ArnavBalyan commented on code in PR #9074: URL: https://github.com/apache/incubator-gluten/pull/9074#discussion_r2063294959
########## backends-velox/src/main/scala/org/apache/gluten/execution/ColumnarCollectTailExec.scala: ########## @@ -0,0 +1,142 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +package org.apache.gluten.execution + +import org.apache.gluten.backendsapi.BackendsApiManager +import org.apache.gluten.columnarbatch.ColumnarBatches +import org.apache.gluten.columnarbatch.VeloxColumnarBatches +import org.apache.gluten.extension.columnar.transition.Convention +import org.apache.gluten.iterator.Iterators + +import org.apache.spark.rdd.RDD +import org.apache.spark.serializer.Serializer +import org.apache.spark.sql.catalyst.expressions.Attribute +import org.apache.spark.sql.catalyst.plans.physical.SinglePartition +import org.apache.spark.sql.execution.{ShuffledColumnarBatchRDD, SparkPlan} +import org.apache.spark.sql.execution.metric.{SQLMetric, SQLShuffleWriteMetricsReporter} +import org.apache.spark.sql.metric.SQLColumnarShuffleReadMetricsReporter +import org.apache.spark.sql.vectorized.ColumnarBatch + +import scala.collection.mutable + +case class ColumnarCollectTailExec( + limit: Int, + child: SparkPlan +) extends ColumnarCollectTailBaseExec(limit, child) { + + private def collectTailRows( + partitionIter: Iterator[ColumnarBatch], + limit: Int + ): Iterator[ColumnarBatch] = { + + if (!partitionIter.hasNext || limit <= 0) { + return Iterator.empty + } + + val tailQueue = new mutable.ListBuffer[ColumnarBatch]() + var totalRowsInTail = 0L + + while (partitionIter.hasNext) { + val batch = partitionIter.next() + val batchRows = batch.numRows() + ColumnarBatches.retain(batch) + tailQueue += batch + totalRowsInTail += batchRows + + while (totalRowsInTail > limit && tailQueue.nonEmpty) { + val front = tailQueue.remove(0) + val frontRows = front.numRows().toLong + val overflow = totalRowsInTail - limit + + if (frontRows <= overflow) { + totalRowsInTail -= frontRows + } else { + val keep = frontRows - overflow + val partial = VeloxColumnarBatches.slice(front, overflow.toInt, keep.toInt) Review Comment: That's a good point, I think this comes down to the cost of slice vs the memory pressure. Had added slice here to keep only the required number of rows in queue similar to Spark. If we delay the slice, in the worst case, it can potentially load all batches into the queue and put oom pressure, since iterator is not aware of the batch sizes or total rows which can be expected, what do you think? -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
