LuciferYang opened a new pull request, #12890:
URL: https://github.com/apache/gluten/pull/12890

   `SUPPORTED_SQL_QUERY_LIST` in `gluten-ut/spark33`'s 
`VeloxSQLQueryTestSettings` lists ten `.sql` files that no surviving module 
lists. One of them, `udf/udf-window.sql`, is in that module's own `ignoreList` 
and never runs, so nine files stop running when the module is removed under 
#12807:
   
   ```
   ansi/datetime-special.sql
   ansi/parse-schema-string.sql
   udf/udf-count.sql
   udf/udf-cross-join.sql
   udf/udf-having.sql
   udf/udf-inner-join.sql
   udf/udf-join-empty-relation.sql
   udf/udf-natural-join.sql
   udf/udf-outer-join.sql
   ```
   
   ### Evidence
   
   All nine pass on Spark 3.4.4 with Velox. I ran `GlutenSQLQueryTestSuite` 
locally with these entries added and `spark.test.home` pointed at a 3.4.4 
source tree: 222 tests, 0 failures. The nine show up as:
   
   ```
   - ansi/datetime-special.sql
   - ansi/parse-schema-string.sql
   - udf/udf-count.sql - Scala UDF               - udf/udf-count.sql - Regular 
Python UDF
   - udf/udf-cross-join.sql - Scala UDF          - udf/udf-cross-join.sql - 
Regular Python UDF
   - udf/udf-having.sql - Scala UDF              - udf/udf-having.sql - Regular 
Python UDF
   - udf/udf-inner-join.sql - Scala UDF          - udf/udf-inner-join.sql - 
Regular Python UDF
   - udf/udf-join-empty-relation.sql - Scala UDF - 
udf/udf-join-empty-relation.sql - Regular Python UDF
   - udf/udf-natural-join.sql - Scala UDF        - udf/udf-natural-join.sql - 
Regular Python UDF
   - udf/udf-outer-join.sql - Scala UDF          - udf/udf-outer-join.sql - 
Regular Python UDF
   ```
   
   The Scalar Pandas UDF variant of each udf file is skipped on my machine 
because pyspark, pandas and pyarrow are not installed there. Draft until CI 
reports on 3.5, 4.0 and 4.1: the golden `.out` files come from each Spark 
version's own resources and I have only checked 3.4.
   
   ### Two things about the entry format
   
   The two sides of the comparison use different matching, so a plain diff of 
the two lists is misleading. `gluten-ut/spark33` matches with `contains`, while 
3.4+ switched to exact names (`Modified for Gluten to use exact name 
matching`), and a UDF sql file expands into one test per UDF kind. The new 
entries therefore use the `... - Scala UDF` form that the surviving modules 
already use for their udf files. Listing only the Scala UDF name still runs the 
Regular Python UDF variant, as the output above shows.
   
   This also drops two entries from `gluten-ut/spark34`:
   
   ```scala
   "udf-union.sql",
   "udf-window.sql",
   ```
   
   They are bare file names, and under exact matching the test names are 
`udf/udf-union.sql - Scala UDF` and so on, so neither can ever match. 
`gluten-ut/spark35` carries the correct form for `udf-union.sql`; spark34 is 
the only module with the broken shape.
   
   Found while inventorying what only runs on `gluten-ut/spark33` before 
removing it. Tracked in #12863.
   
   ## Was this patch authored or co-authored using generative AI tooling?
   
   Generated-by: Claude claude-opus-5
   


-- 
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]

Reply via email to