Use keywords in pdfs to search for existing tags

During processing, keywords stored in PDF metadata are used to look
them up in the tag database and associate any existing tags to the
item.

See #175

modules/extract/src/main/scala/docspell/extract/ExtractResult.scala

@@ -1,39 +1,41 @@
 package docspell.extract
 
-import scala.util.Try
-
 import docspell.common.MimeType
+import docspell.extract.pdfbox.PdfMetaData
 
 sealed trait ExtractResult {
 
   def textOption: Option[String]
 
+  def pdfMeta: Option[PdfMetaData]
 }
 
 object ExtractResult {
 
   case class UnsupportedFormat(mime: MimeType) extends ExtractResult {
     val textOption = None
+    val pdfMeta    = None
   }
   def unsupportedFormat(mt: MimeType): ExtractResult =
     UnsupportedFormat(mt)
 
   case class Failure(ex: Throwable) extends ExtractResult {
     val textOption = None
+    val pdfMeta    = None
   }
   def failure(ex: Throwable): ExtractResult =
     Failure(ex)
 
-  case class Success(text: String) extends ExtractResult {
+  case class Success(text: String, pdfMeta: Option[PdfMetaData]) extends ExtractResult {
     val textOption = Some(text)
   }
-  def success(text: String): ExtractResult =
-    Success(text)
-
-  def fromTry(r: Try[String]): ExtractResult =
-    r.fold(Failure.apply, Success.apply)
+  def success(text: String, pdfMeta: Option[PdfMetaData]): ExtractResult =
+    Success(text, pdfMeta)
 
   def fromEither(e: Either[Throwable, String]): ExtractResult =
-    e.fold(failure, success)
+    e.fold(failure, str => success(str, None))
+
+  def fromEitherResult(e: Either[Throwable, PdfExtract.Result]): ExtractResult =
+    e.fold(failure, r => success(r.txt.value, r.meta))
 
 }

modules/extract/src/main/scala/docspell/extract/Extraction.scala

@@ -40,8 +40,7 @@ object Extraction {
           case MimeType.PdfMatch(_) =>
             PdfExtract
               .get(data, blocker, lang, cfg.pdf.minTextLen, cfg.ocr, logger)
-              .map(_.map(_.value))
-              .map(ExtractResult.fromEither)
+              .map(ExtractResult.fromEitherResult)
 
           case PoiType(mt) =>
             PoiExtract
@@ -103,7 +102,7 @@ object Extraction {
             val cs = mt.charsetOrUtf8
             logger.info(s"File detected as ${mt.asString}. Returning itself as text.") *>
               data.through(Binary.decode(cs)).foldMonoid.compile.last.map { txt =>
-                ExtractResult.success(Text(txt).value)
+                ExtractResult.success(Text(txt).value, None)
               }
 
           case mt =>

modules/extract/src/main/scala/docspell/extract/PdfExtract.scala

@@ -7,9 +7,15 @@ import fs2.Stream
 import docspell.common.{Language, Logger}
 import docspell.extract.internal.Text
 import docspell.extract.ocr.{OcrConfig, TextExtract}
+import docspell.extract.pdfbox.PdfMetaData
 import docspell.extract.pdfbox.PdfboxExtract
 
 object PdfExtract {
+  final case class Result(txt: Text, meta: Option[PdfMetaData])
+  object Result {
+    def apply(t: (Text, Option[PdfMetaData])): Result =
+      Result(t._1, t._2)
+  }
 
   def get[F[_]: Sync: ContextShift](
       in: Stream[F, Byte],
@@ -18,39 +24,39 @@ object PdfExtract {
       stripMinLen: Int,
       ocrCfg: OcrConfig,
       logger: Logger[F]
-  ): F[Either[Throwable, Text]] = {
+  ): F[Either[Throwable, Result]] = {
 
     val runOcr =
       TextExtract.extractOCR(in, blocker, logger, lang.iso3, ocrCfg).compile.lastOrError
 
-    def chooseResult(ocrStr: Text, strippedStr: Text) =
-      if (ocrStr.length > strippedStr.length)
+    def chooseResult(ocrStr: Text, strippedRes: (Text, Option[PdfMetaData])) =
+      if (ocrStr.length > strippedRes._1.length)
         logger.info(
-          s"Using OCR text, as it is longer (${ocrStr.length} > ${strippedStr.length})"
-        ) *> ocrStr.pure[F]
+          s"Using OCR text, as it is longer (${ocrStr.length} > ${strippedRes._1.length})"
+        ) *> Result(ocrStr, strippedRes._2).pure[F]
       else
         logger.info(
-          s"Using stripped text (not OCR), as it is longer (${strippedStr.length} > ${ocrStr.length})"
-        ) *> strippedStr.pure[F]
+          s"Using stripped text (not OCR), as it is longer (${strippedRes._1.length} > ${ocrStr.length})"
+        ) *> Result(strippedRes).pure[F]
 
     //maybe better: inspect the pdf and decide whether ocr or not
     for {
       pdfboxRes <-
-        logger.debug("Trying to strip text from pdf using pdfbox.") *> PdfboxExtract
-          .getText[F](in)
+        logger.debug("Trying to strip text from pdf using pdfbox.") *>
+          PdfboxExtract.getTextAndMetaData[F](in)
       res <- pdfboxRes.fold(
         ex =>
           logger.info(
             s"Stripping text from PDF resulted in an error: ${ex.getMessage}. Trying with OCR. "
-          ) >> runOcr.attempt,
-        str =>
-          if (str.length >= stripMinLen) str.pure[F].attempt
+          ) >> runOcr.map(txt => Result(txt, None)).attempt,
+        pair =>
+          if (pair._1.length >= stripMinLen) Result(pair).pure[F].attempt
           else
             logger
               .info(
-                s"Stripped text from PDF is small (${str.length}). Trying with OCR."
+                s"Stripped text from PDF is small (${pair._1.length}). Trying with OCR."
               ) *>
-              runOcr.flatMap(ocrStr => chooseResult(ocrStr, str)).attempt
+              runOcr.flatMap(ocrStr => chooseResult(ocrStr, pair)).attempt
       )
     } yield res
   }

modules/extract/src/main/scala/docspell/extract/pdfbox/PdfboxExtract.scala

@@ -9,17 +9,17 @@ import cats.effect.Sync
 import cats.implicits._
 import fs2.Stream
 
+import docspell.common.Timestamp
 import docspell.extract.internal.Text
 
 import org.apache.pdfbox.pdmodel.PDDocument
 import org.apache.pdfbox.text.PDFTextStripper
-import docspell.common.Timestamp
 
 object PdfboxExtract {
 
   def getTextAndMetaData[F[_]: Sync](
       data: Stream[F, Byte]
-  ): F[Either[Throwable, (Text, PdfMetaData)]] =
+  ): F[Either[Throwable, (Text, Option[PdfMetaData])]] =
     data.compile
       .to(Array)
       .map(bytes =>
@@ -27,7 +27,7 @@ object PdfboxExtract {
           for {
             txt <- readText(doc)
             md  <- readMetaData(doc)
-          } yield (txt, md)
+          } yield (txt, Some(md).filter(_.nonEmpty))
         }.toEither.flatten
       )
 

modules/extract/src/test/scala/docspell/extract/pdfbox/PdfboxExtractTest.scala

@@ -47,7 +47,7 @@ object PdfboxExtractTest extends SimpleTestSuite {
     val url = ExampleFiles.keywords_pdf.toJavaUrl.fold(sys.error, identity)
     val str = PdfboxExtract.getText(url.openStream()).fold(throw _, identity)
     assert(str.value.startsWith("Keywords in PDF"))
-    val md =  PdfboxExtract.getMetaData(url.openStream()).fold(throw _, identity)
+    val md = PdfboxExtract.getMetaData(url.openStream()).fold(throw _, identity)
     assertEquals(md.author, Some("E.K."))
     assertEquals(md.title, Some("Keywords in PDF"))
     assertEquals(md.subject, Some("This is a subject"))