aokolnychyi commented on code in PR #55518: URL: https://github.com/apache/spark/pull/55518#discussion_r4428977891
########## sql/catalyst/src/main/java/org/apache/spark/sql/connector/write/SupportsColumnUpdates.java: ########## @@ -0,0 +1,114 @@ +/* + * 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.connector.write; + +import org.apache.spark.annotation.Experimental; +import org.apache.spark.sql.connector.expressions.NamedReference; + +/** + * A mix-in interface for {@link RowLevelOperation}. Data sources can implement this interface to + * receive a narrow row containing only the columns declared via {@link #requiredDataAttributes()} + * for updated, copied, and reinserted records, instead of the full table row. + * <p> + * When {@link LogicalWriteInfo#columnUpdateSchema()} is present, updated, copied, and reinserted + * rows follow it: group-based operations receive them through + * {@link DataWriter#writeColumnUpdate(Object, Object)} or + * {@link DataWriter#writeColumnUpdate(Object)}, and operations that mix in {@link SupportsDelta} + * through {@link DeltaWriter#update} and {@link DeltaWriter#reinsert}. Rows the command inserts + * always follow {@link LogicalWriteInfo#schema()} and arrive through + * {@link DataWriter#write(Object)} and {@link DeltaWriter#insert}. When + * {@link LogicalWriteInfo#columnUpdateSchema()} is absent, every row follows + * {@link LogicalWriteInfo#schema()}, as for any other operation. Connectors must decide which + * layout to expect by whether {@link LogicalWriteInfo#columnUpdateSchema()} is present, not by + * the command. A builder that wants full-width rows for some command can check + * {@link RowLevelOperationInfo#command()} and build an operation without this mix-in. Currently + * Spark narrows only UPDATE. + * <p> + * The scan builder returned by {@link #newScanBuilder} should implement + * {@link org.apache.spark.sql.connector.read.SupportsPushDownRequiredColumns}, so that the scan + * does not read columns the command neither references nor declares. Otherwise the scan reads + * every column. For operations that do not mix in {@link SupportsDelta}, only top-level columns + * are pruned: a struct column the command references through a nested field is read in full. + * + * @since 4.4.0 + */ +@Experimental +public interface SupportsColumnUpdates extends RowLevelOperation { + /** + * Returns the data column references required to perform this row-level operation. + * <p> + * When Spark narrows the write, the returned columns become + * {@link LogicalWriteInfo#columnUpdateSchema()}, in declared order. Spark matches each declared + * column with the session's resolver, and {@link LogicalWriteInfo#columnUpdateSchema()} uses the + * table's spelling of each column, which can differ in case from the declared reference. + * Implementations must include every column they want to receive: every column reported by + * {@link RowLevelOperationInfo#updatedColumns()}, plus any columns needed for row lookup or + * routing. Columns that are not declared are absent from the rows the connector receives, so + * the connector must preserve their values itself. Spark may call this method more than once + * for the same operation, and every call must return the same columns. + * <p> + * If the returned array is empty, Spark does not narrow the write and the operation receives + * full-width rows. An UPDATE whose assignments are all identity assignments reports no + * {@link RowLevelOperationInfo#updatedColumns()}, so an operation that declares exactly those + * columns takes this path for such an UPDATE. An empty array is not an opt-out: an UPDATE that + * sets a column is still rejected, because the array misses that column. + * <p> + * An operation that does not mix in {@link SupportsDelta} should locate original rows through + * metadata columns returned by {@link #requiredMetadataAttributes()} that are preserved on + * update ({@link org.apache.spark.sql.connector.catalog.MetadataColumn#PRESERVE_ON_UPDATE}), + * such as a file and a row position. If it locates rows through a data column, its builder + * should build an operation without this mix-in when + * {@link RowLevelOperationInfo#updatedColumns()} contains that column. + * <p> + * Each entry must name a top-level data column of the table. For updates on nested fields such Review Comment: This sounds like a temp limitation? -- 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]
