceeds the level of the
-- message, but the message is added to the list of log messages
-- that will be retrieved by 'getLog' regardless of its verbosity level.
report :: PandocMonad m => LogMessage -> m ()
report msg = do
  verbosity <- getsCommonState stVerbosity
  let level = messageVerbosity msg
  when (level <= verbosity) $ logOutput msg
  modifyCommonState $ \st -> st{ stLog = msg : stLog st }

-- | Determine whether tracing is enabled.  This affects
-- the behavior of 'trace'.  If tracing is not enabled,
-- 'trace' does nothing.
setTrace :: PandocMonad m => Bool -> m ()
setTrace useTracing = modifyCommonState $ \st -> st{stTrace = useTracing}

-- | Set request header to use in HTTP requests.
setRequestHeader :: PandocMonad m
                 => String  -- ^ Header name
                 -> String  -- ^ Value
                 -> m ()
setRequestHeader name val = modifyCommonState $ \st ->
  st{ stRequestHeaders =
       (name, val) : filter (\(n,_) -> n /= name) (stRequestHeaders st)  }

-- | Initialize the media bag.
setMediaBag :: PandocMonad m => MediaBag -> m ()
setMediaBag mb = modifyCommonState $ \st -> st{stMediaBag = mb}

-- Retrieve the media bag.
getMediaBag :: PandocMonad m => m MediaBag
getMediaBag = getsCommonState stMediaBag

-- Insert an item into the media bag.
insertMedia :: PandocMonad m => FilePath -> Maybe MimeType -> BL.ByteString -> m ()
insertMedia fp mime bs = do
  mb <- getMediaBag
  let mb' = MB.insertMedia fp mime bs mb
  setMediaBag mb'

-- Retrieve the input filenames.
getInputFiles :: PandocMonad m => m [FilePath]
getInputFiles = getsCommonState stInputFiles

-- Set the input filenames.
setInputFiles :: PandocMonad m => [FilePath] -> m ()
setInputFiles fs = do
  let sourceURL = case fs of
                    []    -> Nothing
                    (x:_) -> case parseURI x of
                                Just u
                                  | uriScheme u `elem` ["http:","https:"] ->
                                      Just $ show u{ uriQuery = "",
                                                     uriFragment = "" }
                                _ -> Nothing

  modifyCommonState $ \st -> st{ stInputFiles = fs
                               , stSourceURL = sourceURL }

-- Retrieve the output filename.
getOutputFile :: PandocMonad m => m (Maybe FilePath)
getOutputFile = getsCommonState stOutputFile

-- Set the output filename.
setOutputFile :: PandocMonad m => Maybe FilePath -> m ()
setOutputFile mbf = modifyCommonState $ \st -> st{ stOutputFile = mbf }

-- Retrieve the resource path searched by 'fetchItem'.
getResourcePath :: PandocMonad m => m [FilePath]
getResourcePath = getsCommonState stResourcePath

-- Set the resource path searched by 'fetchItem'.
setResourcePath :: PandocMonad m => [FilePath] -> m ()
setResourcePath ps = modifyCommonState $ \st -> st{stResourcePath = ps}

-- Get the POSIX time.
getPOSIXTime :: PandocMonad m => m POSIXTime
getPOSIXTime = utcTimeToPOSIXSeconds <$> getCurrentTime

-- Get the zoned time.
getZonedTime :: PandocMonad m => m ZonedTime
getZonedTime = do
  t <- getCurrentTime
  tz <- getCurrentTimeZone
  return $ utcToZonedTime tz t

-- | Read file, checking in any number of directories.
readFileFromDirs :: PandocMonad m => [FilePath] -> FilePath -> m (Maybe String)
readFileFromDirs [] _ = return Nothing
readFileFromDirs (d:ds) f = catchError
    ((Just . UTF8.toStringLazy) <$> readFileLazy (d </> f))
    (\_ -> readFileFromDirs ds f)

--

-- | 'CommonState' represents state that is used by all
-- instances of 'PandocMonad'.  Normally users should not
-- need to interact with it directly; instead, auxiliary
-- functions like 'setVerbosity' and 'withMediaBag' should be used.
data CommonState = CommonState { stLog          :: [LogMessage]
                                 -- ^ A list of log messages in reverse order
                               , stUserDataDir  :: Maybe FilePath
                                 -- ^ Directory to search for data files
                               , stSourceURL    :: Maybe String
                                 -- ^ Absolute URL + dir of 1st source file
                               , stRequestHeaders :: [(String, String)]
                                 -- ^ Headers to add for HTTP requests
                               , stMediaBag     :: MediaBag
                                 -- ^ Media parsed from binary containers
                               , stTranslations :: Maybe
                                                  (Lang, Maybe Translations)
                                 -- ^ Translations for localization
                               , stInputFiles   :: [FilePath]
                                 -- ^ List of input files from command line
                               , stOutputFile   :: Maybe FilePath
                                 -- ^ Output file from command line
                               , stResourcePath :: [FilePath]
                                 -- ^ Path to search for resources like
                                 -- included images
                               , stVerbosity    :: Verbosity
                                 -- ^ Verbosity level
                               , stTrace        :: Bool
                                 -- ^ Controls whether tracing messages are
                                 -- issued.
                               }

instance Default CommonState where
  def = CommonState { stLog = []
                    , stUserDataDir = Nothing
                    , stSourceURL = Nothing
                    , stRequestHeaders = []
                    , stMediaBag = mempty
                    , stTranslations = Nothing
                    , stInputFiles = []
                    , stOutputFile = Nothing
                    , stResourcePath = ["."]
                    , stVerbosity = WARNING
                    , stTrace = False
                    }

-- | Convert BCP47 string to a Lang, issuing warning
-- if there are problems.
toLang :: PandocMonad m => Maybe String -> m (Maybe Lang)
toLang Nothing = return Nothing
toLang (Just s) =
  case parseBCP47 s of
       Left _ -> do
         report $ InvalidLang s
         return Nothing
       Right l -> return (Just l)

-- | Select the language to use with 'translateTerm'.
-- Note that this does not read a translation file;
-- that is only done the first time 'translateTerm' is
-- used.
setTranslations :: PandocMonad m => Lang -> m ()
setTranslations lang =
  modifyCommonState $ \st -> st{ stTranslations = Just (lang, Nothing) }

-- | Load term map.
getTranslations :: PandocMonad m => m Translations
getTranslations = do
  mbtrans <- getsCommonState stTranslations
  case mbtrans of
       Nothing -> return mempty  -- no language defined
       Just (_, Just t) -> return t
       Just (lang, Nothing) -> do  -- read from file
         let translationFile = "translations/" ++ renderLang lang ++ ".yaml"
         let fallbackFile = "translations/" ++ langLanguage lang ++ ".yaml"
         let getTrans fp = do
               bs <- readDataFile fp
               case readTranslations (UTF8.toString bs) of
                    Left e   -> do
                      report $ CouldNotLoadTranslations (renderLang lang)
                        (fp ++ ": " ++ e)
                      -- make sure we don't try again...
                      modifyCommonState $ \st ->
                        st{ stTranslations = Nothing }
                      return mempty
                    Right t -> do
                      modifyCommonState $ \st ->
                                  st{ stTranslations = Just (lang, Just t) }
                      return t
         catchError (getTrans translationFile)
           (\_ ->
             catchError (getTrans fallbackFile)
               (\e -> do
                 report $ CouldNotLoadTranslations (renderLang lang)
                          $ case e of
                               PandocCouldNotFindDataFileError _ ->
                                 "data file " ++ fallbackFile ++ " not found"
                               _ -> ""
                 -- make sure we don't try again...
                 modifyCommonState $ \st -> st{ stTranslations = Nothing }
                 return mempty))

-- | Get a translation from the current term map.
-- Issue a warning if the term is not defined.
translateTerm :: PandocMonad m => Term -> m String
translateTerm term = do
  translations <- getTranslations
  case lookupTerm term translations of
       Just s -> return s
       Nothing -> do
         report $ NoTranslation (show term)
         return ""

-- | Evaluate a 'PandocIO' operation.
runIO :: PandocIO a -> IO (Either PandocError a)
runIO ma = flip evalStateT def $ runExceptT $ unPandocIO ma

-- | Evaluate a 'PandocIO' operation, handling any errors
-- by exiting with an appropriate message and error status. 
runIOorExplode :: PandocIO a -> IO a
runIOorExplode ma = runIO ma >>= handleError

newtype PandocIO a = PandocIO {
  unPandocIO :: ExceptT PandocError (StateT CommonState IO) a
  } deriving ( MonadIO
             , Functor
             , Applicative
             , Monad
             , MonadError PandocError
             )

-- | Utility function to lift IO errors into 'PandocError's.
liftIOError :: (String -> IO a) -> String -> PandocIO a
liftIOError f u = do
  res <- liftIO $ tryIOError $ f u
  case res of
         Left e  -> throwError $ PandocIOError u e
         Right r -> return r

instance PandocMonad PandocIO where
  lookupEnv = liftIO . IO.lookupEnv
  getCurrentTime = liftIO IO.getCurrentTime
  getCurrentTimeZone = liftIO IO.getCurrentTimeZone
  newStdGen = liftIO IO.newStdGen
  newUniqueHash = hashUnique <$> liftIO IO.newUnique

  openURL u
   | Just u'' <- stripPrefix "data:" u = do
       let mime     = takeWhile (/=',') u''
       let contents = UTF8.fromString $
                       unEscapeString $ drop 1 $ dropWhile (/=',') u''
       return (decodeLenient contents, Just mime)
   | otherwise = do
       let toReqHeader (n, v) = (CI.mk (UTF8.fromString n), UTF8.fromString v)
       customHeaders <- map toReqHeader <$> getsCommonState stRequestHeaders
       report $ Fetching u
       res <- liftIO $ E.try $ withSocketsDo $ do
         let parseReq = parseRequest
         proxy <- tryIOError (getEnv "http_proxy")
         let addProxy' x = case proxy of
                              Left _ -> return x
                              Right pr -> parseReq pr >>= \r ->
                                  return (addProxy (host r) (port r) x)
         req <- parseReq u >>= addProxy'
         let req' = req{requestHeaders = customHeaders ++ requestHeaders req}
         resp <- newManager tlsManagerSettings >>= httpLbs req'
         return (B.concat $ toChunks $ responseBody resp,
                 UTF8.toString `fmap` lookup hContentType (responseHeaders resp))

       case res of
            Right r -> return r
            Left e  -> throwError $ PandocHttpError u e

  readFileLazy s = liftIOError BL.readFile s
  readFileStrict s = liftIOError B.readFile s

  glob = liftIOError IO.glob
  fileExists = liftIOError Directory.doesFileExist
#ifdef EMBED_DATA_FILES
  getDataFileName = return
#else
  getDataFileName = liftIOError Paths.getDataFileName
#endif
  getModificationTime = liftIOError IO.getModificationTime
  getCommonState = PandocIO $ lift get
  putCommonState x = PandocIO $ lift $ put x
  logOutput msg = liftIO $ do
    UTF8.hPutStr stderr $
        "[" ++ show (messageVerbosity msg) ++ "] "
    alertIndent $ lines $ showLogMessage msg

alertIndent :: [String] -> IO ()
alertIndent [] = return ()
alertIndent (l:ls) = do
  UTF8.hPutStrLn stderr l
  mapM_ go ls
  where go l' = do UTF8.hPutStr stderr "  "
                   UTF8.hPutStrLn stderr l'

-- | Specialized version of parseURIReference that disallows
-- single-letter schemes.  Reason:  these are usually windows absolute
-- paths.
parseURIReference' :: String -> Maybe URI
parseURIReference' s =
  case parseURIReference s of
       Just u
         | length (uriScheme u) > 2  -> Just u
         | null (uriScheme u)        -> Just u  -- protocol-relative
       _                             -> Nothing

-- | Set the user data directory in common state.
setUserDataDir :: PandocMonad m
               => Maybe FilePath
               -> m ()
setUserDataDir mbfp = modifyCommonState $ \st -> st{ stUserDataDir = mbfp }

-- | Get the user data directory from common state.
getUserDataDir :: PandocMonad m
               => m (Maybe FilePath)
getUserDataDir = getsCommonState stUserDataDir

-- | Fetch an image or other item from the local filesystem or the net.
-- Returns raw content and maybe mime type.
fetchItem :: PandocMonad m
          => String
          -> m (B.ByteString, Maybe MimeType)
fetchItem s = do
  mediabag <- getMediaBag
  case lookupMedia s mediabag of
    Just (mime, bs) -> return (BL.toStrict bs, Just mime)
    Nothing -> downloadOrRead s

downloadOrRead :: PandocMonad m
               => String
               -> m (B.ByteString, Maybe MimeType)
downloadOrRead s = do
  sourceURL <- getsCommonState stSourceURL
  case (sourceURL >>= parseURIReference' .
                       ensureEscaped, ensureEscaped s) of
    (Just u, s') -> -- try fetching from relative path at source
       case parseURIReference' s' of
            Just u' -> openURL $ show $ u' `nonStrictRelativeTo` u
            Nothing -> openURL s' -- will throw error
    (Nothing, s'@('/':'/':_)) ->  -- protocol-relative URI
       case parseURIReference' s' of
            Just u' -> openURL $ show $ u' `nonStrictRelativeTo` httpcolon
            Nothing -> openURL s' -- will throw error
    (Nothing, s') ->
       case parseURI s' of  -- requires absolute URI
            -- We don't want to treat C:/ as a scheme:
            Just u' | length (uriScheme u') > 2 -> openURL (show u')
            Just u' | uriScheme u' == "file:" ->
                 readLocalFile $ dropWhile (=='/') (uriPath u')
            _ -> readLocalFile fp -- get from local file system
   where readLocalFile f = do
             resourcePath <- getResourcePath
             cont <- if isRelative f
                        then withPaths resourcePath readFileStrict f
                        else readFileStrict f
             return (cont, mime)
         httpcolon = URI{ uriScheme = "http:",
                          uriAuthority = Nothing,
                          uriPath = "",
                          uriQuery = "",
                          uriFragment = "" }
         dropFragmentAndQuery = takeWhile (\c -> c /= '?' && c /= '#')
         fp = unEscapeString $ dropFragmentAndQuery s
         mime = case takeExtension fp of
                     ".gz" -> getMimeType $ dropExtension fp
                     ".svgz" -> getMimeType $ dropExtension fp ++ ".svg"
                     x     -> getMimeType x
         ensureEscaped = escapeURIString isAllowedInURI . map convertSlash
         convertSlash '\\' = '/'
         convertSlash x    = x

-- Retrieve default reference.docx.
getDefaultReferenceDocx :: PandocMonad m => m Archive
getDefaultReferenceDocx = do
  let paths = ["[Content_Types].xml",
               "_rels/.rels",
               "docProps/app.xml",
               "docProps/core.xml",
               "word/document.xml",
               "word/fontTable.xml",
               "word/footnotes.xml",
               "word/numbering.xml",
               "word/settings.xml",
               "word/webSettings.xml",
               "word/styles.xml",
               "word/_rels/document.xml.rels",
               "word/_rels/footnotes.xml.rels",
               "word/theme/theme1.xml"]
  let toLazy = BL.fromChunks . (:[])
  let pathToEntry path = do
        epochtime <- (floor . utcTimeToPOSIXSeconds) <$> getCurrentTime
        contents <- toLazy <$> readDataFile ("docx/" ++ path)
        return $ toEntry path epochtime contents
  datadir <- getUserDataDir
  mbArchive <- case datadir of
                    Nothing   -> return Nothing
                    Just d    -> do
                       exists <- fileExists (d </> "reference.docx")
                       if exists
                          then return (Just (d </> "reference.docx"))
                          else return Nothing
  case mbArchive of
     Just arch -> toArchive <$> readFileLazy arch
     Nothing   -> foldr addEntryToArchive emptyArchive <$>
                     mapM pathToEntry paths

-- Retrieve default reference.odt.
getDefaultReferenceODT :: PandocMonad m => m Archive
getDefaultReferenceODT = do
  let paths = ["mimetype",
               "manifest.rdf",
               "styles.xml",
               "content.xml",
               "meta.xml",
               "settings.xml",
               "Configurations2/accelerator/current.xml",
               "Thumbnails/thumbnail.png",
               "META-INF/manifest.xml"]
  let pathToEntry path = do epochtime <- floor `fmap` getPOSIXTime
                            contents <- (BL.fromChunks . (:[])) `fmap`
                                          readDataFile ("odt/" ++ path)
                            return $ toEntry path epochtime contents
  datadir <- getUserDataDir
  mbArchive <- case datadir of
                    Nothing   -> return Nothing
                    Just d    -> do
                       exists <- fileExists (d </> "reference.odt")
                       if exists
                          then return (Just (d </> "reference.odt"))
                          else return Nothing
  case mbArchive of
     Just arch -> toArchive <$> readFileLazy arch
     Nothing   -> foldr addEntryToArchive emptyArchive <$>
                     mapM pathToEntry paths

-- | Read file from user data directory or,
-- if not found there, from Cabal data directory.
readDataFile :: PandocMonad m => FilePath -> m B.ByteString
readDataFile fname = do
  datadir <- getUserDataDir
  case datadir of
       Nothing -> readDefaultDataFile fname
       Just userDir -> do
         exists <- fileExists (userDir </> fname)
         if exists
            then readFileStrict (userDir </> fname)
            else readDefaultDataFile fname

-- | Read file from from Cabal data directory.
readDefaultDataFile :: PandocMonad m => FilePath -> m B.ByteString
readDefaultDataFile "reference.docx" =
  (B.concat . BL.toChunks . fromArchive) <$> getDefaultReferenceDocx
readDefaultDataFile "reference.odt" =
  (B.concat . BL.toChunks . fromArchive) <$> getDefaultReferenceODT
readDefaultDataFile fname =
#ifdef EMBED_DATA_FILES
  case lookup (makeCanonical fname) dataFiles of
    Nothing       -> throwError $ PandocCouldNotFindDataFileError fname
    Just contents -> return contents
  where makeCanonical = Posix.joinPath . transformPathParts . splitDirectories
        transformPathParts = reverse . foldl go []
        go as     "."  = as
        go (_:as) ".." = as
        go as     x    = x : as
#else
  getDataFileName fname' >>= checkExistence >>= readFileStrict
    where fname' = if fname == "MANUAL.txt" then fname else "data" </> fname

checkExistence :: PandocMonad m => FilePath -> m FilePath
checkExistence fn = do
  exists <- fileExists fn
  if exists
     then return fn
     else throwError $ PandocCouldNotFindDataFileError fn
#endif

withPaths :: PandocMonad m => [FilePath] -> (FilePath -> m a) -> FilePath -> m a
withPaths [] _ fp = throwError $ PandocResourceNotFound fp
withPaths (p:ps) action fp =
  catchError (action (p </> fp))
             (\_ -> withPaths ps action fp)

-- | Fetch local or remote resource (like an image) and provide data suitable
-- for adding it to the MediaBag.
fetchMediaResource :: PandocMonad m
              => String -> m (FilePath, Maybe MimeType, BL.ByteString)
fetchMediaResource src = do
  (bs, mt) <- downloadOrRead src
  let ext = fromMaybe (takeExtension src)
                      (mt >>= extensionFromMimeType)
  let bs' = BL.fromChunks [bs]
  let basename = showDigest $ sha1 bs'
  let fname = basename <.> ext
  return (fname, mt, bs')

-- | Traverse tree, filling media bag for any images that
-- aren't already in the media bag.
fillMediaBag :: PandocMonad m => Pandoc -> m Pandoc
fillMediaBag d = walkM handleImage d
  where handleImage :: PandocMonad m => Inline -> m Inline
        handleImage (Image attr lab (src, tit)) = catchError
          (do mediabag <- getMediaBag
              case lookupMedia src mediabag of
                Just (_, _) -> return $ Image attr lab (src, tit)
                Nothing -> do
                  (fname, mt, bs) <- fetchMediaResource src
                  insertMedia fname mt bs
                  return $ Image attr lab (fname, tit))
          (\e ->
              case e of
                PandocResourceNotFound _ -> do
                  report $ CouldNotFetchResource src
                            "replacing image with description"
                  -- emit alt text
                  return $ Span ("",["image"],[]) lab
                PandocHttpError u er -> do
                  report $ CouldNotFetchResource u
                            (show er ++ "\rReplacing image with description.")
                  -- emit alt text
                  return $ Span ("",["image"],[]) lab
                _ -> throwError e)
        handleImage x = return x

-- | Extract media from the mediabag into a directory.
extractMedia :: FilePath -> Pandoc -> PandocIO Pandoc
extractMedia dir d = do
  media <- getMediaBag
  case [fp | (fp, _, _) <- mediaDirectory media] of
        []  -> return d
        fps -> do
          mapM_ (writeMedia dir media) fps
          return $ walk (adjustImagePath dir fps) d

-- Write the contents of a media bag to a path.
writeMedia :: FilePath -> MediaBag -> FilePath -> PandocIO ()
writeMedia dir mediabag subpath = do
  -- we join and split to convert a/b/c to a\b\c on Windows;
  -- in zip containers all paths use /
  let fullpath = dir </> normalise subpath
  let mbcontents = lookupMedia subpath mediabag
  case mbcontents of
       Nothing -> throwError $ PandocResourceNotFound subpath
       Just (_, bs) -> do
         report $ Extracting fullpath
         liftIOError (createDirectoryIfMissing True) (takeDirectory fullpath)
         liftIOError (\p -> BL.writeFile p bs) fullpath

adjustImagePath :: FilePath -> [FilePath] -> Inline -> Inline
adjustImagePath dir paths (Image attr lab (src, tit))
   | src `elem` paths = Image attr lab (dir ++ "/" ++ src, tit)
adjustImagePath _ _ x = x

-- | The 'PureState' contains ersatz representations
-- of things that would normally be obtained through IO.
data PureState = PureState { stStdGen     :: StdGen
                           , stWord8Store :: [Word8] -- should be
                                                     -- inifinite,
                                                     -- i.e. [1..]
                           , stUniqStore  :: [Int] -- should be
                                                   -- inifinite and
                                                   -- contain every
                                                   -- element at most
                                                   -- once, e.g. [1..]
                           , stEnv :: [(String, String)]
                           , stTime :: UTCTime
                           , stTimeZone :: TimeZone
                           , stReferenceDocx :: Archive
                           , stReferenceODT :: Archive
                           , stFiles :: FileTree
                           , stUserDataFiles :: FileTree
                           , stCabalDataFiles :: FileTree
                           }

instance Default PureState where
  def = PureState { stStdGen = mkStdGen 1848
                  , stWord8Store = [1..]
                  , stUniqStore = [1..]
                  , stEnv = [("USER", "pandoc-user")]
                  , stTime = posixSecondsToUTCTime 0
                  , stTimeZone = utc
                  , stReferenceDocx = emptyArchive
                  , stReferenceODT = emptyArchive
                  , stFiles = mempty
                  , stUserDataFiles = mempty
                  , stCabalDataFiles = mempty
                  }


getPureState :: PandocPure PureState
getPureState = PandocPure $ lift $ lift get

getsPureState :: (PureState -> a) -> PandocPure a
getsPureState f = f <$> getPureState

putPureState :: PureState -> PandocPure ()
putPureState ps= PandocPure $ lift $ lift $ put ps

modifyPureState :: (PureState -> PureState) -> PandocPure ()
modifyPureState f = PandocPure $ lift $ lift $ modify f


data FileInfo = FileInfo { infoFileMTime :: UTCTime
                         , infoFileContents :: B.ByteString
                         }

newtype FileTree = FileTree {unFileTree :: M.Map FilePath FileInfo}
  deriving (Monoid)

getFileInfo :: FilePath -> FileTree -> Maybe FileInfo
getFileInfo fp tree = M.lookup fp $ unFileTree tree

-- | Add the specified file to the FileTree. If file
-- is a directory, add its contents recursively.
addToFileTree :: FileTree -> FilePath -> IO FileTree
addToFileTree (FileTree treemap) fp = do
  isdir <- doesDirectoryExist fp
  if isdir
     then do -- recursively add contents of directories
       let isSpecial ".." = True
           isSpecial "."  = True
           isSpecial _    = False
       fs <- (map (fp </>) . filter (not . isSpecial)) <$> getDirectoryContents fp
       foldM addToFileTree (FileTree treemap) fs
     else do
       contents <- B.readFile fp
       mtime <- IO.getModificationTime fp
       return $ FileTree $
                M.insert fp FileInfo{ infoFileMTime = mtime
                                    , infoFileContents = contents } treemap

newtype PandocPure a = PandocPure {
  unPandocPure :: ExceptT PandocError
                  (StateT CommonState (State PureState)) a
  } deriving ( Functor
             , Applicative
             , Monad
             , MonadError PandocError
             )

-- Run a 'PandocPure' operation.
runPure :: PandocPure a -> Either PandocError a
runPure x = flip evalState def $
            flip evalStateT def $
            runExceptT $
            unPandocPure x

instance PandocMonad PandocPure where
  lookupEnv s = do
    env <- getsPureState stEnv
    return (lookup s env)

  getCurrentTime = getsPureState stTime

  getCurrentTimeZone = getsPureState stTimeZone

  newStdGen = do
    g <- getsPureState stStdGen
    let (_, nxtGen) = next g
    modifyPureState $ \st -> st { stStdGen = nxtGen }
    return g

  newUniqueHash = do
    uniqs <- getsPureState stUniqStore
    case uniqs of
      u : us -> do
        modifyPureState $ \st -> st { stUniqStore = us }
        return u
      _ -> M.fail "uniq store ran out of elements"
  openURL u = throwError $ PandocResourceNotFound u
  readFileLazy fp = do
    fps <- getsPureState stFiles
    case infoFileContents <$> getFileInfo fp fps of
      Just bs -> return (BL.fromStrict bs)
      Nothing -> throwError $ PandocResourceNotFound fp
  readFileStrict fp = do
    fps <- getsPureState stFiles
    case infoFileContents <$> getFileInfo fp fps of
      Just bs -> return bs
      Nothing -> throwError $ PandocResourceNotFound fp

  glob s = do
    FileTree ftmap <- getsPureState stFiles
    return $ filter (match (compile s)) $ M.keys ftmap

  fileExists fp = do
    fps <- getsPureState stFiles
    case getFileInfo fp fps of
         Nothing -> return False
         Just _  -> return True

  getDataFileName fp = return $ "data/" ++ fp

  getModificationTime fp = do
    fps <- getsPureState stFiles
    case infoFileMTime <$> getFileInfo fp fps of
      Just tm -> return tm
      Nothing -> throwError $ PandocIOError fp
                    (userError "Can't get modification time")

  getCommonState = PandocPure $ lift get
  putCommonState x = PandocPure $ lift $ put x

  logOutput _msg = return ()

-- This requires UndecidableInstances.  We could avoid that
-- by repeating the definitions below for every monad transformer
-- we use: ReaderT, WriterT, StateT, RWST.  But this seems to
-- be harmless.
instance (MonadTrans t, PandocMonad m, Functor (t m),
          MonadError PandocError (t m), Monad (t m),
          Applicative (t m)) => PandocMonad (t m) where
  lookupEnv = lift . lookupEnv
  getCurrentTime = lift getCurrentTime
  getCurrentTimeZone = lift getCurrentTimeZone
  newStdGen = lift newStdGen
  newUniqueHash = lift newUniqueHash
  openURL = lift . openURL
  readFileLazy = lift . readFileLazy
  readFileStrict = lift . readFileStrict
  glob = lift . glob
  fileExists = lift . fileExists
  getDataFileName = lift . getDataFileName
  getModificationTime = lift . getModificationTime
  getCommonState = lift getCommonState
  putCommonState = lift . putCommonState
  logOutput = lift . logOutput

#if MIN_VERSION_base(4,8,0)
instance {-# OVERLAPS #-} PandocMonad m => PandocMonad (ParsecT s st m) where
#else
instance PandocMonad m => PandocMonad (ParsecT s st m) where
#endif
  lookupEnv = lift . lookupEnv
  getCurrentTime = lift getCurrentTime
  getCurrentTimeZone = lift getCurrentTimeZone
  newStdGen = lift newStdGen
  newUniqueHash = lift newUniqueHash
  openURL = lift . openURL
  readFileLazy = lift . readFileLazy
  readFileStrict = lift . readFileStrict
  glob = lift . glob
  fileExists = lift . fileExists
  getDataFileName = lift . getDataFileName
  getModificationTime = lift . getModificationTime
  getCommonState = lift getCommonState
  putCommonState = lift . putCommonState
  trace msg = do
    tracing <- getsCommonState stTrace
    when tracing $ do
      pos <- getPosition
      Debug.Trace.trace
        ("[trace] Parsed " ++ msg ++ " at line " ++
            show (sourceLine pos) ++
            if sourceName pos == "chunk"
               then " of chunk"
               else "")
        (return ())
  logOutput = lift . logOutput