JOIN improvements #1439
Closed
joocer
started this conversation in
Improvements
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Following investigations into TPCH benchmark failures, a problem with how cross joins with three or more tables is handled has been found.
This issue is throughout the entire handling of multi way cross joins, starting with the logical planner, as such it will likely be a non-trivial activity to remediate.
A number of issues and improvements have been opened recently relating to joins, this refactoring is an opportunity to address these too.
The intention is to rewrite INNER JOINS as CROSS JOINS, and move the join conditions to be filters. These can then be reapplied to the joins by the optimizer, which can make decisions about a cost-effective order to perform these activities.
This will require some information about the size and selectivity of the relations and columns. This initially will be extrapolated from the first morsel.
Beta Was this translation helpful? Give feedback.
All reactions