chore: add attrs

1 files changed, 21 insertions, 2 deletions

diff --git a/app/Main.hs b/app/Main.hs
index 57e015d..b6441c8 100644
--- a/app/Main.hs
+++ b/app/Main.hs
@@ -9,6 +9,7 @@ import Control.Arrow (second)
 import Control.Concurrent.ParallelIO.Local (parallel, withPool)
 import Control.Exception (Exception, throw, throwIO)
 import Control.Monad (join, when)
+import Data.Aeson qualified as J
 import Data.Attoparsec.Text qualified as A
 import Data.ByteString.Lazy qualified as LB
 import Data.List
@@ -19,6 +20,7 @@ import Data.Text qualified as T
 import Data.Text.Encoding qualified as T
 import Debug.Trace
 import GHC.Conc (getNumProcessors)
+import GHC.Generics (Generic)
 import System.Directory
 import System.FilePath
 import System.IO.Temp (withSystemTempDirectory)
@@ -31,6 +33,7 @@ main =
   parMapM_ ocr
     =<< sort
       . map ("originals" </>)
+      . filter (not . (".attrs" `isSuffixOf`))
       . filter (not . (".bak" `isSuffixOf`))
       <$> listDirectory "originals"
 
@@ -40,9 +43,10 @@ debug s x =
 
 ocr :: FilePath -> IO ()
 ocr input = withSystemTempDirectory (takeBaseName input) $ \tmp -> do
-  hasText <-
-    (not . T.null) . T.strip . T.decodeUtf8 . LB.toStrict
+  originalText <-
+    T.decodeUtf8 . LB.toStrict
       <$> sh (printf "pdftotext '%s' -" input)
+  let hasText = (not . T.null) . T.strip $ originalText
   when (not hasText) do
     let fn suffix = tmp </> takeBaseName input <> suffix
     pdfInfo <- parsePdfInfo <$> sh (printf "pdfinfo '%s'" input)
@@ -75,6 +79,12 @@ ocr input = withSystemTempDirectory (takeBaseName input) $ \tmp -> do
     copyFile (fn ".pdf") (takeDirectory input </> "." <> takeBaseName input <.> "pdf")
     printf "~ renameFile %s %s\n" (takeDirectory input </> "." <> takeBaseName input <.> "pdf") input
     renameFile (takeDirectory input </> "." <> takeBaseName input <.> "pdf") input
+  let attrsFile = takeDirectory input </> takeBaseName input <.> "attrs"
+  doesAttrsFileExist <- doesFileExist attrsFile
+  when (not doesAttrsFileExist) $
+    J.encodeFile attrsFile Attrs {..}
+  Just attrs <- J.decodeFileStrict attrsFile
+  print (attrs :: Attrs)
 
 ocr1 :: FilePath -> FilePath -> IO FilePath
 ocr1 tmp input = do
@@ -86,6 +96,15 @@ ocr1 tmp input = do
     )
   pure (takeBaseName input <.> "pdf")
 
+data Attrs = Attrs
+  { originalText :: T.Text
+  }
+  deriving (Show, Generic, Eq)
+
+instance J.ToJSON Attrs
+
+instance J.FromJSON Attrs
+
 data PdfInfo = PdfInfo
   { numPages :: Int,
     pageSize :: (Double, Double)