unikdahal opened a new issue, #6692: URL: https://github.com/apache/datafusion-comet/issues/6692
Part of #6240. Related: #5122, #5649. ## What is the problem? On Spark 3.5+, an Iceberg merge-on-read `DELETE`, `UPDATE` or `MERGE` writes through Spark's `WriteDelta` command. Comet's split Iceberg write path (`IcebergWrite` + `IcebergCommit`) does not recognise it today, so those writes either stay on Spark's stock plan or are caught by checks written for copy-on-write and append writes. There is no explicit, tested statement of what happens to a position-delta write. The native position-delta writer in #6240 needs this first. Before native code can route data rows and position deletes, Comet must recognise `WriteDelta`, extract what a delta write needs (the operation column and the data, row-id and metadata projections), and keep running Iceberg's JVM `DeltaWriter` for it. That makes later native work a swap of the writer behind a known dispatch point rather than a planner change. ## Proposed solution The first item of the incremental split in #6240, with behaviour unchanged for users: - Recognise Iceberg `WriteDelta` on Spark 3.5+; Spark 3.4 keeps the stock plan. - Carry only what the JVM delta path needs: the operation and Spark's projection state, using Spark's `RowDeltaUtils` operation constants instead of copies of the numeric ids. - Run position-delta rows through Iceberg's JVM `DeltaWriter`. The native Iceberg writer stays out of the delta path (`CometIcebergWriteExec` is never used for position-delta rows). - Keep row-level write summaries, writer reporting, AQE behaviour and failure cleanup the same as the stock Iceberg path. - Fall back cleanly when the version-specific logical-plan extraction cannot be done, with a stated reason. ## Scope In scope: planning, extraction and JVM dispatch for position-delta writes, with tests. Out of scope (later items in #6240): a native delta operator or protobuf, a native position-delete writer or file layout contract, loading previous deletes, and a native feature flag. No Rust changes or dependencies. ## Acceptance criteria - `DELETE`, `UPDATE` and `MERGE` results match the stock Iceberg JVM path. - A position-delta write plans without `CometIcebergWriteExec`. - Copy-on-write and merge-on-read `MERGE` plans match the stock path with AQE on and off. - Row-level snapshot summaries match where Spark exposes them. - Writer reporting classifies the intercepted JVM `WriteDelta` write. - Task commit location extraction covers position-delta data and delete files. - Unsupported shapes fall back with a reason. -- 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]
