soumilshah1995 opened a new pull request, #16370: URL: https://github.com/apache/iceberg/pull/16370
## Summary Kafka Connect uses Iceberg’s generic `Record` model with a `Void` engine schema. Parquet variant shredding was ineffective on that path because the generic `ParquetFormatModel` did not use a variant shredding analyzer / row copier, and `RecordVariantShreddingAnalyzer` could not resolve VARIANT columns (engine `resolveColumnIndex` is a dead end for `Void`). This PR wires `RecordVariantShreddingAnalyzer` and `Record::copy` into `GenericFormatModels` and analyzes VARIANT columns by Iceberg `Schema#columns()` order so buffered inference and shredded Parquet columns work for Connect. ## Changes - `GenericFormatModels`: register `ParquetFormatModel` with `RecordVariantShreddingAnalyzer` + `Record::copy`. - `RecordVariantShreddingAnalyzer`: implement `analyzeVariantColumns` using positional indices aligned with `Record#get`. ## Config (Connect) Table write properties (e.g. via `iceberg.tables.write-props`): - `write.parquet.shred-variants=true` - `write.parquet.variant-inference-buffer-size=<rows>` ## Test plan - [ ] `./gradlew :iceberg-data:check :iceberg-kafka-connect:iceberg-kafka-connect:check` (or CI green). - [ ] Connect sink writing VARIANT with `write.parquet.shred-variants=true`; inspect Parquet for `typed_value` paths / higher physical column count vs `false`. - [ ] Regression: Connect append with shredding disabled still succeeds. -- 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]
