andygrove opened a new issue, #5693:
URL: https://github.com/apache/datafusion-comet/issues/5693
### Describe the bug
After an identity partition field is dropped and a column is re-added, the
native writer cannot resolve the partition source column and fails the write:
```
org.apache.comet.CometNativeException: Unexpected => No column with source
column id 2 in schema
Schema { struct: StructType { fields: [NestedField { id: 1, name: "id",
required: true,
field_type: Primitive(Long), ... } ...
at org.apache.comet.Native.executePlan(Native Method)
at
org.apache.spark.sql.comet.CometIcebergWriteExec.drainAvroPayload(CometIcebergWriteExec.scala:276)
at
org.apache.spark.sql.comet.CometIcebergWriteExec.$anonfun$doExecute$9(CometIcebergWriteExec.scala:181)
```
The partition spec still references source column id 2, which no longer
exists in the current schema, and the native manifest encoding treats that as
fatal. iceberg-java tolerates it.
### Steps to reproduce
Spark 4.1.3, Iceberg 1.11.0, `spark.comet.iceberg.write.enabled=true` plus
the Iceberg Spark SQL test setup from `dev/diffs/iceberg/1.11.0.diff`:
```
./gradlew -DsparkVersions=4.1 -DscalaVersion=2.13 -DflinkVersions=
-DkafkaVersions= \
:iceberg-spark:iceberg-spark-extensions-4.1_2.13:test \
--tests '*TestAlterTablePartitionFields*' -Pquick=true -x javadoc
```
Failing case: `testReaddColumnAfterIdentityPartitionDrop`.
### Expected behavior
The write succeeds against a table whose partition spec references a source
column that is no longer in the current schema, matching iceberg-java, or the
operator declines the write at planning time so it falls back to the JVM writer
instead of failing the task.
### Additional context
Found by turning the two Iceberg write flags on by default in #5677.
Same test class and same `encode_data_files_as_manifest` phase as #5691,
which panics on a related partition-spec-evolution case. One fix to how the
writer resolves the spec against the current schema may close both.
Part of #5649.
--
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]
