The GitHub Actions job "CI" on fluss.git/feat/fix-partition-pruning-types has failed. Run started by GitHub user fresh-borzoni (triggered by fresh-borzoni).
Head commit for run: e83a8393770934074bbaa53680b6e1dfa8dac0cd / Anton Borisov <[email protected]> [flink][spark] Fix partition pruning for non-string partition keys Partition predicate pushdown stringified literals and partition values before evaluation, so range comparisons fell back to string lexicographic order. An INT partition column with values 2 and 10 under WHERE pt > 2 lex-compared "10" < "2" and incorrectly dropped partition 10. Add PartitionUtils.toPartitionRow and PartitionUtils.partitionRowType in fluss-common. Use them from SparkPartitionPredicate and FlinkSourceEnumerator; drop the stringify step in FlinkTableSource and delete StringifyPredicateVisitor. The stringifier was also hiding two latent gaps in LeafPredicate.get: BYTES had no case (UnsupportedOperationException) and TIMESTAMP_WITH_LOCAL_TIME_ZONE used getTimestampNtz instead of getTimestampLtz (ClassCastException). Both exercised by testStreamingReadAllPartitionTypePushDown; fix in the same file. Regression test for the partition pruning bug added with an INT partition column and a range predicate in SparkLogTableReadTest and FlinkTableSourceITCase. Closes #3292. Report URL: https://github.com/apache/fluss/actions/runs/25907175963 With regards, GitHub Actions via GitBox
