imbajin opened a new pull request, #3405: URL: https://github.com/apache/hugegraph/pull/3405
## Purpose of the PR Fixes apache/hugegraph#3397. Source proposal: [org #337](https://github.com/hugegraph/hugegraph/pull/337), using the same source branch. When a client selects precise union deduplication and the union switches to SortShuffle, repeated keys can disappear entirely. For example, 20,000 keys scanned twice return no rows. The reducer now emits each comparator-equal group once, including the final group. ## Main Changes - Correct the existing sorted reducer and preserve its pending result across iterator probes. - Add regressions with fresh rows across map flushes, mixed singleton/repeated groups, repeated probes and EOF. - Keep the existing name-based comparison and in-memory behavior. The separate content-hashing fix remains in [ASF #3404](https://github.com/apache/hugegraph/pull/3404). ## Verifying these changes - [x] Need tests: both new regressions fail on the unchanged reducer and pass with this fix. - [x] Java 17 full clean compile and the Store core test suite pass. - [x] EditorConfig formatting and whitespace checks pass; independent review found no blocking issues. The regressions trigger the Union sorting threshold, not SortShuffle's separate physical file-spill threshold. They do not validate a running RocksDB/Store cluster. Default deduplication settings remain unchanged. ## Documentation Status - [x] `Doc - No Need`: restores the existing internal union contract without changing API or configuration. -- 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]
