Add an inspect command that gathers metrics about a Jelly file

build.sbt

@@ -38,6 +38,7 @@ lazy val root = (project in file("."))
       "eu.ostrzyciel.jelly" %% "jelly-jena" % jellyV,
       "com.github.alexarchambault" %% "case-app" % "2.1.0-M30",
       "org.scalatest" %% "scalatest" % "3.2.19" % Test,
+      "org.yaml" % "snakeyaml" % "2.4" % Test,
     ),
     scalacOptions ++= Seq(
       "-Wunused:imports",

src/main/scala/eu/neverblink/jelly/cli/App.scala

@@ -22,4 +22,5 @@ object App extends CommandsEntryPoint:
     Version,
     RdfFromJelly,
     RdfToJelly,
+    RdfInspect,
   )

src/main/scala/eu/neverblink/jelly/cli/Exceptions.scala

@@ -1,6 +1,5 @@
 package eu.neverblink.jelly.cli
 
-import com.google.protobuf.InvalidProtocolBufferException
 import org.apache.jena.riot.RiotException
 
 /** Contains a set of common jelly-cli exceptions with custom output messages.
@@ -22,7 +21,7 @@ case class JellyTranscodingError(message: String)
     extends CriticalException(s"Jelly transcoding error: $message")
 case class JenaRiotException(e: RiotException)
     extends CriticalException(s"Jena RDF I/O exception: ${e.getMessage}")
-case class InvalidJellyFile(e: InvalidProtocolBufferException)
+case class InvalidJellyFile(e: Exception)
     extends CriticalException(s"Invalid Jelly file: ${e.getMessage}")
 case class InvalidFormatSpecified(format: String, validFormats: String)
     extends CriticalException(

src/main/scala/eu/neverblink/jelly/cli/command/rdf/RdfFromJelly.scala

@@ -3,9 +3,9 @@ import caseapp.*
 import eu.neverblink.jelly.cli.*
 import eu.neverblink.jelly.cli.command.rdf.RdfFormat.*
 import eu.neverblink.jelly.cli.command.rdf.RdfFormat.Jena.*
+import eu.neverblink.jelly.cli.util.JellyUtil
 import eu.ostrzyciel.jelly.convert.jena.riot.JellyLanguage
 import eu.ostrzyciel.jelly.core.proto.v1.RdfStreamFrame
-import eu.ostrzyciel.jelly.core.IoUtils
 import org.apache.jena.riot.system.StreamRDFWriter
 import org.apache.jena.riot.{Lang, RDFParser}
 
@@ -25,7 +25,7 @@ case class RdfFromJellyOptions(
     @ExtraName("out-format") outputFormat: Option[String] = None,
 ) extends HasJellyCommandOptions
 
-object RdfFromJelly extends RdfCommand[RdfFromJellyOptions, RdfFormat.Writeable]:
+object RdfFromJelly extends RdfTranscodeCommand[RdfFromJellyOptions, RdfFormat.Writeable]:
 
   override def names: List[List[String]] = List(
     List("rdf", "from-jelly"),
@@ -83,30 +83,10 @@ object RdfFromJelly extends RdfCommand[RdfFromJellyOptions, RdfFormat.Writeable]
       outputStream.write(frame.getBytes)
 
     try {
-      iterateRdfStream(inputStream, outputStream).zipWithIndex.foreach {
+      JellyUtil.iterateRdfStream(inputStream).zipWithIndex.foreach {
         case (maybeFrame, frameIndex) =>
           writeFrameToOutput(maybeFrame, frameIndex)
       }
     } finally {
       outputStream.flush()
     }
-
-  /** This method reads the Jelly file and returns an iterator of RdfStreamFrame
-    * @param inputStream
-    * @param outputStream
-    * @return
-    */
-  private def iterateRdfStream(
-      inputStream: InputStream,
-      outputStream: OutputStream,
-  ): Iterator[RdfStreamFrame] =
-    IoUtils.autodetectDelimiting(inputStream) match
-      case (false, newIn) =>
-        // Non-delimited Jelly file
-        // In this case, we can only read one frame
-        Iterator(RdfStreamFrame.parseFrom(newIn))
-      case (true, newIn) =>
-        // Delimited Jelly file
-        // In this case, we can read multiple frames
-        Iterator.continually(RdfStreamFrame.parseDelimitedFrom(newIn))
-          .takeWhile(_.isDefined).map(_.get)

src/main/scala/eu/neverblink/jelly/cli/command/rdf/RdfInspect.scala

@@ -0,0 +1,93 @@
+package eu.neverblink.jelly.cli.command.rdf
+
+import caseapp.{ExtraName, Recurse}
+import caseapp.core.RemainingArgs
+import eu.neverblink.jelly.cli.util.{FrameInfo, JellyUtil, MetricsPrinter}
+import eu.neverblink.jelly.cli.*
+import eu.ostrzyciel.jelly.core.proto.v1.*
+
+import java.io.InputStream
+
+case class RdfInspectOptions(
+    @Recurse
+    common: JellyCommandOptions = JellyCommandOptions(),
+    @ExtraName("to") outputFile: Option[String] = None,
+    @ExtraName("per-frame") perFrame: Boolean = false,
+) extends HasJellyCommandOptions
+
+object RdfInspect extends JellyCommand[RdfInspectOptions]:
+
+  override def names: List[List[String]] = List(
+    List("rdf", "inspect"),
+  )
+
+  override final def group = "rdf"
+
+  override def doRun(options: RdfInspectOptions, remainingArgs: RemainingArgs): Unit =
+    val (inputStream, outputStream) =
+      this.getIoStreamsFromOptions(remainingArgs.remaining.headOption, options.outputFile)
+    val printer = inspectJelly(inputStream)
+    if options.perFrame then printer.printPerFrame(outputStream)
+    else printer.printAggregate(outputStream)
+
+  private def inspectJelly(
+      inputStream: InputStream,
+  ): MetricsPrinter =
+
+    inline def computeMetrics(
+        frame: RdfStreamFrame,
+        frameIndex: Int,
+        printer: MetricsPrinter,
+    ): Unit =
+      val metrics = new FrameInfo()
+      frame.rows.foreach(r => metricsForRow(r, metrics))
+      printer.frameInfo += metrics
+
+    try {
+      val allRows = JellyUtil.iterateRdfStream(inputStream).toList
+      // we need to check if the first frame contains options
+      val streamOptions = checkOptions(allRows)
+      val printer = new MetricsPrinter(streamOptions)
+      // We compute the metrics for each frame
+      // and then sum them all during the printing if desired
+      allRows.zipWithIndex.foreach { case (maybeFrame, frameIndex) =>
+        computeMetrics(maybeFrame, frameIndex, printer)
+      }
+      printer
+    } catch {
+      case e: Exception =>
+        throw InvalidJellyFile(e)
+    }
+
+  private def metricsForRow(
+      row: RdfStreamRow,
+      metadata: FrameInfo,
+  ): Unit =
+    row.row match {
+      case r: RdfTriple => metadata.tripleCount += 1
+      case r: RdfQuad => metadata.quadCount += 1
+      case r: RdfNameEntry => metadata.nameCount += 1
+      case r: RdfPrefixEntry => metadata.prefixCount += 1
+      case r: RdfNamespaceDeclaration => metadata.namespaceCount += 1
+      case r: RdfDatatypeEntry => metadata.datatypeCount += 1
+      case r: RdfGraphStart => metadata.graphStartCount += 1
+      case r: RdfGraphEnd => metadata.graphEndCount += 1
+      case r: RdfStreamOptions => metadata.optionCount += 1
+    }
+
+  /** Checks whether the first frame in the stream contains options and returns them.
+    * @param allFrames
+    *   The list of all frames in the stream.
+    * @return
+    *   The options from the first frame.
+    * @throws RuntimeException
+    *   If the first frame does not contain options or if there are no frames in the stream.
+    */
+  private def checkOptions(allFrames: List[RdfStreamFrame]): RdfStreamOptions =
+    if allFrames.isEmpty then throw new RuntimeException("No frames in the stream.")
+    if allFrames.head.rows.isEmpty then throw new RuntimeException("No rows in the frame.")
+    val frameRows = allFrames.head.rows
+    frameRows.head.row match {
+      case r: RdfStreamOptions => r
+      case _ => throw new RuntimeException("First row of the frame is not an options row.")
+    }

src/main/scala/eu/neverblink/jelly/cli/command/rdf/RdfToJelly.scala

@@ -40,7 +40,7 @@ case class RdfToJellyOptions(
     delimited: Boolean = true,
 ) extends HasJellyCommandOptions
 
-object RdfToJelly extends RdfCommand[RdfToJellyOptions, RdfFormat.Readable]:
+object RdfToJelly extends RdfTranscodeCommand[RdfToJellyOptions, RdfFormat.Readable]:
 
   override def names: List[List[String]] = List(
     List("rdf", "to-jelly"),

src/main/scala/eu/neverblink/jelly/cli/command/rdf/RdfCommand.scala → src/main/scala/eu/neverblink/jelly/cli/command/rdf/RdfTranscodeCommand.scala

@@ -12,8 +12,8 @@ import java.io.{InputStream, OutputStream}
 
 /** This abstract class is responsible for the common logic in both RDF parsing commands
   */
-abstract class RdfCommand[T <: HasJellyCommandOptions: {Parser, Help}, F <: RdfFormat](using
-    tt: TypeTest[RdfFormat, F],
+abstract class RdfTranscodeCommand[T <: HasJellyCommandOptions: {Parser, Help}, F <: RdfFormat](
+    using tt: TypeTest[RdfFormat, F],
 ) extends JellyCommand[T]:
 
   override final def group = "rdf"

src/main/scala/eu/neverblink/jelly/cli/util/JellyUtil.scala

@@ -0,0 +1,27 @@
+package eu.neverblink.jelly.cli.util
+
+import eu.ostrzyciel.jelly.core.IoUtils
+import eu.ostrzyciel.jelly.core.proto.v1.RdfStreamFrame
+
+import java.io.InputStream
+
+object JellyUtil:
+  /** This method reads the Jelly file and returns an iterator of RdfStreamFrame
+    *
+    * @param inputStream
+    * @param outputStream
+    * @return
+    */
+  def iterateRdfStream(
+      inputStream: InputStream,
+  ): Iterator[RdfStreamFrame] =
+    IoUtils.autodetectDelimiting(inputStream) match
+      case (false, newIn) =>
+        // Non-delimited Jelly file
+        // In this case, we can only read one frame
+        Iterator(RdfStreamFrame.parseFrom(newIn))
+      case (true, newIn) =>
+        // Delimited Jelly file
+        // In this case, we can read multiple frames
+        Iterator.continually(RdfStreamFrame.parseDelimitedFrom(newIn))
+          .takeWhile(_.isDefined).map(_.get)

src/main/scala/eu/neverblink/jelly/cli/util/MetricsPrinter.scala

@@ -0,0 +1,110 @@
+package eu.neverblink.jelly.cli.util
+
+import eu.neverblink.jelly.cli.util.MetricsPrinter.{formatOptions, formatStats}
+import eu.neverblink.jelly.cli.util.YamlDocBuilder.YamlMap
+import eu.ostrzyciel.jelly.core.proto.v1.RdfStreamOptions
+
+import java.io.OutputStream
+import scala.collection.mutable.ListBuffer
+
+/** This class is used to store the metrics for a single frame
+  */
+class FrameInfo:
+  var optionCount: Int = 0
+  var nameCount: Int = 0
+  var namespaceCount: Int = 0
+  var tripleCount: Int = 0
+  var quadCount: Int = 0
+  var prefixCount: Int = 0
+  var datatypeCount: Int = 0
+  var graphStartCount: Int = 0
+  var graphEndCount: Int = 0
+
+  def +=(other: FrameInfo): FrameInfo = {
+    this.optionCount += other.optionCount
+    this.nameCount += other.nameCount
+    this.namespaceCount += other.namespaceCount
+    this.tripleCount += other.tripleCount
+    this.quadCount += other.quadCount
+    this.prefixCount += other.prefixCount
+    this.datatypeCount += other.datatypeCount
+    this.graphStartCount += other.graphStartCount
+    this.graphEndCount += other.graphEndCount
+    this
+  }
+
+end FrameInfo
+
+class MetricsPrinter(printOptions: RdfStreamOptions):
+
+  var frameInfo: ListBuffer[FrameInfo] = ListBuffer.empty
+
+  def printPerFrame(o: OutputStream): Unit = {
+    val options = formatOptions(options = printOptions)
+    val yamlFrames = YamlDocBuilder.YamlList(frameInfo.map { frame =>
+      formatStats(frame)
+    }.toSeq)
+    val fullString =
+      YamlDocBuilder.build(
+        YamlMap(
+          "stream_options" -> options,
+          "frames" -> yamlFrames,
+        ),
+      )
+    o.write(fullString.getBytes)
+
+  }
+
+  def printAggregate(o: OutputStream): Unit = {
+    val sumCounts = frameInfo.reduce(_ += _)
+    val options = formatOptions(options = printOptions)
+    val fullString =
+      YamlDocBuilder.build(
+        YamlMap(
+          "stream_options" -> options,
+          "frames" -> formatStats(sumCounts),
+        ),
+      )
+    o.write(fullString.getBytes)
+  }
+
+end MetricsPrinter
+
+object MetricsPrinter:
+
+  /** This method converts a boolean to an integer
+    */
+  private def boolToInt(b: Boolean): Int =
+    if b then 1 else 0
+
+  def formatOptions(
+      options: RdfStreamOptions,
+  ): YamlDocBuilder.YamlMap =
+    YamlDocBuilder.YamlMap(
+      "rdf_star" -> YamlDocBuilder.YamlInt(boolToInt(options.rdfStar)),
+      "stream_name" -> YamlDocBuilder.YamlString(options.streamName),
+      "generalized_statements" -> YamlDocBuilder.YamlInt(boolToInt(options.generalizedStatements)),
+      "version" -> YamlDocBuilder.YamlInt(options.version),
+      "max_datatype_table_size" -> YamlDocBuilder.YamlInt(options.maxDatatypeTableSize),
+      "max_name_table_size" -> YamlDocBuilder.YamlInt(options.maxNameTableSize),
+      "max_prefix_table_size" -> YamlDocBuilder.YamlInt(options.maxPrefixTableSize),
+      "logical_type" -> YamlDocBuilder.YamlInt(options.logicalType.value),
+      "physical_type" -> YamlDocBuilder.YamlInt(options.physicalType.value),
+    )
+
+  def formatStats(
+      frame: FrameInfo,
+  ): YamlDocBuilder.YamlMap =
+    YamlDocBuilder.YamlMap(
+      "option_count" -> YamlDocBuilder.YamlInt(frame.optionCount),
+      "name_count" -> YamlDocBuilder.YamlInt(frame.nameCount),
+      "namespace_count" -> YamlDocBuilder.YamlInt(frame.namespaceCount),
+      "triple_count" -> YamlDocBuilder.YamlInt(frame.tripleCount),
+      "quad_count" -> YamlDocBuilder.YamlInt(frame.quadCount),
+      "prefix_count" -> YamlDocBuilder.YamlInt(frame.prefixCount),
+      "datatype_count" -> YamlDocBuilder.YamlInt(frame.datatypeCount),
+      "graph_start_count" -> YamlDocBuilder.YamlInt(frame.graphStartCount),
+      "graph_end_count" -> YamlDocBuilder.YamlInt(frame.graphEndCount),
+    )
+
+end MetricsPrinter