andygrove opened a new pull request, #4048:
URL: https://github.com/apache/datafusion-comet/pull/4048
## Which issue does this PR close?
Closes #2946.
## Rationale for this change
The Spark 4.0 `ShimSparkErrorConverter` translates native FileNotFound
errors into a `SparkFileNotFoundException` with error class
`_LEGACY_ERROR_TEMP_2055`. That error class exists in Spark 3.4 and 3.5 but was
removed in Spark 4.0, so Spark's `ErrorClassesJsonReader` throws an
`INTERNAL_ERROR` when it tries to look up the message template ("Cannot find
main error class '_LEGACY_ERROR_TEMP_2055'").
This manifests as test failures on
`org.apache.spark.sql.hive.HiveMetadataCacheSuite` (and any other suite that
uses `checkError`/`checkErrorMatchPVals` against
`FAILED_READ_FILE.FILE_NOT_EXIST`) when run against Spark 4.0 with Comet
enabled:
```
null did not equal "FAILED_READ_FILE.FILE_NOT_EXIST"
```
with underlying cause:
```
SparkException: [INTERNAL_ERROR] Cannot find main error class
'_LEGACY_ERROR_TEMP_2055'
```
The previously ignored `sql_hive-1` CI job for Spark 4.0 (issue #2946)
surfaces these failures once the job is re-enabled.
## What changes are included in this PR?
In the Spark 4.0 shim, delegate to
`QueryExecutionErrors.fileNotExistError(path, cause)`, which is the 4.0
replacement for `readCurrentFileNotFoundError`. It produces a `SparkException`
with the expected error class `FAILED_READ_FILE.FILE_NOT_EXIST` and `path`
parameter that tests assert on. The 3.4 and 3.5 shims are unchanged.
## How are these changes tested?
Covered by the existing Spark test
`org.apache.spark.sql.hive.HiveMetadataCacheSuite` (run in the `sql_hive-1` CI
job for Spark 4.0), which previously failed on:
- `SPARK-16337 temporary view refresh`
- `view refresh`
- `partitioned table is cached when partition pruning is true`
- `partitioned table is cached when partition pruning is false`
Re-enabling that job will be handled separately.
--
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]
