c21 opened a new pull request #31892: URL: https://github.com/apache/spark/pull/31892
<!-- Thanks for sending a pull request! Here are some tips for you: 1. If this is your first time, please read our contributor guidelines: https://spark.apache.org/contributing.html 2. Ensure you have added or run the appropriate tests for your PR: https://spark.apache.org/developer-tools.html 3. If the PR is unfinished, add '[WIP]' in your PR title, e.g., '[WIP][SPARK-XXXX] Your PR title ...'. 4. Be sure to keep the PR description updated to reflect all changes. 5. Please write your PR title to summarize what this PR proposes. 6. If possible, provide a concise example to reproduce the issue for a faster review. 7. If you want to add a new configuration, please read the guideline first for naming configurations in 'core/src/main/scala/org/apache/spark/internal/config/ConfigEntry.scala'. --> ### What changes were proposed in this pull request? <!-- Please clarify what changes you are proposing. The purpose of this section is to outline the changes and how this PR fixes the issue. If possible, please consider writing useful notes for better and faster reviews in your PR. See the examples below. 1. If you refactor some codes with changing classes, showing the class hierarchy will help reviewers. 2. If you fix some SQL features, you can provide some references of other DBMSes. 3. If there is design documentation, please add the link. 4. If there is a discussion in the mailing list, please add the link. --> This PR is to fix the LIMIT code-gen bug in https://issues.apache.org/jira/browse/SPARK-34796, where the counter variable from `BaseLimitExec` is not initialized but used in code-gen. This is because the limit counter variable will be used in upstream operators (LIMIT's child plan, e.g. `ColumnarToRowExec` operator for early termination), but in the same stage, there can be some operators doing the shortcut and not calling `BaseLimitExec`'s `doConsume()`, e.g. [HashJoin.codegenInner](https://github.com/apache/spark/blob/master/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/HashJoin.scala#L402). So if we have query that `LocalLimit - BroadcastHashJoin - FileScan` in the same stage, the whole stage code-gen compilation will be failed. Here is an example: ``` test("failed limit query") { withTable("left_table", "empty_right_table", "output_table") { spark.range(5).toDF("k").write.saveAsTable("left_table") spark.range(0).toDF("k").write.saveAsTable("empty_right_table") withSQLConf(SQLConf.ADAPTIVE_EXECUTION_ENABLED.key -> "false") { spark.sql("CREATE TABLE output_table (k INT) USING parquet") spark.sql( s""" |INSERT INTO TABLE output_table |SELECT t1.k FROM left_table t1 |JOIN empty_right_table t2 |ON t1.k = t2.k |LIMIT 3 |""".stripMargin) } } } ``` Query plan: ``` Execute InsertIntoHadoopFsRelationCommand file:/Users/chengsu/spark/sql/core/spark-warehouse/org.apache.spark.sql.SQLQuerySuite/output_table, false, Parquet, Map(path -> file:/Users/chengsu/spark/sql/core/spark-warehouse/org.apache.spark.sql.SQLQuerySuite/output_table), Append, CatalogTable( Database: default Table: output_table Created Time: Thu Mar 18 21:46:26 PDT 2021 Last Access: UNKNOWN Created By: Spark 3.2.0-SNAPSHOT Type: MANAGED Provider: parquet Location: file:/Users/chengsu/spark/sql/core/spark-warehouse/org.apache.spark.sql.SQLQuerySuite/output_table Schema: root |-- k: integer (nullable = true) ), org.apache.spark.sql.execution.datasources.InMemoryFileIndex@b25d08b, [k] +- *(3) Project [ansi_cast(k#228L as int) AS k#231] +- *(3) GlobalLimit 3 +- Exchange SinglePartition, ENSURE_REQUIREMENTS, [id=#179] +- *(2) LocalLimit 3 +- *(2) Project [k#228L] +- *(2) BroadcastHashJoin [k#228L], [k#229L], Inner, BuildRight, false :- *(2) Filter isnotnull(k#228L) : +- *(2) ColumnarToRow : +- FileScan parquet default.left_table[k#228L] Batched: true, DataFilters: [isnotnull(k#228L)], Format: Parquet, Location: InMemoryFileIndex(1 paths)[file:/Users/chengsu/spark/sql/core/spark-warehouse/org.apache.spark.sq..., PartitionFilters: [], PushedFilters: [IsNotNull(k)], ReadSchema: struct<k:bigint> +- BroadcastExchange HashedRelationBroadcastMode(List(input[0, bigint, false]),false), [id=#173] +- *(1) Filter isnotnull(k#229L) +- *(1) ColumnarToRow +- FileScan parquet default.empty_right_table[k#229L] Batched: true, DataFilters: [isnotnull(k#229L)], Format: Parquet, Location: InMemoryFileIndex(1 paths)[file:/Users/chengsu/spark/sql/core/spark-warehouse/org.apache.spark.sq..., PartitionFilters: [], PushedFilters: [IsNotNull(k)], ReadSchema: struct<k:bigint> ``` Codegen failure - https://gist.github.com/c21/ea760c75b546d903247582be656d9d66 . The uninitialized variable `_limit_counter_1` from `LocalLimitExec` is referenced in `ColumnarToRowExec`, but `BroadcastHashJoinExec` does not call `LocalLimitExec.doConsume()` to initialize the counter variable. The fix is to move the counter variable initialization to `doProduce()`, as in whole stage code-gen framework, `doProduce()` will definitely be called if upstream operators `doProduce()`/`doConsume()` is called. Note: this only happens in AQE disabled case, because we have an AQE optimization rule [EliminateUnnecessaryJoin](https://github.com/apache/spark/blob/master/sql/core/src/main/scala/org/apache/spark/sql/execution/adaptive/EliminateUnnecessaryJoin.scala#L69) to change the whole query to an empty `LocalRelation` if inner join broadcast side is empty with AQE enabled. ### Why are the changes needed? <!-- Please clarify why the changes are needed. For instance, 1. If you propose a new API, clarify the use case for a new API. 2. If you fix a bug, you can clarify why it is a bug. --> Fix query failure. ### Does this PR introduce _any_ user-facing change? <!-- Note that it means *any* user-facing change including all aspects such as the documentation fix. If yes, please clarify the previous behavior and the change this PR proposes - provide the console output, description and/or an example to show the behavior difference if possible. If possible, please also clarify if this is a user-facing change compared to the released Spark versions or within the unreleased branches such as master. If no, write 'No'. --> No. ### How was this patch tested? <!-- If tests were added, say they were added here. Please make sure to add some test cases that check the changes thoroughly including negative and positive cases if possible. If it was tested in a way different from regular unit tests, please clarify how you tested step by step, ideally copy and paste-able, so that other reviewers can test and check, and descendants can verify in the future. If tests were not added, please describe why they were not added and/or why it was difficult to add. --> Added unit test in `SQLQuerySuite.scala`. -- 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 --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org