[
https://issues.apache.org/jira/browse/FLINK-2579?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14716855#comment-14716855
]
Arnaud Linz commented on FLINK-2579:
------------------------------------
I understand, it may not be such a good idea.
What I had in mind was :
// Batch
public ExecutionConfig getConfig()
public int getParallelism()
public void setParallelism(int parallelism)
public void setNumberOfExecutionRetries(int numberOfExecutionRetries)
public int getNumberOfExecutionRetries()
public void addDefaultKryoSerializer(Class<?> type, Serializer<?> serializer)
public void addDefaultKryoSerializer(Class<?> type, Class<? extends
Serializer<?>> serializerClass)
public void registerTypeWithKryoSerializer(Class<?> type, Serializer<?>
serializer)
public void registerTypeWithKryoSerializer(Class<?> type, Class<? extends
Serializer<?>> serializerClass)
public void registerType(Class<?> type)
public JobExecutionResult execute()
public JobExecutionResult execute(String paramString)
// And maybe
public String getExecutionPlan()
public static LocalEnvironment createLocalEnvironment()
public static LocalEnvironment createLocalEnvironment(int parallelism)
public static void setDefaultLocalParallelism(int parallelism)
// Stream
public ExecutionConfig getConfig()
public int getParallelism()
public StreamExecutionEnvironment setParallelism(int parallelism)
public void setNumberOfExecutionRetries(int numberOfExecutionRetries)
public int getNumberOfExecutionRetries()
public void addDefaultKryoSerializer(Class<?> type, Serializer<?> serializer)
public void addDefaultKryoSerializer(Class<?> type, Class<? extends
Serializer<?>> serializerClass)
public void registerTypeWithKryoSerializer(Class<?> type, Serializer<?>
serializer)
public void registerTypeWithKryoSerializer(Class<?> type, Class<? extends
Serializer<?>> serializerClass)
public void registerType(Class<?> type)
public JobExecutionResult execute()
public JobExecutionResult execute(String paramString)
public String getExecutionPlan()
// And maybe
public static LocalStreamEnvironment createLocalEnvironment()
public static LocalStreamEnvironment createLocalEnvironment(int parallelism)
public static void setDefaultLocalParallelism(int parallelism)
but it's a small set of API that thus it may not worth the effort. However, one
should try to make a few rules on API syntax to help users, for instance decide
whether "setXXX" returns void or this, it seems that "this" is the exception on
the StreamExecutionEnvironment :-)
> StreamExecutionEnvironment & ExecutionEnvironment do not share interface but
> have a lot in common
> -------------------------------------------------------------------------------------------------
>
> Key: FLINK-2579
> URL: https://issues.apache.org/jira/browse/FLINK-2579
> Project: Flink
> Issue Type: Improvement
> Components: Core
> Reporter: Arnaud Linz
> Priority: Minor
> Original Estimate: 24h
> Remaining Estimate: 24h
>
> Both classes
> org.apache.flink.streaming.api.environment.StreamExecutionEnvironment and
> org.apache.flink.api.java.ExecutionEnvironment have a lot in common (same
> methods for kryo registration, fromCollection, etc) but are not related by a
> java contract.
> That leads to annoying differences, for instance :
> StreamExecutionEnvironment.setParallelism() returns 'this' where as
> ExecutionEnvironment.setParallelism() has not return value.
> They have specificities, but maybe they should both implement a common
> Interface to make sure that the common signatures are coherent?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
