[
https://issues.apache.org/jira/browse/FLINK-15973?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Dian Fu updated FLINK-15973:
----------------------------
Description:
For the following job:
{code}
t_env.register_function("inc", inc)
table.select("inc(id) as inc_id") \
.where("inc_id > 0") \
.insert_into("sink")
{code}
The execution plan is as following:
{code}
StreamExecPythonCalc(select=inc(f0) AS inc_id))
+- StreamExecCalc(select=id AS f0, where=>(f0, 0))
+--- StreamExecPythonCalc(select=id, inc(f0) AS f0))
+-----StreamExecCalc(select=id, id AS f0))
+-------StreamExecTableSourceScan(fields=id)
{code}
The plan is not the best. It could be optimized as following:
{code}
StreamExecPythonCalc(select=f0)
+- StreamExecCalc(select=f0, where=>(f0, 0))
+--- StreamExecPythonCalc(select=inc(f0) AS f0))
+-----StreamExecCalc(select=id, id AS f0))
+-------StreamExecTableSourceScan(fields=id)
{code}
was:
For the following job:
{code}
t_env.register_function("inc", inc)
table.select("inc(id) as inc_id") \
.where("inc_id > 0") \
.insert_into("sink")
{code}
The execution plan is as following:
{code}
StreamExecPythonCalc(select=inc(f0) AS inc_id))
+- StreamExecCalc(select=id AS f0, where=>(f0, 0))
+--- StreamExecPythonCalc(select=id, inc(f0) AS f0))
+-----StreamExecCalc(select=id, id AS f0))
+-------StreamExecTableSourceScan(fields=id)
{code}
The plan is not the best. It should be as following:
{code}
StreamExecPythonCalc(select=f0)
+- StreamExecCalc(select=f0, where=>(f0, 0))
+--- StreamExecPythonCalc(select=inc(f0) AS f0))
+-----StreamExecCalc(select=id, id AS f0))
+-------StreamExecTableSourceScan(fields=id)
{code}
> Optimize the execution plan where it refers the Python UDF result field in
> the where clause
> -------------------------------------------------------------------------------------------
>
> Key: FLINK-15973
> URL: https://issues.apache.org/jira/browse/FLINK-15973
> Project: Flink
> Issue Type: Improvement
> Components: API / Python
> Reporter: Dian Fu
> Priority: Major
> Fix For: 1.11.0
>
>
> For the following job:
> {code}
> t_env.register_function("inc", inc)
> table.select("inc(id) as inc_id") \
> .where("inc_id > 0") \
> .insert_into("sink")
> {code}
> The execution plan is as following:
> {code}
> StreamExecPythonCalc(select=inc(f0) AS inc_id))
> +- StreamExecCalc(select=id AS f0, where=>(f0, 0))
> +--- StreamExecPythonCalc(select=id, inc(f0) AS f0))
> +-----StreamExecCalc(select=id, id AS f0))
> +-------StreamExecTableSourceScan(fields=id)
> {code}
> The plan is not the best. It could be optimized as following:
> {code}
> StreamExecPythonCalc(select=f0)
> +- StreamExecCalc(select=f0, where=>(f0, 0))
> +--- StreamExecPythonCalc(select=inc(f0) AS f0))
> +-----StreamExecCalc(select=id, id AS f0))
> +-------StreamExecTableSourceScan(fields=id)
> {code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
