[
https://issues.apache.org/jira/browse/FLINK-13081?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
godfrey he updated FLINK-13081:
-------------------------------
Description:
in flink planner, a query will be optimized while calling
{{TableEnvironment#insertInto}} or {{TableEnvironment#sqlUpdate}}. however, if
a job has multiple sinks (means {{insertInto}} or {{sqlUpdate}} will be called
multiple times), the final job contains several independent sub-graphs. In most
cases, there is duplicate computing in a multiple sinks job. so in blink
planner, multiple sinks queries will be optimized together to avoid duplicate
computing. a query will not be optimized in {{insertInto}} and {{sqlUpdate}}.
instead, queries will be optimized before executing.
this issue aims to support above case.
two methods will be added into {{TableEnvironment}}:
{code:java}
// explain multiple-sinks plan
String explain(boolean extended);
// Triggers the program execution
// in blink planner, queries will be optimized together in this method
JobExecutionResult execute(String jobName) throws Exception;
{code}
to make sure the behavior of flink planner is same as before, a
{{isLazyOptMode}} filed is added into {{EnvironmentSettings}}, which tells the
table environment should optimize the query immediately in
{{insertInto}}/{{sqlUpdate}} methods for flink planner({{isLazyOptMode}} is
always false) or in execute method for blink planner({{isLazyOptMode}} is
always true) .
was:
in flink planner, a query will be optimized while calling
{{TableEnvironment#insertInto}} or {{TableEnvironment#sqlUpdate}}. however, if
a job has multiple sinks (means {{insertInto}} or {{sqlUpdate}} will be called
multiple times), the final job contains several independent sub-graphs. In most
cases, there is duplicate computing in a multiple sinks job. so in blink
planner, multiple sinks queries will be optimized together to avoid duplicate
computing. a query will not be optimized in {{insertInto}} and {{sqlUpdate}}.
instead, queries will be optimized before executing.
this issue aims to support above case.
two methods will be added into {{TableEnvironment}}:
{code:java}
// explain multiple-sinks plan
String explain(boolean extended);
// Triggers the program execution
// in blink planner, queries will be optimized together in this method
JobExecutionResult execute(String jobName) throws Exception;
{code}
to make sure the behavior of flink planner is same as before, a
{{isLazyOptMode}} filed is added into {{EnvironmentSettings}}, which tells the
table environment should optimize the query immediately in
{{insertInto}}/{{sqlUpdate}} methods(isLazyOptMode=false) for flink planner or
in execute method for blink planner.
> Supports explain and execute DAG plan
> -------------------------------------
>
> Key: FLINK-13081
> URL: https://issues.apache.org/jira/browse/FLINK-13081
> Project: Flink
> Issue Type: New Feature
> Components: Table SQL / Planner
> Reporter: godfrey he
> Assignee: godfrey he
> Priority: Major
>
> in flink planner, a query will be optimized while calling
> {{TableEnvironment#insertInto}} or {{TableEnvironment#sqlUpdate}}. however,
> if a job has multiple sinks (means {{insertInto}} or {{sqlUpdate}} will be
> called multiple times), the final job contains several independent
> sub-graphs. In most cases, there is duplicate computing in a multiple sinks
> job. so in blink planner, multiple sinks queries will be optimized together
> to avoid duplicate computing. a query will not be optimized in {{insertInto}}
> and {{sqlUpdate}}. instead, queries will be optimized before executing.
> this issue aims to support above case.
> two methods will be added into {{TableEnvironment}}:
> {code:java}
> // explain multiple-sinks plan
> String explain(boolean extended);
> // Triggers the program execution
> // in blink planner, queries will be optimized together in this method
> JobExecutionResult execute(String jobName) throws Exception;
> {code}
> to make sure the behavior of flink planner is same as before, a
> {{isLazyOptMode}} filed is added into {{EnvironmentSettings}}, which tells
> the table environment should optimize the query immediately in
> {{insertInto}}/{{sqlUpdate}} methods for flink planner({{isLazyOptMode}} is
> always false) or in execute method for blink planner({{isLazyOptMode}} is
> always true) .
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
