JingsongLi opened a new pull request, #8250: URL: https://github.com/apache/paimon/pull/8250
## Summary This PR adds an optional BTree global index build mode that can merge discontinuous RowID ranges into one index file span when RowID gaps are permanent and will not be filled later. The new option is disabled by default. ## Changes - Add `btree-index.build.merge-row-ranges` for opt-in sparse RowID range merging during BTree index builds. - Preserve `IndexedSplit` row ranges during local index building so actual reads still use the real data ranges while index metadata can cover the enclosing span. - Wire the option through core, Flink, and Spark BTree index builders. - Relax global index RowID existence conflict checks only when the option is enabled and the span endpoints are covered by current data files. - Add tests for default option behavior, sparse range grouping, conflict compatibility, and table-level reads through a sparse BTree index. ## Testing - `mvn -pl paimon-core -am -Pfast-build -DfailIfNoTests=false -Dtest=BtreeGlobalIndexTableTest#testBuildBTreeGlobalIndexWithDiscontinuousRowIdRanges test` - `mvn -pl paimon-core -am -Pfast-build -DfailIfNoTests=false -Dtest=BTreeGlobalIndexBuilderSplitTest,BtreeGlobalIndexTableTest#testBuildBTreeGlobalIndexWithDiscontinuousRowIdRanges,ConflictDetectionTest test` - `mvn -pl paimon-common -am -Pfast-build -DfailIfNoTests=false -Dtest=BTreeIndexOptionsTest test` - `mvn -pl paimon-flink/paimon-flink-common,paimon-spark/paimon-spark-common -am -Pfast-build -DskipTests compile` - `git diff --check` ## Notes The option should only be enabled when RowID gaps are permanent. With the default value, existing BTree index build behavior is unchanged. -- 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]
