revans2 commented on a change in pull request #29067: URL: https://github.com/apache/spark/pull/29067#discussion_r457558925
########## File path: sql/core/src/main/scala/org/apache/spark/sql/columnar/CachedBatchSerializer.scala ########## @@ -0,0 +1,291 @@ +/* + * 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.spark.sql.columnar + +import org.apache.spark.annotation.{DeveloperApi, Since} +import org.apache.spark.internal.Logging +import org.apache.spark.rdd.RDD +import org.apache.spark.sql.catalyst.InternalRow +import org.apache.spark.sql.catalyst.dsl.expressions._ +import org.apache.spark.sql.catalyst.expressions.{And, Attribute, AttributeReference, BindReferences, EqualNullSafe, EqualTo, Expression, GreaterThan, GreaterThanOrEqual, In, IsNotNull, IsNull, Length, LessThan, LessThanOrEqual, Literal, Or, Predicate, StartsWith} +import org.apache.spark.sql.execution.SparkPlan +import org.apache.spark.sql.execution.columnar.{ColumnStatisticsSchema, PartitionStatistics} +import org.apache.spark.sql.internal.SQLConf +import org.apache.spark.sql.types.{AtomicType, BinaryType, StructType} +import org.apache.spark.sql.vectorized.ColumnarBatch + +/** + * Basic interface that all cached batches of data must support. This is primarily to allow + * for metrics to be handled outside of the encoding and decoding steps in a standard way. + */ +@DeveloperApi +@Since("3.1.0") +trait CachedBatch { + def numRows: Int + def sizeInBytes: Long +} + +/** + * Provides APIs for compressing, filtering, and decompressing SQL data that will be + * persisted/cached. + */ +@DeveloperApi +@Since("3.1.0") +trait CachedBatchSerializer extends Serializable { Review comment: The major goal was to expose some way to provide custom code for compressing/decompressing the cached data. Decompression was already really fast and had APIs for columnar output vs non-columnar output so I mostly just kept them in place. `convertForCache` does have a goal to be able to let someone process columnar input, but because of how the `SparkPlan` is created that is not a simple task. `QueryExecution.executedPlan` which is the source of the `SparkPlan` will always produce row based output. That was one of the design goals for the columnar execution work. It had to be truly transparent. Currently the only way to get around that is by walking the SparkPlan to check if the last stage is a code generation phase with only a child that is doing a columnar to row transition. If so then it is safe to remove that and process it directly as columnar. If you want me to create a trait for `SparkPlan` that exposes a simpler API just for caching, I can, but it will not be able to expose anything columnar unless we make larger changes. I am fine with making those larger changes, I just want to be sure that the community is willing to accept them conceptually before I start to touch a lot of the code. ---------------------------------------------------------------- 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: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
