andygrove opened a new pull request, #7053: URL: https://github.com/apache/datafusion-comet/pull/7053
## Which issue does this PR close? Closes #3427. ## Rationale for this change Spark's `ParquetWriteSupport` records the Spark version in each file's footer as `org.apache.spark.version`. Spark's datetime rebase logic reads it back to tell files written by Spark 3.0+ from older ones, and `ParquetIOSuite` asserts that it is present. The native Parquet writer built its writer properties with only the compression codec, so its files had no such key and "Write Spark version into Parquet metadata" failed when the native writer was enabled. ## What changes are included in this PR? - Add an optional `spark_version` field to the `ParquetWriter` proto message. - Both Scala serde paths (`CometDataWritingCommand` for Spark 3.x and `CometWriteFiles` for Spark 4.0+) set it to `SPARK_VERSION_SHORT`. - `ParquetWriterExec` writes it to the footer as `org.apache.spark.version` when present, and keeps it across `with_new_children`. A plan serialized without the field writes no key, as before. ## How are these changes tested? - New Rust test `test_parquet_writer_records_spark_version_in_footer` covers both the present and absent cases. - New Scala test `parquet write records the Spark version in the footer` in `CometParquetWriterSuite` does a native write and checks the footer against `SPARK_VERSION_SHORT`. It passes on the default Spark profile. - The native `parquet_writer` tests, clippy and `cargo fmt --check` are clean. - I have not run the Spark SQL test suite or the Spark 3.4/3.5 profiles. -- 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]
