FileRead.hs

-- | Module with function to read file in with pandoc and discard everything superfluous.
module Data.Text.WordCount.FileRead where

import Text.Pandoc
import Filesystem.Path.CurrentOS as F
import qualified Data.Text.Lazy as TL
import qualified Data.Text.Lazy.IO as TLIO
import qualified Data.ByteString.Lazy as BSL

eitherError = either (error . show) id

-- | Process a file given its filename. Return text only, discarding superflouous material.
processFile :: String -> IO TL.Text
processFile filepath = case (extension . decodeString $ filepath) of
    (Just "md") -> TL.pack . writePlain def . filterCode . eitherError . readMarkdown def . TL.unpack <$> TLIO.readFile filepath
    (Just "dbk") -> TL.pack . writePlain def . filterCode . eitherError . readDocBook def . TL.unpack <$> TLIO.readFile filepath
    (Just "docx") -> TL.pack . writePlain def . filterCode . fst . eitherError . readDocx def <$> BSL.readFile filepath
    (Just "epub") -> TL.pack . writePlain def . filterCode . fst . eitherError . readEPUB def <$> BSL.readFile filepath
    (Just "html") -> TL.pack . writePlain def . filterCode . eitherError . readHtml def . TL.unpack <$> TLIO.readFile filepath
    (Just "tex") -> TL.pack . writePlain def . filterCode . eitherError . readLaTeX def . TL.unpack <$> TLIO.readFile filepath
    (Just "xml") -> TL.pack . writePlain def . filterCode . eitherError . readOPML def . TL.unpack <$> TLIO.readFile filepath
    (Just "odt") -> TL.pack . writePlain def . filterCode . fst . eitherError . readOdt def <$> BSL.readFile filepath
    (Just "rst") -> TL.pack . writePlain def . filterCode . eitherError . readRST def . TL.unpack <$> TLIO.readFile filepath
    (Just "textile") -> TL.pack . writePlain def . filterCode . eitherError . readTextile def . TL.unpack <$> TLIO.readFile filepath
    _ -> TLIO.readFile filepath

-- | Filter out code and tables from the document
filterCode :: Pandoc -> Pandoc
filterCode (Pandoc meta content) = Pandoc meta $ filter rightBlock content
    where rightBlock CodeBlock { } = False
          rightBlock Table { }     = False
          rightBlock _             = True