cloud-fan commented on a change in pull request #26973: [SPARK-30323][SQL] Support filters pushdown in CSV datasource URL: https://github.com/apache/spark/pull/26973#discussion_r365641683
########## File path: sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/csv/CSVFilters.scala ########## @@ -0,0 +1,212 @@ +/* + * 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.catalyst.csv + +import scala.util.Try + +import org.apache.spark.sql.catalyst.InternalRow +import org.apache.spark.sql.catalyst.expressions._ +import org.apache.spark.sql.internal.SQLConf +import org.apache.spark.sql.sources +import org.apache.spark.sql.types.{BooleanType, StructType} + +/** + * An instance of the class compiles filters to predicates and allows to + * apply the predicates to an internal row with partially initialized values + * converted from parsed CSV fields. + * + * @param filters The filters pushed down to CSV datasource. + * @param dataSchema The full schema with all fields in CSV files. + * @param requiredSchema The schema with only fields requested by the upper layer. + * @param columnPruning true if CSV parser can read sub-set of columns otherwise false. + */ +class CSVFilters( + filters: Seq[sources.Filter], + dataSchema: StructType, + requiredSchema: StructType, + columnPruning: Boolean) { + require(checkFilters(), "All filters must be applicable to the data schema.") + + /** + * The schema to read from the underlying CSV parser. + * It combines the required schema and the fields referenced by filters. + */ + val readSchema: StructType = { + if (columnPruning) { + val refs = filters.flatMap(_.references).toSet + val readFields = dataSchema.filter { field => + requiredSchema.contains(field) || refs.contains(field.name) + } + StructType(readFields) + } else { + dataSchema + } + } + + /** + * Converted filters to predicates and grouped by maximum field index + * in the read schema. For example, if an filter refers to 2 attributes + * attrA with field index 5 and attrB with field index 10 in the read schema: + * 0 === $"attrA" or $"attrB" < 100 + * the filter is compiled to a predicate, and placed to the `predicates` + * array at the position 10. In this way, if there is a row with initialized + * fields from the 0 to 10 index, the predicate can be applied to the row + * to check that the row should be skipped or not. + * Multiple predicates with the same maximum reference index are combined + * by the `And` expression. + */ + private val predicates: Array[BasePredicate] = { + val len = readSchema.fields.length + val groupedPredicates = Array.fill[BasePredicate](len)(null) + if (SQLConf.get.csvFilterPushDown) { + val groupedExprs = Array.fill(len)(Seq.empty[Expression]) + for (filter <- filters) { + val expr = CSVFilters.filterToExpression(filter, toRef) + val refs = filter.references + if (refs.isEmpty) { + // For example, AlwaysTrue and AlwaysFalse doesn't have any references + for (i <- 0 until len) { + groupedExprs(i) ++= expr Review comment: for literals, shall we only put it in position 0? It only needs to be evaluated once. ---------------------------------------------------------------- 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 --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org
