andygrove opened a new pull request, #4149:
URL: https://github.com/apache/datafusion-comet/pull/4149

   ## Which issue does this PR close?
   
   Closes #.
   
   ## Rationale for this change
   
   Add native support for Spark's `slice(array, start, length)` expression so 
it runs on Comet instead of falling back to Spark.
   
   The `datafusion-spark` crate already ships a `SparkSlice`, but it is not 
Spark-compatible: when a negative `start` lies before the beginning of the 
array (e.g. `slice([a], -2, 2)`), it returns the first element instead of an 
empty array. We can upstream the fix later; for now this PR ships a Comet-local 
implementation.
   
   ## What changes are included in this PR?
   
   - `native/spark-expr/src/array_funcs/array_slice.rs`: new `SparkArraySlice` 
UDF (`spark_array_slice`) implementing Spark's slice semantics, including 
1-based indexing, negative-start-from-end, error on `start = 0` or `length < 
0`, and clamping length to the array end. Supports both `List` and `LargeList` 
element storage.
   - `native/spark-expr/src/comet_scalar_funcs.rs`: register the new UDF.
   - `spark/src/main/scala/org/apache/comet/serde/arrays.scala`: `CometSlice` 
serde casts the start/length args to `Long` and serialises a call to 
`spark_array_slice`, promising `containsNull = true` to match DataFusion's list 
nullability.
   - `spark/src/main/scala/org/apache/comet/serde/QueryPlanSerde.scala`: 
register `Slice` in `arrayExpressions`.
   
   ## How are these changes tested?
   
   - 12 native unit tests in `array_slice.rs` covering positive / negative / 
zero / overflowing start, length 0, length past end, null inputs, empty arrays, 
and the error cases.
   - New SQL test file 
`spark/src/test/resources/sql-tests/expressions/array/slice.sql` covering 
all-literal, column + literal, and column-only argument combinations across 
boolean, tinyint, smallint, int, bigint, float, double, decimal, date, 
timestamp, timestamp_ntz, string, and nested array element types, plus the 
negative-start-overflow case that exposed the upstream bug.


-- 
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