jackylk commented on a change in pull request #3542: [CARBONDATA-3640] Insert from stage command support partition table URL: https://github.com/apache/carbondata/pull/3542#discussion_r361828062
########## File path: integration/spark2/src/main/scala/org/apache/spark/sql/execution/command/management/CarbonInsertFromStageCommand.scala ########## @@ -269,11 +281,90 @@ case class CarbonInsertFromStageCommand( SparkSQLUtil.sessionState(spark).newHadoopConf() ).map { row => (row._1, FailureCauses.NONE == row._2._2.failureCauses) - } + } + + LOGGER.info(s"finish data loading, time taken ${System.currentTimeMillis() - start}ms") + } + + /** + * Start global sort loading of partition table + */ + private def startLoadingWithPartition( + spark: SparkSession, + table: CarbonTable, + loadModel: CarbonLoadModel, + stageInput: Seq[StageInput] + ): Unit = { + val partitionDataList = listPartitionFiles(stageInput) + val start = System.currentTimeMillis() + var index = 0 + partitionDataList.map( + partitionData => { + index = index + 1 + val partition = partitionData._1 + val splits = partitionData._2 + LOGGER.info(s"start to load ${splits.size} files into " + + s"${table.getDatabaseName}.${table.getTableName}") + val dataFrame = createInputDataFrameOfInternalRow(spark, table, splits) + val columns = dataFrame.columns + val header = columns.mkString(",") + val selectColumns = columns.filter(!partition.contains(_)) + val selectedDataFrame = dataFrame.select(selectColumns.head, selectColumns.tail : _*) + val loadCommand = CarbonLoadDataCommand( + databaseNameOp = Option(table.getDatabaseName), + tableName = table.getTableName, + factPathFromUser = null, + dimFilesPath = Seq(), + options = scala.collection.immutable.Map("fileheader" -> header), + isOverwriteTable = false, + inputSqlString = null, + dataFrame = Some(selectedDataFrame), + updateModel = None, + tableInfoOp = None, + internalOptions = Map.empty, + partition = partition + ) + loadCommand.run(spark) + } + ) LOGGER.info(s"finish data loading, time taken ${System.currentTimeMillis() - start}ms") } + private def listPartitionFiles(stageInputs : Seq[StageInput]): + Seq[(Map[String, Option[String]], Seq[InputSplit])] = { + val partitionMap = new util.HashMap[Map[String, Option[String]], util.List[InputSplit]]() + stageInputs.foreach( + stageInput => { + val locations = stageInput.getLocations.asScala + locations.foreach( + location => { + val partition = location.getPartitions.asScala.map(t => (t._1, Option(t._2))).toMap Review comment: Why Option is required? Can't we remove it? ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services