rdtr opened a new pull request, #12146:
URL: https://github.com/apache/gluten/pull/12146
Description:
## Summary
- Enable 7 of 11 TODO SQL query test files in spark41
`VeloxSQLQueryTestSettings` (#11916)
- 6 tests were already passing on current main — they just needed
verification and uncomment
- Fix golden file for `type-coercion-edge-cases.sql`: Spark 4.1 changed
`MAP` field access behavior — `col2.field` on
`MAP<BIGINT, DOUBLE>` now uses semi-structured field extraction (returns
`struct<field:double>`) instead of attempting
`CAST("field" AS BIGINT)` which threw `CAST_INVALID_INPUT` in ANSI mode
### Enabled tests
| File | What was needed |
|------|----------------|
| cast.sql | Already passing — uncomment TODO |
| describe.sql | Already passing — uncomment TODO |
| literals.sql | Already passing — uncomment TODO |
| nonansi/cast.sql | Already passing (imports cast.sql) — uncomment TODO |
| nonansi/literals.sql | Already passing (imports literals.sql) —
uncomment TODO |
| nonansi/st-functions.sql | Already passing — uncomment TODO |
| type-coercion-edge-cases.sql | Golden file update for Spark 4.1 behavior
change |
### Remaining 4 TODOs (triaged, not in this PR)
- `subquery/exists-subquery/exists-orderby-limit.sql` -> ANSI fallback
causes `INTERNAL_ERROR` on Spark 4.1
`OneRowRelationExec` null session
- `scripting/randomly_generated_scripts.sql` -> NPE in SQL scripting
exception handler condition handling
- `st-functions.sql` -> Geospatial types (GEOGRAPHY/GEOMETRY) new in Spark
4.1, not supported in Velox
- `variant-field-extractions.sql` -> Variant type not supported (#11371)
I will check them too.
## Test plan
- [ ] CI: spark-test-spark41-slow (runs `GlutenSQLQueryTestSuite` with
`ExtendedSQLTest` tag)
--
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]
