Add option to decrypt PDFs during conversion

Refs: #1074

build.sbt

@@ -275,6 +275,7 @@ val common = project
   .settings(testSettingsMUnit)
   .settings(
     name := "docspell-common",
+    addCompilerPlugin(Dependencies.kindProjectorPlugin),
     libraryDependencies ++=
       Dependencies.fs2 ++
         Dependencies.circe ++
@@ -409,7 +410,8 @@ val convert = project
     name := "docspell-convert",
     libraryDependencies ++=
       Dependencies.flexmark ++
-        Dependencies.twelvemonkeys
+        Dependencies.twelvemonkeys ++
+        Dependencies.pdfbox
   )
   .dependsOn(common, files % "compile->compile;test->test")
 

modules/common/src/main/scala/docspell/common/Logger.scala

@@ -7,12 +7,13 @@
 package docspell.common
 
 import cats.effect.Sync
+import fs2.Stream
 
 import docspell.common.syntax.all._
 
 import org.log4s.{Logger => Log4sLogger}
 
-trait Logger[F[_]] {
+trait Logger[F[_]] { self =>
 
   def trace(msg: => String): F[Unit]
   def debug(msg: => String): F[Unit]
@@ -21,6 +22,25 @@ trait Logger[F[_]] {
   def error(ex: Throwable)(msg: => String): F[Unit]
   def error(msg: => String): F[Unit]
 
+  final def s: Logger[Stream[F, *]] = new Logger[Stream[F, *]] {
+    def trace(msg: => String): Stream[F, Unit] =
+      Stream.eval(self.trace(msg))
+
+    def debug(msg: => String): Stream[F, Unit] =
+      Stream.eval(self.debug(msg))
+
+    def info(msg: => String): Stream[F, Unit] =
+      Stream.eval(self.info(msg))
+
+    def warn(msg: => String): Stream[F, Unit] =
+      Stream.eval(self.warn(msg))
+
+    def error(msg: => String): Stream[F, Unit] =
+      Stream.eval(self.error(msg))
+
+    def error(ex: Throwable)(msg: => String): Stream[F, Unit] =
+      Stream.eval(self.error(ex)(msg))
+  }
 }
 
 object Logger {

modules/convert/src/main/scala/docspell/convert/Conversion.scala

@@ -42,8 +42,12 @@ object Conversion {
       ): F[A] =
         TikaMimetype.resolve(dataType, in).flatMap {
           case MimeType.PdfMatch(_) =>
+            val pdfStream =
+              if (cfg.decryptPdf.enabled)
+                in.through(RemovePdfEncryption(logger, cfg.decryptPdf.passwords))
+              else in
             OcrMyPdf
-              .toPDF(cfg.ocrmypdf, lang, cfg.chunkSize, logger)(in, handler)
+              .toPDF(cfg.ocrmypdf, lang, cfg.chunkSize, logger)(pdfStream, handler)
 
           case MimeType.HtmlMatch(mt) =>
             val cs = mt.charsetOrUtf8

modules/convert/src/main/scala/docspell/convert/ConvertConfig.scala

@@ -6,11 +6,12 @@
 
 package docspell.convert
 
+import docspell.convert.ConvertConfig.DecryptPdf
 import docspell.convert.extern.OcrMyPdfConfig
 import docspell.convert.extern.{TesseractConfig, UnoconvConfig, WkHtmlPdfConfig}
 import docspell.convert.flexmark.MarkdownConfig
 
-case class ConvertConfig(
+final case class ConvertConfig(
     chunkSize: Int,
     convertedFilenamePart: String,
     maxImageSize: Int,
@@ -18,5 +19,11 @@ case class ConvertConfig(
     wkhtmlpdf: WkHtmlPdfConfig,
     tesseract: TesseractConfig,
     unoconv: UnoconvConfig,
-    ocrmypdf: OcrMyPdfConfig
+    ocrmypdf: OcrMyPdfConfig,
+    decryptPdf: DecryptPdf
 )
+
+object ConvertConfig {
+
+  final case class DecryptPdf(enabled: Boolean, passwords: List[String])
+}

modules/convert/src/main/scala/docspell/convert/RemovePdfEncryption.scala

@@ -0,0 +1,88 @@
+/*
+ * Copyright 2020 Eike K. & Contributors
+ *
+ * SPDX-License-Identifier: AGPL-3.0-or-later
+ */
+
+package docspell.convert
+
+import java.io.ByteArrayOutputStream
+
+import cats.effect._
+import fs2.{Chunk, Pipe, Stream}
+
+import docspell.common.Logger
+
+import org.apache.pdfbox.pdmodel.PDDocument
+import org.apache.pdfbox.pdmodel.encryption.InvalidPasswordException
+
+/** Using PDFBox, the incoming pdf is loaded while trying the given passwords. */
+object RemovePdfEncryption {
+
+  def apply[F[_]: Sync](
+      logger: Logger[F],
+      passwords: List[String]
+  ): Pipe[F, Byte, Byte] =
+    apply(logger, Stream.emits(passwords))
+
+  def apply[F[_]: Sync](
+      logger: Logger[F],
+      passwords: Stream[F, String]
+  ): Pipe[F, Byte, Byte] = {
+    val pws = passwords.cons1("")
+    in =>
+      pws
+        .flatMap(pw => in.through(openPdf[F](logger, pw)))
+        .head
+        .flatMap { doc =>
+          if (doc.isEncrypted) {
+            logger.s.debug("Removing protection/encryption from PDF").drain ++
+              Stream.eval(Sync[F].delay(doc.setAllSecurityToBeRemoved(true))).drain ++
+              toStream[F](doc)
+          } else {
+            in
+          }
+        }
+        .ifEmpty(
+          logger.s
+            .info(
+              s"None of the passwords helped to read the given PDF!"
+            )
+            .drain ++ in
+        )
+  }
+
+  private def openPdf[F[_]: Sync](
+      logger: Logger[F],
+      pw: String
+  ): Pipe[F, Byte, PDDocument] = {
+    def alloc(bytes: Array[Byte]): F[Option[PDDocument]] =
+      Sync[F].delay(load(bytes, pw))
+
+    def free(doc: Option[PDDocument]): F[Unit] =
+      Sync[F].delay(doc.foreach(_.close()))
+
+    val log =
+      if (pw.isEmpty) Stream.empty
+      else logger.s.debug(s"Try opening PDF with password: ${pw.take(2)}***").drain
+
+    in =>
+      Stream
+        .eval(in.compile.to(Array))
+        .flatMap(bytes => log ++ Stream.bracket(alloc(bytes))(free))
+        .flatMap(opt => opt.map(Stream.emit).getOrElse(Stream.empty))
+  }
+
+  private def load(bytes: Array[Byte], pw: String): Option[PDDocument] =
+    try Option(PDDocument.load(bytes, pw))
+    catch {
+      case _: InvalidPasswordException =>
+        None
+    }
+
+  private def toStream[F[_]](doc: PDDocument): Stream[F, Byte] = {
+    val baos = new ByteArrayOutputStream()
+    doc.save(baos)
+    Stream.chunk(Chunk.array(baos.toByteArray))
+  }
+}

modules/convert/src/test/scala/docspell/convert/ConversionTest.scala

@@ -74,7 +74,8 @@ class ConversionTest extends FunSuite with FileChecks {
         Duration.seconds(20)
       ),
       target
-    )
+    ),
+    ConvertConfig.DecryptPdf(true, Nil)
   )
 
   val conversion =

modules/convert/src/test/scala/docspell/convert/FileChecks.scala

@@ -9,6 +9,8 @@ package docspell.convert
 import java.nio.charset.StandardCharsets
 import java.nio.file.Files
 
+import scala.util.Try
+
 import cats.data.Kleisli
 import cats.effect.IO
 import cats.effect.unsafe.implicits.global
@@ -19,6 +21,9 @@ import docspell.common._
 import docspell.convert.ConversionResult.Handler
 import docspell.files.TikaMimetype
 
+import org.apache.pdfbox.pdmodel.PDDocument
+import org.apache.pdfbox.pdmodel.encryption.InvalidPasswordException
+
 trait FileChecks {
 
   implicit class FileCheckOps(p: Path) {
@@ -34,15 +39,46 @@ trait FileChecks {
 
     def isPlainText: Boolean =
       isType(MimeType.text("plain"))
+
+    def isUnencryptedPDF: Boolean =
+      Try(PDDocument.load(p.toNioPath.toFile)).map(_.close()).isSuccess
+  }
+
+  implicit class ByteStreamOps(delegate: Stream[IO, Byte]) {
+    def isNonEmpty: IO[Boolean] =
+      delegate.head.compile.last.map(_.isDefined)
+
+    def isType(mime: MimeType): IO[Boolean] =
+      TikaMimetype.detect(delegate, MimeTypeHint.none).map(_ == mime)
+
+    def isPDF: IO[Boolean] =
+      isType(MimeType.pdf)
+
+    def isUnencryptedPDF: IO[Boolean] =
+      delegate.compile
+        .to(Array)
+        .map(PDDocument.load(_))
+        .map(_.close())
+        .map(_ => true)
+
+    def isEncryptedPDF: IO[Boolean] =
+      delegate.compile
+        .to(Array)
+        .map(PDDocument.load(_))
+        .attempt
+        .map(e =>
+          e.fold(
+            _.isInstanceOf[InvalidPasswordException],
+            doc => {
+              doc.close();
+              false
+            }
+          )
+        )
   }
 
   def storeFile(file: Path): Pipe[IO, Byte, Path] =
-    in =>
+    fs2.io.file.Files[IO].writeAll(file).andThen(s => s ++ Stream.emit(file))
-      Stream
-        .eval(
-          in.compile.to(Array).flatMap(bytes => IO(Files.write(file.toNioPath, bytes)))
-        )
-        .map(p => File.path(p))
 
   def storePdfHandler(file: Path): Handler[IO, Path] =
     storePdfTxtHandler(file, file.resolveSibling("unexpected.txt")).map(_._1)

modules/convert/src/test/scala/docspell/convert/RemovePdfEncryptionTest.scala

@@ -0,0 +1,75 @@
+/*
+ * Copyright 2020 Eike K. & Contributors
+ *
+ * SPDX-License-Identifier: AGPL-3.0-or-later
+ */
+
+package docspell.convert
+
+import cats.effect.IO
+import fs2.Stream
+
+import docspell.common.Logger
+import docspell.files.ExampleFiles
+
+import munit.CatsEffectSuite
+
+class RemovePdfEncryptionTest extends CatsEffectSuite with FileChecks {
+  val logger: Logger[IO] = Logger.log4s(org.log4s.getLogger)
+
+  val protectedPdf = ExampleFiles.secured_protected_test123_pdf.readURL[IO](16 * 1024)
+  val encryptedPdf = ExampleFiles.secured_encrypted_test123_pdf.readURL[IO](16 * 1024)
+  val plainPdf = ExampleFiles.letter_en_pdf.readURL[IO](16 * 1024)
+
+  test("have encrypted pdfs") {
+    for {
+      _ <- assertIO(encryptedPdf.isEncryptedPDF, true)
+      _ <- assertIO(encryptedPdf.isEncryptedPDF, true)
+    } yield ()
+  }
+
+  test("decrypt pdf") {
+    encryptedPdf
+      .through(RemovePdfEncryption(logger, List("test123")))
+      .isUnencryptedPDF
+      .map(assert(_))
+  }
+
+  test("decrypt pdf with multiple passwords") {
+    encryptedPdf
+      .through(RemovePdfEncryption(logger, List("xy123", "123xy", "test123", "abc123")))
+      .isUnencryptedPDF
+      .map(assert(_))
+  }
+
+  test("remove protection") {
+    protectedPdf
+      .through(RemovePdfEncryption(logger, Nil))
+      .isUnencryptedPDF
+      .map(assert(_))
+  }
+
+  test("read unprotected pdf") {
+    plainPdf
+      .through(RemovePdfEncryption(logger, Nil))
+      .isUnencryptedPDF
+      .map(assert(_))
+  }
+
+  test("decrypt with multiple passwords, stop on first") {
+    val passwords: Stream[IO, String] =
+      Stream("test123") ++ Stream.raiseError[IO](new Exception("is not called"))
+    val decrypt = RemovePdfEncryption(logger, passwords)
+    encryptedPdf
+      .through(decrypt)
+      .isUnencryptedPDF
+      .map(assert(_))
+  }
+
+  test("return input stream if nothing helps") {
+    encryptedPdf
+      .through(RemovePdfEncryption(logger, List("a", "b")))
+      .isEncryptedPDF
+      .map(assert(_))
+  }
+}

modules/joex/src/main/resources/reference.conf

@@ -586,6 +586,21 @@ Docpell Update Check
       }
       working-dir = ${java.io.tmpdir}"/docspell-convert"
     }
+
+    # Allows to try to decrypt a PDF with encryption or protection. If
+    # enabled, a PDFs encryption or protection will be removed during
+    # conversion.
+    #
+    # For encrypted PDFs, this is necessary to be processed, because
+    # docspell needs to read it. It also requires to specify a
+    # password here. All passwords are tried when reading a PDF.
+    #
+    # This is enabled by default, using an empty password list. This
+    # removes protection from PDFs, which is better for processing.
+    decrypt-pdf = {
+      enabled = true
+      passwords = []
+    }
   }
 
   # The same section is also present in the rest-server config. It is