Sergey Nuyanzin created FLINK-33485:
---------------------------------------
Summary: Optimize the EXISTS sub-query by Metadata RowCount
Key: FLINK-33485
URL: https://issues.apache.org/jira/browse/FLINK-33485
Project: Flink
Issue Type: Improvement
Components: Table SQL / API
Affects Versions: 1.18.0
Reporter: Sergey Nuyanzin
If the sub-query is guaranteed to produce at least one row, just return TRUE.
If the sub-query is guaranteed to produce no row, just return FALSE.
inspired by CALCITE-5117 however since there is {{FlinkSubQueryRemoveRule}}
then it shold be adopted accordingly
examples
{code:sql}
SELECT * FROM T2 WHERE EXISTS (SELECT SUM(a1), COUNT(*) FROM T1 WHERE 1=2)
{code}
aggregation functions always return 1 row even if there is an empty table then
we could just replace this query with
{code:sql}
SELECT * FROM T2
{code}
another example
{code:sql}
SELECT * FROM MyTable WHERE NOT EXISTS (SELECT a FROM MyTable LIMIT 0)
{code}
{{LIMIT 0}} means no rows so it cold be optimized to
{code:sql}
SELECT * FROM MyTable
{code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
