[
https://issues.apache.org/jira/browse/CALCITE-4143?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17166235#comment-17166235
]
Jiatao Tao commented on CALCITE-4143:
-------------------------------------
Hi [~julianhyde] [~liupengcheng]
A case I met like that(Simplified, just for explanation, the SQL is
meaningless):
SQL:
{code:java}
SELECT
u.name
from
USERS u
LEFT JOIN JOBS j ON u.ID = j.ID
and j.name = 'Bob'
WHERE
u.name = 'Alice'
{code}
Optimized plan:
{code:java}
1 EnumerableProject(name=[$1])
2 EnumerableHashJoin(condition=[=($0, $2)], joinType=[left])
3 EnumerableProject(ID=[$0], NAME=[$1])
4 EnumerableFilter(condition=[=($1, _UTF-8'Alice')])
5 EnumerableTableScan(table=[[default_ns, RECR, USERS]])
6 EnumerableProject(ID=[$0], NAME=[$1])
7 EnumerableFilter(condition=[=($1, _UTF-8'Bob')])
8 EnumerableTableScan(table=[[default_ns, RECR, JOBS]])
{code}
See line 6, we only use job#name as a filter, we do not need to project that
after the filter.
After debugging, I found at first, the filter is above the prj.
> Trim fields after XXProjectTransposeRule
> ----------------------------------------
>
> Key: CALCITE-4143
> URL: https://issues.apache.org/jira/browse/CALCITE-4143
> Project: Calcite
> Issue Type: Improvement
> Components: core
> Reporter: Jiatao Tao
> Assignee: Jiatao Tao
> Priority: Major
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
