ajantha-bhat commented on a change in pull request #3538: [CARBONDATA-3637]
Optimize insert into flow
URL: https://github.com/apache/carbondata/pull/3538#discussion_r376348551
##########
File path:
integration/spark2/src/main/scala/org/apache/spark/sql/execution/command/management/CarbonInsertIntoCommand.scala
##########
@@ -17,85 +17,518 @@
package org.apache.spark.sql.execution.command.management
-import org.apache.spark.sql.{CarbonDatasourceHadoopRelation, Dataset, Row,
SparkSession}
-import org.apache.spark.sql.catalyst.plans.logical.{GlobalLimit, LogicalPlan}
-import org.apache.spark.sql.execution.command.{AtomicRunnableCommand,
DataCommand}
-import org.apache.spark.storage.StorageLevel
-
-import org.apache.carbondata.common.logging.{LogService, LogServiceFactory}
-import org.apache.carbondata.core.constants.CarbonCommonConstants
-import org.apache.carbondata.core.util.{CarbonProperties,
ThreadLocalSessionInfo}
-
-case class CarbonInsertIntoCommand(
- relation: CarbonDatasourceHadoopRelation,
- child: LogicalPlan,
- overwrite: Boolean,
- partition: Map[String, Option[String]])
+import java.text.SimpleDateFormat
+import java.util
+
+import scala.collection.JavaConverters._
+import scala.collection.mutable
+
+import org.apache.spark.rdd.RDD
+import org.apache.spark.sql.{AnalysisException, CarbonEnv, DataFrame, Dataset,
Row, SparkSession}
+import org.apache.spark.sql.catalyst.InternalRow
+import org.apache.spark.sql.catalyst.catalog.HiveTableRelation
+import org.apache.spark.sql.catalyst.expressions.{Alias, AttributeReference,
Literal, NamedExpression}
+import org.apache.spark.sql.catalyst.plans.logical.{LogicalPlan, Project}
+import org.apache.spark.sql.execution.command.{AtomicRunnableCommand,
DataLoadTableFileMapping}
+import org.apache.spark.sql.execution.datasources.LogicalRelation
+import org.apache.spark.sql.types.{StructField, StructType}
+import org.apache.spark.util.CausedBy
+
+import org.apache.carbondata.common.logging.LogServiceFactory
+import org.apache.carbondata.converter.SparkDataTypeConverterImpl
+import org.apache.carbondata.core.datastore.impl.FileFactory
+import org.apache.carbondata.core.indexstore.PartitionSpec
+import org.apache.carbondata.core.metadata.schema.SchemaEvolutionEntry
+import org.apache.carbondata.core.metadata.schema.table.{CarbonTable,
TableInfo, TableSchema}
+import org.apache.carbondata.core.metadata.schema.table.column.ColumnSchema
+import org.apache.carbondata.core.statusmanager.{LoadMetadataDetails,
SegmentStatus, SegmentStatusManager}
+import org.apache.carbondata.core.util.{CarbonProperties, DataTypeUtil,
ThreadLocalSessionInfo}
+import org.apache.carbondata.core.util.path.CarbonTablePath
+import org.apache.carbondata.events.exception.PreEventException
+import org.apache.carbondata.events.OperationContext
+import org.apache.carbondata.processing.loading.TableProcessingOperations
+import org.apache.carbondata.processing.loading.exception.NoRetryException
+import org.apache.carbondata.processing.loading.model.CarbonLoadModel
+import org.apache.carbondata.processing.util.CarbonLoaderUtil
+import org.apache.carbondata.spark.rdd.CarbonDataRDDFactory
+import org.apache.carbondata.spark.util.CarbonScalaUtil
+
+/*
+* insert into without df, by just using logical plan
+*
+*/
+case class CarbonInsertIntoCommand(databaseNameOp: Option[String],
+ tableName: String,
+ options: Map[String, String],
+ isOverwriteTable: Boolean,
+ var logicalPlan: LogicalPlan,
+ var dimFilesPath: Seq[DataLoadTableFileMapping] = Seq(),
+ var inputSqlString: String = null,
+ var tableInfoOp: Option[TableInfo] = None,
+ var internalOptions: Map[String, String] = Map.empty,
+ var partition: Map[String, Option[String]] = Map.empty,
+ var operationContext: OperationContext = new OperationContext)
extends AtomicRunnableCommand {
- var loadCommand: CarbonLoadDataCommand = _
+ var table: CarbonTable = _
+
+ var logicalPartitionRelation: LogicalRelation = _
+
+ var sizeInBytes: Long = _
+
+ var currPartitions: util.List[PartitionSpec] = _
+
+ var parentTablePath: String = _
+
+ val LOGGER = LogServiceFactory.getLogService(this.getClass.getName)
+
+ var scanResultRdd: RDD[InternalRow] = _
+
+ var timeStampFormat: SimpleDateFormat = _
+
+ var dateFormat: SimpleDateFormat = _
+
+ var finalPartition: Map[String, Option[String]] = Map.empty
+
+ var isInsertIntoWithConverterFlow: Boolean = false
+
+ var dataFrame: DataFrame = _
override def processMetadata(sparkSession: SparkSession): Seq[Row] = {
- setAuditTable(relation.carbonTable.getDatabaseName,
relation.carbonTable.getTableName)
- val LOGGER = LogServiceFactory.getLogService(this.getClass.getName)
- def containsLimit(plan: LogicalPlan): Boolean = {
- plan find {
- case limit: GlobalLimit => true
- case other => false
- } isDefined
+ if (!tableInfoOp.isDefined) {
+ throw new RuntimeException(" table info must be present when logical
relation exist")
}
-
+ // If logical plan is unresolved, need to convert it to resolved.
+ dataFrame = Dataset.ofRows(sparkSession, logicalPlan)
+ logicalPlan = dataFrame.queryExecution.analyzed
+ var isInsertFromTable = false
+ logicalPlan.collect {
+ case _: LogicalRelation =>
+ isInsertFromTable = true
+ }
+ // Currently projection re-ordering is based on schema ordinal,
+ // for some scenarios in alter table scenario, schema ordinal logic cannot
be applied.
+ // So, sending it to old flow
+ // TODO: Handle this in future, this must use new flow.
+ if (!isInsertFromTable || isAlteredSchema(tableInfoOp.get.getFactTable)) {
+ isInsertIntoWithConverterFlow = true
+ }
+ if (isInsertIntoWithConverterFlow) {
+ return Seq.empty
+ }
+ setAuditTable(CarbonEnv.getDatabaseName(databaseNameOp)(sparkSession),
tableName)
ThreadLocalSessionInfo
.setConfigurationToCurrentThread(sparkSession.sessionState.newHadoopConf())
- val isPersistEnabledUserValue = CarbonProperties.getInstance
- .getProperty(CarbonCommonConstants.CARBON_INSERT_PERSIST_ENABLED,
- CarbonCommonConstants.CARBON_INSERT_PERSIST_ENABLED_DEFAULT)
- val isPersistRequired =
- isPersistEnabledUserValue.equalsIgnoreCase("true") ||
containsLimit(child)
- val df =
- if (isPersistRequired) {
- LOGGER.info("Persist enabled for Insert operation")
- Dataset.ofRows(sparkSession, child).persist(
- StorageLevel.fromString(
- CarbonProperties.getInstance.getInsertIntoDatasetStorageLevel))
+ val (sizeInBytes, table, dbName, logicalPartitionRelation, finalPartition)
= CommonLoadUtils
+ .processMetadataCommon(
+ sparkSession,
+ databaseNameOp,
+ tableName,
+ tableInfoOp,
+ partition)
+ this.sizeInBytes = sizeInBytes
+ this.table = table
+ this.logicalPartitionRelation = logicalPartitionRelation
+ this.finalPartition = finalPartition
+ setAuditTable(dbName, tableName)
+ Seq.empty
+ }
+
+ override def processData(sparkSession: SparkSession): Seq[Row] = {
+ if (isInsertIntoWithConverterFlow) {
+ return CarbonInsertIntoWithDf(
+ databaseNameOp,
+ tableName,
+ options,
+ isOverwriteTable,
+ dimFilesPath,
+ dataFrame,
+ inputSqlString,
+ None,
+ tableInfoOp,
+ internalOptions,
+ partition).process(sparkSession)
+ }
+ val carbonProperty: CarbonProperties = CarbonProperties.getInstance()
+ val dbName = CarbonEnv.getDatabaseName(databaseNameOp)(sparkSession)
+ val hadoopConf = sparkSession.sessionState.newHadoopConf()
+ carbonProperty.addProperty("zookeeper.enable.lock", "false")
+ val factPath = ""
+ currPartitions = CommonLoadUtils.getCurrentParitions(sparkSession, table)
+ CommonLoadUtils.setNumberOfCoresWhileLoading(sparkSession, carbonProperty)
+ val optionsFinal: util.Map[String, String] =
+ CommonLoadUtils.getFinalLoadOptions(
+ carbonProperty, table, options)
+ val carbonLoadModel: CarbonLoadModel = CommonLoadUtils.prepareLoadModel(
+ hadoopConf,
+ factPath,
+ optionsFinal, parentTablePath, table, isDataFrame = true,
internalOptions, partition, options)
+
+ val (tf, df) = CommonLoadUtils.getTimeAndDateFormatFromLoadModel(
+ carbonLoadModel)
Review comment:
ok
----------------------------------------------------------------
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
