szehon-ho commented on code in PR #55969: URL: https://github.com/apache/spark/pull/55969#discussion_r3269504915
########## sql/pipelines/src/main/scala/org/apache/spark/sql/pipelines/autocdc/Scd1BatchProcessor.scala: ########## @@ -0,0 +1,59 @@ +/* + * 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. + * + * The schema of the returned dataframe matches the schema of the microbatch exactly. + */ + def deduplicateMicrobatch(microbatchDf: DataFrame): DataFrame = { Review Comment: The scaladoc documents tie-breaking and null sequencing behavior; consider adding tests for: 1. **Equal `sequencing` for the same key** — even a lightweight test that documents non-determinism (or runs twice) would lock in the contract. 2. **Null `sequencing`** — `max_by` has subtle null ordering (see `DataFrameAggregateSuite` "max_by"); worth defining expected CDC behavior or asserting we reject nulls upstream. 3. **Single row per key (no-op)** — cheap sanity check that one input row passes through unchanged. Not blocking if you prefer to add these when merge logic lands. ########## sql/pipelines/src/test/scala/org/apache/spark/sql/pipelines/autocdc/Scd1BatchProcessorSuite.scala: ########## @@ -0,0 +1,232 @@ +/* + * 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.SparkFunSuite +import org.apache.spark.sql.{functions => F, Row} +import org.apache.spark.sql.classic.DataFrame +import org.apache.spark.sql.test.SharedSparkSession +import org.apache.spark.sql.types._ + +class Scd1BatchProcessorSuite extends SparkFunSuite with SharedSparkSession { Review Comment: Nit: Spark convention for SQL tests using `checkAnswer` is to extend `QueryTest` explicitly, e.g. `class Scd1BatchProcessorSuite extends QueryTest with SharedSparkSession` (rather than `SparkFunSuite with SharedSparkSession`). `SharedSparkSession` already extends `QueryTest`, so this works today — just consistency with other `sql/core` / pipelines suites. ########## sql/pipelines/src/main/scala/org/apache/spark/sql/pipelines/autocdc/Scd1BatchProcessor.scala: ########## @@ -0,0 +1,59 @@ +/* + * 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. + * + * The schema of the returned dataframe matches the schema of the microbatch exactly. + */ + def deduplicateMicrobatch(microbatchDf: 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 = OutOfOrderCdcMergeUtils.tempColName("__winning_row") + + val allMicrobatchColumns = + microbatchDf.columns + .map(colName => F.col(QuotingUtils.quoteIdentifier(colName))) + .toImmutableArraySeq + + microbatchDf + .groupBy(changeArgs.keys.map(k => F.col(k.quoted)): _*) Review Comment: If `changeArgs.keys` is empty, `groupBy()` collapses the entire microbatch into a single group (one output row). Worth guarding with `require(changeArgs.keys.nonEmpty, ...)` here or validating at `ChangeArgs` construction in the registration PR. ########## sql/pipelines/src/main/scala/org/apache/spark/sql/pipelines/autocdc/Scd1BatchProcessor.scala: ########## @@ -0,0 +1,59 @@ +/* + * 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. + * + * The schema of the returned dataframe matches the schema of the microbatch exactly. + */ + def deduplicateMicrobatch(microbatchDf: 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 = OutOfOrderCdcMergeUtils.tempColName("__winning_row") Review Comment: `tempColName` generates a fresh UUID on every `deduplicateMicrobatch` call, so the logical plan column name differs across invocations. Fine for correctness; just a heads-up if you later add plan-golden / EXPLAIN tests — you may want a stable internal name with a collision-safe prefix instead. Non-blocking. -- 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]
