Hufflepdf/examples/getText.hs

import Control.Monad ((>=>))
import qualified Data.ByteString.Char8 as BS (readFile)
import Data.Map (Map, foldlWithKey, mapWithKey)
import qualified Data.Map as Map (empty, insert, toList, union)
import qualified Data.Text as Text (unpack)
import qualified Data.Text.IO as Text (putStrLn)
import PDF (UnifiedLayers(..), parseDocument)
import PDF.Box (Box(..))
import PDF.Content.Text (Chunks(..))
import PDF.Layer (Layer)
import PDF.Pages (Contents(..), Page(..), PageNumber(..), Pages(..), cacheFonts)
import System.Environment (getArgs)
import System.Exit (die)
import System.IO (BufferMode(..), hSetBuffering, stdout)
import Text.Printf (printf)

displayPage :: Page -> IO ()
displayPage =
    r Contents
  >=> mapM (\(objectId, content) -> r Chunks content >>= return . mixIn objectId)
  >=> fusion
  >=> sequence_ . mapWithKey (\k v -> putStrLn $ printf "%s: %s" (show k) (Text.unpack v))
  -- >=> mapM_ Text.putStrLn
  where
    mixIn :: (Ord k1, Ord k2) => k1 -> Map k2 v -> Map (k1, k2) v
    mixIn prefix =
      foldlWithKey (\m k v -> Map.insert (prefix, k) v m) Map.empty
    fusion :: (Monad m, Ord a) => [Map a b] -> m (Map a b)
    fusion = return . foldl Map.union Map.empty

getAll :: Layer -> IO ()
getAll layer = (cacheFonts $ r Pages layer) >>= mapM_ (displayPage . snd) . Map.toList

get :: Int -> Layer -> IO ()
get n layer = (cacheFonts $ r (P n) layer) >>= displayPage

onDoc :: FilePath -> (Layer -> IO ()) -> IO ()
onDoc inputFile f = do
    (parseDocument <$> BS.readFile inputFile)
  >>= either die (r UnifiedLayers >=> f)

main :: IO ()
main = do
  hSetBuffering stdout LineBuffering
  args <- getArgs
  case args of
    [inputFile] -> onDoc inputFile getAll
    [inputFile, pageNumber] -> onDoc inputFile (get $ read pageNumber)
    _ -> die "Syntax: getText INPUT_FILE [PAGE_NUMBER]"