voonhous commented on PR #19687: URL: https://github.com/apache/hudi/pull/19687#issuecomment-5412609320
Changes beyond the rounds above: - Hive guards (legacy and file-group-reader): honour nested column pruning (`select s.other` reads, `s.inner` fails fast); the legacy anchor is the parsed `struct<metadata:binary,value:binary>` node, not a type-string scan; `columns.types` is screened, parse failures bail out. Legs: count(*), schema-on-read, bootstrap, INT96, nested paths, both list layouts, a `typed_value` sibling. - Sort columns: every sink trims the config list (Spark strategies, consistent-bucket partitioner, Flink `ClusteringOperator`); `id, ts` passed validation and failed in the job. Consistent-bucket and inline-clustering legs. - Tests: `TestParquetSchemaEvolutionUtils` for the schema-on-read arms, a partitioned bootstrap leg, #19558's test restored, duplicate and NULL-blind legs gone, scaffolding shared. - CI: #18961's `testApplyInferredShredding` checked `typed_value` nullability through the `Variant` accessor, which `fix(schema)` unwraps to the value type; the assertion now reads the field schema (nullable on the wire either way). Caveat: no CI lane builds Spark 4.0; its module tests ran locally only. -- 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]
