sunchao commented on a change in pull request #34659: URL: https://github.com/apache/spark/pull/34659#discussion_r839086101
########## File path: sql/core/src/main/java/org/apache/spark/sql/execution/datasources/parquet/ParquetColumnVector.java ########## @@ -0,0 +1,379 @@ +/* + * 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.execution.datasources.parquet; + +import java.util.ArrayList; +import java.util.List; +import java.util.Set; + +import com.google.common.base.Preconditions; +import org.apache.spark.memory.MemoryMode; +import org.apache.spark.sql.execution.vectorized.OffHeapColumnVector; +import org.apache.spark.sql.execution.vectorized.OnHeapColumnVector; +import org.apache.spark.sql.execution.vectorized.WritableColumnVector; +import org.apache.spark.sql.types.ArrayType; +import org.apache.spark.sql.types.DataType; +import org.apache.spark.sql.types.DataTypes; +import org.apache.spark.sql.types.MapType; +import org.apache.spark.sql.types.StructType; + +/** + * Contains necessary information representing a Parquet column, either of primitive or nested type. + */ +final class ParquetColumnVector { + private final ParquetColumn column; + private final List<ParquetColumnVector> children; + private final WritableColumnVector vector; + + /** + * Repetition & Definition levels + * These are allocated only for leaf columns; for non-leaf columns, they simply maintain + * references to that of the former. + */ + private WritableColumnVector repetitionLevels; + private WritableColumnVector definitionLevels; + + /** Whether this column is primitive (i.e., leaf column) */ + private final boolean isPrimitive; + + /** Reader for this column - only set if 'isPrimitive' is true */ + private VectorizedColumnReader columnReader; + + ParquetColumnVector( + ParquetColumn column, + WritableColumnVector vector, + int capacity, + MemoryMode memoryMode, + Set<ParquetColumn> missingColumns) { + + DataType sparkType = column.sparkType(); + if (!sparkType.sameType(vector.dataType())) { + throw new IllegalArgumentException("Spark type: " + sparkType + + " doesn't match the type: " + vector.dataType() + " in column vector"); + } + + this.column = column; + this.vector = vector; + this.children = new ArrayList<>(); + this.isPrimitive = column.isPrimitive(); + + if (missingColumns.contains(column)) { + vector.setAllNull(); + return; + } + + if (isPrimitive) { + // TODO: avoid allocating these if not necessary, for instance, the node is of top-level + // and is not repeated, or the node is not top-level but its max repetition level is 0. + repetitionLevels = allocateLevelsVector(capacity, memoryMode); + definitionLevels = allocateLevelsVector(capacity, memoryMode); + } else { + Preconditions.checkArgument(column.children().size() == vector.getNumChildren()); + for (int i = 0; i < column.children().size(); i++) { + ParquetColumnVector childCv = new ParquetColumnVector(column.children().apply(i), + vector.getChild(i), capacity, memoryMode, missingColumns); + children.add(childCv); + + // Only use levels from non-missing child, this can happen if only some but not all + // fields of a struct are missing. + if (!childCv.vector.isAllNull()) { + this.repetitionLevels = childCv.repetitionLevels; + this.definitionLevels = childCv.definitionLevels; + } + } + + // This can happen if all the fields of a struct are missing, in which case we should mark + // the struct itself as a missing column + if (repetitionLevels == null) { + vector.setAllNull(); + } + } + } + + /** + * Returns all the children of this column. + */ + List<ParquetColumnVector> getChildren() { + return children; + } + + /** + * Returns all the leaf columns in depth-first order. + */ + List<ParquetColumnVector> getLeaves() { + List<ParquetColumnVector> result = new ArrayList<>(); + getLeavesHelper(this, result); + return result; + } + + private static void getLeavesHelper(ParquetColumnVector vector, List<ParquetColumnVector> coll) { + if (vector.isPrimitive) { + coll.add(vector); + } else { + for (ParquetColumnVector child : vector.children) { + getLeavesHelper(child, coll); + } + } + } + + /** + * Assembles this column and calculate collection offsets recursively. + * This is a no-op for primitive columns. + */ + void assemble() { + // nothing to do if the column itself is missing + if (vector.isAllNull()) return; + + DataType type = column.sparkType(); + if (type instanceof ArrayType || type instanceof MapType) { + for (ParquetColumnVector child : children) { + child.assemble(); + } + calculateCollectionOffsets(); + } else if (type instanceof StructType) { + for (ParquetColumnVector child : children) { + child.assemble(); + } + calculateStructOffsets(); + } + } + + /** + * Resets this Parquet column vector, which includes resetting all the writable column vectors + * (used to store values, definition levels, and repetition levels) for this and all its children. + */ + void reset() { + // nothing to do if the column itself is missing + if (vector.isAllNull()) return; + + vector.reset(); + repetitionLevels.reset(); + definitionLevels.reset(); + for (ParquetColumnVector child : children) { + child.reset(); + } + } + + /** + * Returns the {@link ParquetColumn} of this column vector. + */ + ParquetColumn getColumn() { + return this.column; + } + + /** + * Returns the writable column vector used to store values. + */ + WritableColumnVector getValueVector() { + return this.vector; + } + + /** + * Returns the writable column vector used to store repetition levels. + */ + WritableColumnVector getRepetitionLevelVector() { + return this.repetitionLevels; + } + + /** + * Returns the writable column vector used to store definition levels. + */ + WritableColumnVector getDefinitionLevelVector() { + return this.definitionLevels; + } + + /** + * Returns the column reader for reading a Parquet column. + */ + VectorizedColumnReader getColumnReader() { + return this.columnReader; + } + + /** + * Sets the column vector to 'reader'. Note this can only be called on a primitive Parquet + * column. + */ + void setColumnReader(VectorizedColumnReader reader) { + if (!isPrimitive) { + throw new IllegalStateException("Can't set reader for non-primitive column"); + } + this.columnReader = reader; + } + + private void calculateCollectionOffsets() { + int maxDefinitionLevel = column.definitionLevel(); + int maxElementRepetitionLevel = column.repetitionLevel(); + + ParquetColumn child = column.children().apply(0); + int maxElementDefinitionLevel = child.definitionLevel(); + if (!child.required()) maxElementDefinitionLevel -= 1; + + // There are 4 cases when calculating definition levels: + // 1. definitionLevel == maxDefinitionLevel + // ==> value is defined and not null + // 2. definitionLevel == maxDefinitionLevel - 1 + // ==> value is null + // 3. definitionLevel < maxDefinitionLevel - 1 + // ==> value doesn't exist since one of its optional parents is null + // 4. definitionLevel > maxDefinitionLevel + // ==> value is a nested element within an array or map Review comment: Yea `maxDefinitionLevel` and `maxRepetitionLevel` is for the collection itself here. For collection elements, the naming would be `maxElementDefinitionLevel` and `maxElementRepetitionLevel`. -- 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]
