JingsongLi opened a new pull request, #9442: URL: https://github.com/apache/paimon/pull/9442
## What is the purpose of the change Add an exact byte-oriented FM global index for CONTAINS predicates, with candidate-aware refinement for vector and sibling-index pre-filtering. The implementation uses independently readable partitions, blocked four-ary wavelet ranks, sampled suffix locations, demand-paged reads, checksummed compressed blocks, and exact stored-value verification for dense matches. Flink and Spark use the default unsorted global-index build topology. ## Brief change log - add candidate-aware coarse and exact CONTAINS refinement contracts - add streaming partitioned FM index writer and demand-paged reader - add versioned V1 storage format, manifest metadata, checksums, compression, and golden fixture - add bounded-memory SA-IS construction and candidate partition task pruning - register fmindex and document Flink and Spark creation procedures ## Verifying this change - paimon-common full suite: 12,459 tests passed - FM index tests: 24 tests passed - Flink SortedIndexTopoBuilderTest: 10 tests passed - Spark DefaultGlobalIndexTopoBuilderTest: 6 tests passed - Spark FMGlobalIndexProcedureTest: 1 test passed - Spotless check passed for all affected reactor modules This PR intentionally contains no N-Gram implementation or registration. -- 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]
