Skip to content

Commit

Permalink
Streamline extern-conv stdin/infile
Browse files Browse the repository at this point in the history
  • Loading branch information
eikek committed Feb 18, 2020
1 parent 0dcc008 commit 5869e2e
Show file tree
Hide file tree
Showing 6 changed files with 67 additions and 92 deletions.
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,13 @@ object SystemCommand {
def mapArgs(f: String => String): Config =
Config(program, args.map(f), timeout)

def replace(repl: Map[String, String]): Config =
mapArgs(s =>
repl.foldLeft(s) {
case (res, (k, v)) =>
res.replace(k, v)
})

def toCmd: List[String] =
program :: args.toList

Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -13,50 +13,28 @@ object ExternConv {
cmdCfg: SystemCommand.Config,
wd: Path,
chunkSize: Int,
useStdin: Boolean,
blocker: Blocker,
logger: Logger[F]
): Pipe[F, Byte, Byte] =
in =>
Stream.resource(File.withTempDir[F](wd, s"docspell-$name")).flatMap { dir =>
val out = dir.resolve("out.pdf")
val inFile = dir.resolve("infile").toAbsolutePath.normalize
val out = dir.resolve("out.pdf").toAbsolutePath.normalize
val sysCfg =
cmdCfg.mapArgs(_.replace("{{outfile}}", out.toAbsolutePath.normalize.toString))

SystemCommand
.execSuccess[F](sysCfg, blocker, logger, Some(dir), in)
.flatMap(result =>
logResult(name, result, logger) ++ readResult[F](
out,
result,
blocker,
chunkSize,
logger
)
cmdCfg.replace(
Map("{{outfile}}" -> out.toString) ++
(if (!useStdin) Map("{{infile}}" -> inFile.toString)
else Map.empty)
)
}

def toPDFviaFile[F[_]: Sync: ContextShift](
name: String,
cmdCfg: SystemCommand.Config,
wd: Path,
chunkSize: Int,
blocker: Blocker,
logger: Logger[F]
): Pipe[F, Byte, Byte] =
in =>
Stream.resource(File.withTempDir[F](wd, s"docspell-$name")).flatMap { dir =>
val inFile = dir.resolve("infile")
val out = dir.resolve("out.pdf")
val sysCfg =
cmdCfg.mapArgs(
_.replace("{{outfile}}", out.toAbsolutePath.normalize.toString)
.replace("{{infile}}", inFile.toAbsolutePath.normalize.toString)
)
val createInput: Pipe[F, Byte, Unit] =
if (useStdin) _ => Stream.emit(())
else storeDataToFile(name, blocker, logger, inFile)

(Stream.eval(logger.debug(s"Storing input to file ${inFile} for running $name")).drain ++
Stream.eval(storeFile(in, inFile, blocker))).flatMap { _ =>
in.through(createInput).flatMap { _ =>
SystemCommand
.execSuccess[F](sysCfg, blocker, logger, Some(dir))
.execSuccess[F](sysCfg, blocker, logger, Some(dir), if (useStdin) in else Stream.empty)
.flatMap(result =>
logResult(name, result, logger) ++ readResult[F](
out,
Expand All @@ -69,7 +47,7 @@ object ExternConv {
}
}

private def readResult[F[_]: Sync: ContextShift](
def readResult[F[_]: Sync: ContextShift](
out: Path,
result: SystemCommand.Result,
blocker: Blocker,
Expand All @@ -91,6 +69,11 @@ object ExternConv {
)
}

private def storeDataToFile[F[_]: Sync: ContextShift](name: String, blocker: Blocker, logger: Logger[F], inFile: Path): Pipe[F, Byte, Unit] =
in =>
Stream.eval(logger.debug(s"Storing input to file ${inFile} for running $name")).drain ++
Stream.eval(storeFile(in, inFile, blocker))

private def logResult[F[_]: Sync](
name: String,
result: SystemCommand.Result,
Expand Down
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
package docspell.convert.extern

object Tesseract {

}
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,7 @@ object Unoconv {
blocker: Blocker,
logger: Logger[F],
): Pipe[F, Byte, Byte] =
ExternConv.toPDFviaFile[F]("unoconv", cfg.cmd, cfg.workingDir, chunkSize, blocker, logger)
ExternConv.toPDF[F]("unoconv", cfg.cmd, cfg.workingDir, chunkSize, false, blocker, logger)


}
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,7 @@ object WkHtmlPdf {
blocker: Blocker,
logger: Logger[F],
): Pipe[F, Byte, Byte] =
ExternConv.toPDF[F]("wkhtmltopdf", cfg.cmd, cfg.workingDir, chunkSize, blocker, logger)
ExternConv.toPDF[F]("wkhtmltopdf", cfg.cmd, cfg.workingDir, chunkSize, true, blocker, logger)


}
90 changes: 35 additions & 55 deletions modules/extract/src/main/scala/docspell/extract/ocr/Ocr.scala
Original file line number Diff line number Diff line change
Expand Up @@ -19,12 +19,10 @@ object Ocr {
): F[Option[String]] =
File.withTempDir(config.ghostscript.workingDir, "extractpdf").use { wd =>
runGhostscript(pdf, config, wd, blocker, logger)
.flatMap({ tmpImg =>
runTesseractFile(tmpImg, blocker, logger, lang, config)
})
.fold1(_ + "\n\n\n" + _).
compile.
last
.flatMap(tmpImg => runTesseractFile(tmpImg, blocker, logger, lang, config))
.fold1(_ + "\n\n\n" + _)
.compile
.last
}

/** Extract the text from the given image file
Expand All @@ -47,12 +45,10 @@ object Ocr {
): F[Option[String]] =
File.withTempDir(config.ghostscript.workingDir, "extractpdf").use { wd =>
runGhostscriptFile(pdf, config.ghostscript.command, wd, blocker, logger)
.flatMap({ tif =>
runTesseractFile(tif, blocker, logger, lang, config)
})
.fold1(_ + "\n\n\n" + _).
compile.
last
.flatMap(tif => runTesseractFile(tif, blocker, logger, lang, config))
.fold1(_ + "\n\n\n" + _)
.compile
.last
}

def extractImageFile[F[_]: Sync: ContextShift](
Expand All @@ -68,31 +64,27 @@ object Ocr {
* files are stored to a temporary location on disk and returned.
*/
private[extract] def runGhostscript[F[_]: Sync: ContextShift](
pdf: Stream[F, Byte],
cfg: OcrConfig,
wd: Path,
blocker: Blocker,
logger: Logger[F]
pdf: Stream[F, Byte],
cfg: OcrConfig,
wd: Path,
blocker: Blocker,
logger: Logger[F]
): Stream[F, Path] = {
val xargs =
if (cfg.pageRange.begin > 0)
s"-dLastPage=${cfg.pageRange.begin}" +: cfg.ghostscript.command.args
else cfg.ghostscript.command.args
val cmd = cfg.ghostscript.command
.copy(args = xargs)
.mapArgs(
replace(
Map(
"{{infile}}" -> "-",
"{{outfile}}" -> "%d.tif"
)
.replace(
Map(
"{{infile}}" -> "-",
"{{outfile}}" -> "%d.tif"
)
)
SystemCommand
.execSuccess(cmd, blocker, logger, wd = Some(wd), stdin = pdf)
.evalMap({ _ =>
File.listFiles(pathEndsWith(".tif"), wd)
})
.evalMap(_ => File.listFiles(pathEndsWith(".tif"), wd))
.flatMap(fs => Stream.emits(fs))
}

Expand All @@ -106,19 +98,15 @@ object Ocr {
blocker: Blocker,
logger: Logger[F]
): Stream[F, Path] = {
val cmd = ghostscript.mapArgs(
replace(
Map(
"{{infile}}" -> pdf.toAbsolutePath.toString,
"{{outfile}}" -> "%d.tif"
)
val cmd = ghostscript.replace(
Map(
"{{infile}}" -> pdf.toAbsolutePath.toString,
"{{outfile}}" -> "%d.tif"
)
)
SystemCommand
.execSuccess[F](cmd, blocker, logger, wd = Some(wd))
.evalMap({ _ =>
File.listFiles(pathEndsWith(".tif"), wd)
})
.evalMap(_ => File.listFiles(pathEndsWith(".tif"), wd))
.flatMap(fs => Stream.emits(fs))
}

Expand All @@ -136,20 +124,20 @@ object Ocr {
logger: Logger[F]
): Stream[F, Path] = {
val targetFile = img.resolveSibling("u-" + img.getFileName.toString).toAbsolutePath
val cmd = unpaper.mapArgs(
replace(
Map(
"{{infile}}" -> img.toAbsolutePath.toString,
"{{outfile}}" -> targetFile.toString
)
val cmd = unpaper.replace(
Map(
"{{infile}}" -> img.toAbsolutePath.toString,
"{{outfile}}" -> targetFile.toString
)
)
SystemCommand.execSuccess[F](cmd, blocker, logger, wd = Some(wd)).map(_ => targetFile).handleErrorWith {
th =>
SystemCommand
.execSuccess[F](cmd, blocker, logger, wd = Some(wd))
.map(_ => targetFile)
.handleErrorWith { th =>
logger
.warn(s"Unpaper command failed: ${th.getMessage}. Using input file for text extraction.")
Stream.emit(img)
}
}
}

/** Run tesseract on the given image file and return the extracted
Expand All @@ -165,9 +153,8 @@ object Ocr {
// tesseract cannot cope with absolute filenames
// so use the parent as working dir
runUnpaperFile(img, config.unpaper.command, img.getParent, blocker, logger).flatMap { uimg =>
val cmd = config.tesseract.command.mapArgs(
replace(Map("{{file}}" -> uimg.getFileName.toString, "{{lang}}" -> fixLanguage(lang)))
)
val cmd = config.tesseract.command
.replace(Map("{{file}}" -> uimg.getFileName.toString, "{{lang}}" -> fixLanguage(lang)))
SystemCommand.execSuccess[F](cmd, blocker, logger, wd = Some(uimg.getParent)).map(_.stdout)
}

Expand All @@ -182,17 +169,10 @@ object Ocr {
config: OcrConfig
): Stream[F, String] = {
val cmd = config.tesseract.command
.mapArgs(replace(Map("{{file}}" -> "stdin", "{{lang}}" -> fixLanguage(lang))))
.replace(Map("{{file}}" -> "stdin", "{{lang}}" -> fixLanguage(lang)))
SystemCommand.execSuccess(cmd, blocker, logger, stdin = img).map(_.stdout)
}

private def replace(repl: Map[String, String]): String => String =
s =>
repl.foldLeft(s) {
case (res, (k, v)) =>
res.replace(k, v)
}

private def fixLanguage(lang: String): String =
lang match {
case "de" => "deu"
Expand Down

0 comments on commit 5869e2e

Please sign in to comment.