opt: support lookup anti/semi join#38625
Conversation
|
This change is |
justinj
left a comment
There was a problem hiding this comment.
the 1-line change (+ tests) in the opt codebase lgtm :) should we consider tagging in someone from exec? I think @jordanlewis shepherded the original PR for this
Reviewable status:
complete! 0 of 0 LGTMs obtained (waiting on @andy-kimball and @justinj)
jordanlewis
left a comment
There was a problem hiding this comment.
Reviewable status:
pkg/sql/distsqlpb/processors_sql.proto, line 181 at r1 (raw file):
// If populated (lookup join), the `joinReader` will perform a lookup join // and the "internal columns" will be the concatenation of the input stream // columns followed by the table columns (except for semi/anti join).
nit: might as well spell it out here. (except for semi/anti join, which don't output any table columns).
Enable support for lookup anti/semi joins. Some adjustments were needed in the distsql physical planner and the library, but nothing in the optimizer itself. Verified that the TPCH query that changed is faster (~0.9s vs ~1.3s on single node SF1). Release note: None
0b451e6 to
9e9555b
Compare
RaduBerinde
left a comment
There was a problem hiding this comment.
TFTRs!
bors r+
Reviewable status:
38625: opt: support lookup anti/semi join r=RaduBerinde a=RaduBerinde Enable support for lookup anti/semi joins. Some adjustments were needed in the distsql physical planner and the library, but nothing in the optimizer itself. Verified that the TPCH query that changed is faster (~0.9s vs ~1.3s on single node SF1). Release note: None Co-authored-by: Radu Berinde <radu@cockroachlabs.com>
Build succeeded |
Enable support for lookup anti/semi joins. Some adjustments were
needed in the distsql physical planner and the library, but nothing in
the optimizer itself.
Verified that the TPCH query that changed is faster (~0.9s vs ~1.3s on
single node SF1).
Release note: None