datafusion
datafusion copied to clipboard
POC: Reduce `Arc` cloning on hashmap build side
Which issue does this PR close?
- Closes #.
Rationale for this change
What changes are included in this PR?
Are these changes tested?
Are there any user-facing changes?
I've noticed that it is possible for interleave to perform worse than take despite the Arc clones from take. This happens twice as well for equal_row_arr and build_batch_from_indices.
I've noticed that it is possible for
interleaveto perform worse thantakedespite theArcclones fromtake. This happens twice as well forequal_row_arrandbuild_batch_from_indices.
Yes that will be the tricky part. We gain some speed during probing by concatenating the left side (which itself can be slow) into a singe batch.