1 files changed, 52 insertions, 0 deletions
diff --git a/app/Main.hs b/app/Main.hs
new file mode 100644
index 0000000..075414a
--- /dev/null
+++ b/app/Main.hs
@@ -0,0 +1,52 @@
+module Main where
+
+import Control.Exception (Exception, throwIO)
+import Data.ByteString.Lazy qualified as LB
+import Data.List
+import Data.String (IsString (fromString))
+import System.Directory
+import System.FilePath
+import System.IO.Temp (withSystemTempDirectory)
+import System.Process.Typed
+import Text.Printf (printf)
+
+main :: IO ()
+main = do
+  let input = "0000001.pdf"
+
+  withSystemTempDirectory input $ \tmp -> do
+    sh_ (printf "pdftoppm '%s' '%s' -png -r 300" input (tmp </> input))
+    imageInputs <- sort <$> listDirectory tmp
+    outputs <-
+      mapM
+        ( \imageInput -> do
+            sh_
+              ( printf
+                  "tesseract '%s' '%s' pdf -psm 1 -oem 1"
+                  (tmp </> imageInput)
+                  (tmp </> imageInput)
+              )
+            pure (imageInput <.> ".pdf")
+        )
+        imageInputs
+    sh_ ("pdfunite " ++ intercalate " " (map (printf "'%s'" . (tmp </>)) outputs ++ [printf "'%s'" (tmp </> input)]))
+    copyFile input (input <.> "bak")
+    copyFile (tmp </> input) ("." <> input)
+    renameFile ("." <> input) input
+    LB.putStr =<< sh (printf "pdftotext '%s' -" input)
+
+data ProcessException = ProcessException Int LB.ByteString
+  deriving (Show)
+
+instance Exception ProcessException
+
+sh :: String -> IO LB.ByteString
+sh cmd = do
+  -- printf "+ %s\n" cmd
+  (exitCode, out, err) <- readProcess (fromString cmd)
+  case exitCode of
+    ExitSuccess -> return out
+    ExitFailure exitCode' -> throwIO $ ProcessException exitCode' err
+
+sh_ :: String -> IO ()
+sh_ = fmap (\_ -> ()) . sh