Arnab Karmakar has posted comments on this change. ( https://gerrit.cloudera.org/24973 )
Change subject: IMPALA-15365: Iceberg UUID partition transform support ...................................................................... Patch Set 2: (9 comments) Thanks for the comments! https://gerrit.cloudera.org/#/c/24973/1//COMMIT_MSG Commit Message: https://gerrit.cloudera.org/#/c/24973/1//COMMIT_MSG@9 PS1, Line 9: Adds partition transform support for the Iceberg UUID type. The Icebe > Please mention it is not an Impala limitation, but the spec only allows the Done https://gerrit.cloudera.org/#/c/24973/1//COMMIT_MSG@13 PS1, Line 13: and ALTER TABLE ... SET PARTITION SPEC > UUIDs use unsigned byte-wise comparison: https://github.com/apache/iceberg/ Done https://gerrit.cloudera.org/#/c/24973/1//COMMIT_MSG@14 PS1, Line 14: : > Please mention that this is a bug in the Iceberg Java library: https://gith Done https://gerrit.cloudera.org/#/c/24973/1/fe/src/main/java/org/apache/impala/common/IcebergPartitionPredicateConverter.java File fe/src/main/java/org/apache/impala/common/IcebergPartitionPredicateConverter.java: https://gerrit.cloudera.org/#/c/24973/1/fe/src/main/java/org/apache/impala/common/IcebergPartitionPredicateConverter.java@71 PS1, Line 71: partition. != and NOT IN only prune manifests whose summary holds a single v > Holds for writers that use the Java library (Impala, Trino, Spark). A write Done https://gerrit.cloudera.org/#/c/24973/1/fe/src/main/java/org/apache/impala/common/IcebergPartitionPredicateConverter.java@73 PS1, Line 73: with the Iceberg Java library's signed c > != and NOT_IN could be allowed as well. Done https://gerrit.cloudera.org/#/c/24973/1/fe/src/main/java/org/apache/impala/common/IcebergPartitionPredicateConverter.java@74 PS1, Line 74: use the library (Impa > Instead of using canUsePartitionKeyScan(), consider extracting logic to a n Done https://gerrit.cloudera.org/#/c/24973/1/fe/src/main/java/org/apache/impala/common/IcebergPredicateConverter.java File fe/src/main/java/org/apache/impala/common/IcebergPredicateConverter.java: https://gerrit.cloudera.org/#/c/24973/1/fe/src/main/java/org/apache/impala/common/IcebergPredicateConverter.java@475 PS1, Line 475: The Iceberg Java library compares UUID > Same as in commit message: it is a bug in the Iceberg Java Library, not in Done https://gerrit.cloudera.org/#/c/24973/1/fe/src/main/java/org/apache/impala/common/IcebergPredicateConverter.java@476 PS1, Line 476: // use unsigned byte order, so Iceberg could skip data files that contain matching > When UUID is part of identity-partitioning, we could still push down = and Thanks for pointing it out! They can be pushed down with the narrowed down forms. The identity partition would be simpler but since bucket transform might need some changes in planning and its own set of tests, it'd be cleaner to do it in a follow-up patch. Filed IMPALA-15476. https://gerrit.cloudera.org/#/c/24973/1/testdata/data/README File testdata/data/README: https://gerrit.cloudera.org/#/c/24973/1/testdata/data/README@1035 PS1, Line 1035: iceberg_uuid_test_part > Each of the 4 appends writes a single-file manifest, so the partition summa Done -- To view, visit https://gerrit.cloudera.org/24973 To unsubscribe, visit https://gerrit.cloudera.org/settings Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I1dc3af4418285d6ea1d489912419be470e668f29 Gerrit-Change-Number: 24973 Gerrit-PatchSet: 2 Gerrit-Owner: Arnab Karmakar <[email protected]> Gerrit-Reviewer: Arnab Karmakar <[email protected]> Gerrit-Reviewer: Impala Public Jenkins <[email protected]> Gerrit-Reviewer: Zoltan Borok-Nagy <[email protected]> Gerrit-Comment-Date: Sat, 03 Oct 2026 17:47:03 +0000 Gerrit-HasComments: Yes
