Aleksey Plekhanov created IGNITE-27489:
------------------------------------------

             Summary: Calcite engine. Merge projects into join
                 Key: IGNITE-27489
                 URL: https://issues.apache.org/jira/browse/IGNITE-27489
             Project: Ignite
          Issue Type: Improvement
            Reporter: Aleksey Plekhanov
            Assignee: Aleksey Plekhanov


It's an common case when project rel node is placed on top of join node, since 
almost always we skip some columns from join inputs. For example, for join like:
{code:java}
SELECT ... FROM emp JOIN dept ON emp.deptno = dept.deptno
{code}
Only one {{deptno}} column is enough for user.
Join constructs row as concatination of both left and right hand, and project 
rel node construct yet another row. We can optimize such an approach, merge 
project node into join node and construct only one row based on two inputs. We 
using the same approach for filtering (see ExpressionFactoryImpl#biPredicate)



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to