jonathanc-n commented on code in PR #16716:
URL: https://github.com/apache/datafusion/pull/16716#discussion_r2194904410
##########
datafusion/physical-plan/src/joins/utils.rs:
##########
@@ -928,6 +929,55 @@ pub(crate) fn build_batch_from_indices(
Ok(RecordBatch::try_new(Arc::new(schema.clone()), columns)?)
}
+/// Returns a new [RecordBatch] resulting of a join where the build/left side
is empty.
+/// The resulting batch has [Schema] `schema`.
+pub(crate) fn build_batch_empty_build_side(
+ schema: &Schema,
+ build_batch: &RecordBatch,
+ probe_batch: &RecordBatch,
+ column_indices: &[ColumnIndex],
+ join_type: JoinType,
+) -> Result<RecordBatch> {
+ match join_type {
+ // these join types only return data if the left side is not empty, so
we return an
+ // empty RecordBatch
+ JoinType::Inner
Review Comment:
> Thinking about this, I think a more generic version of this would be
switching small left sides (e.g < 10 rows) to using cross join 🤔
Is this including for equijoin conditions? I think the performance seemed
slow when there was a larger right table for doing this with nested loop join
which follows a similar algorithm. It is probably a memory issue due to the
cartesian product.
--
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]
