[
https://issues.apache.org/jira/browse/FLINK-3589?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15187857#comment-15187857
]
Greg Hogan commented on FLINK-3589:
-----------------------------------
I had a go at implementing this across the codebase using a new constant in
{{ExecutionConfig}}, as {{ExecutionConfig}} already contains
{{PARALLELISM_AUTO_MAX}} and is in flink-core so is visible from other modules.
> Allow setting Operator parallelism to default value
> ---------------------------------------------------
>
> Key: FLINK-3589
> URL: https://issues.apache.org/jira/browse/FLINK-3589
> Project: Flink
> Issue Type: Improvement
> Components: Java API
> Affects Versions: 1.1.0
> Reporter: Greg Hogan
> Assignee: Greg Hogan
> Priority: Minor
>
> User's can override the parallelism for a single operator by calling
> {{Operator.setParallelism}}, which accepts a positive value. {{Operator}}
> uses {{-1}} to indicate default parallelism. It would be nice to name and
> accept this default value.
> This would enable user algorithms to allow configurable parallelism while
> still chaining operator methods.
> For example, currently:
> {code}
> private int parallelism;
> ...
> public void setParallelism(int parallelism) {
> this.parallelism = parallelism;
> }
> ...
> MapOperator<Edge<K,LongValue>, Edge<K,IntValue>> newEdges =
> edges
> .map(new MyMapFunction<K>())
> .name("My map function");
> if (parallelism > 0) {
> newEdges.setParallelism(parallelism);
> }
> {code}
> Could be simplified to:
> {code}
> private int parallelism = Operator.DEFAULT_PARALLELISM;
> ...
> public void setParallelism(int parallelism) {
> this.parallelism = parallelism;
> }
> ...
> DataSet<Edge<K,IntValue>> newEdges = edges
> .map(new MyMapFunction<K>())
> .setParallelism(parallelism)
> .name("My map function");
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
