AnishMahto commented on code in PR #55969: URL: https://github.com/apache/spark/pull/55969#discussion_r3277079887
########## sql/pipelines/src/main/scala/org/apache/spark/sql/pipelines/autocdc/Scd1BatchProcessor.scala: ########## @@ -0,0 +1,67 @@ +/* + * 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.pipelines.autocdc + +import org.apache.spark.sql.{functions => F} +import org.apache.spark.sql.catalyst.util.QuotingUtils +import org.apache.spark.sql.classic.DataFrame +import org.apache.spark.util.ArrayImplicits._ + +/** + * Per-microbatch processor for SCD Type 1 AutoCDC flows, complying to the specified [[changeArgs]] + * configuration. + */ +case class Scd1BatchProcessor(changeArgs: ChangeArgs) { + /** + * Deduplicate the incoming CDC microbatch by key, keeping the most recent event per key + * as ordered by [[ChangeArgs.sequencing]]. + * + * For SCD1 we only care about the most recent (by sequence value) event per key. When + * multiple events share the same key and the same sequence value, the row selected is + * non-deterministic and undefined. + * + * @param validatedMicrobatch A microbatch that has already been validated such that the + * sequencing column should not contain null values, and its data type + * should support ordering. + * + * The schema of the returned dataframe matches the schema of the microbatch exactly. + */ + def deduplicateMicrobatch(validatedMicrobatch: DataFrame): DataFrame = { + // The `max_by` API can only return a single column, so pack/unpack the entire row into a + // temporary column before and after the `max_by` operation. + val winningRowCol = Scd1BatchProcessor.winningRowColName + + val allMicrobatchColumns = + validatedMicrobatch.columns + .map(colName => F.col(QuotingUtils.quoteIdentifier(colName))) + .toImmutableArraySeq + + validatedMicrobatch + .groupBy(changeArgs.keys.map(k => F.col(k.quoted)): _*) + .agg( + F.max_by(F.struct(allMicrobatchColumns: _*), changeArgs.sequencing) + .as(winningRowCol) + ) + .select(F.col(s"$winningRowCol.*")) + } +} + +object Scd1BatchProcessor { + // Columns prefixed with `__spark_autocdc_` are reserved for internal SDP AutoCDC processing. + private val winningRowColName = "__spark_autocdc_winning_row" Review Comment: Here are my thoughts and what I did: - Even if we add a validation during microbatch processing, its not really failing early; we're already deep in the flow execution code path - What we should do that is idiomatic to SDP is when we introduce the flow construction/analysis logic to the dataflow graph engine, we validate that none of the analyzed schemas contain the reserved prefix. We can additionally verify none of the `ChangeArg` unqualified references contain the reserved prefix during `ChangeArg` construction at flow registration time - this will give us true eager validation _before_ flow execution - I added a test to lock in the analysis exception in the event the microbatch has a reserved column name: `deduplicateMicrobatch fails when a key column collides with the reserved name`. The spark exception makes it clear exactly which column name is ambiguous due to duplicate -- 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]
