Gabriel39 opened a new pull request, #66206: URL: https://github.com/apache/doris/pull/66206
## Proposed changes - support reading Iceberg V3 Variant columns from Parquet through File Scanner V2 - reject unsupported Variant reads through legacy scanners and non-Parquet file formats - preserve decoded Parquet Variant physical columns in a `ColumnVariantV2` shredded state - reuse shredded typed-value leaf columns directly for Variant path expressions without reconstructing canonical Variant rows - retain late unshredding for operations that require complete Variant values - cover schema evolution, time travel, expression evaluation, mixed file formats, and typed-value page pruning in the Iceberg regression suite ## Why Reconstructing every scanned Variant row through `VariantBatchBuilder` copies data that is already available in shredded Parquet columns. Keeping the physical tree alive lets typed path predicates and expressions consume the decoded leaf column directly while preserving a compatible fallback for unshredded paths. ## Validation - relevant BE production and test objects compiled successfully - 49 related BE unit tests passed, covering `ColumnVariantV2`, Parquet Variant materialization, and Variant element extraction - Iceberg regression suites explicitly enable `ENABLE_VARIANT_V2` -- 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]
