[
https://issues.apache.org/jira/browse/SPARK-25369?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16649495#comment-16649495
]
Sean Owen commented on SPARK-25369:
-----------------------------------
This won't work. The problem is that an API function accepting, say,
{{java.util.function.Function}} is not {{Serializable}} . There are easy ways
for the caller to force it to be so, but otherwise, callers will find their
tasks will never serialize.
I think we have to keep our interface functions, which all {{extend
Serializable}}. The upside is that this is a good reason not to change this API
at all.
We can however retrofit the interfaces to also extend their Java counterparts,
I believe. This might buy a little more interoperability for free.
> Replace Java shim functional interfaces like java.api.Function with Java 8
> equivalents
> --------------------------------------------------------------------------------------
>
> Key: SPARK-25369
> URL: https://issues.apache.org/jira/browse/SPARK-25369
> Project: Spark
> Issue Type: Improvement
> Components: Spark Core
> Affects Versions: 2.4.0
> Reporter: Sean Owen
> Priority: Minor
> Labels: release-notes
>
> In Spark 3, we should remove interfaces like
> org.apache.spark.api.java.function.Function and replace with
> java.util.function equivalents, for better compatibility with Java 8. This
> would let callers pass, in more cases, an existing functional object in Java
> rather than wrap in a lambda.
> It's possible to have the functional interfaces in Spark just extend Java 8
> functional interfaces to interoperate better with existing code, but might be
> as well to remove them in Spark 3 to clean up.
> A partial list of transitions from Spark to Java interfaces:
> * Function -> Function
> * Function0 -> Supplier
> * Function2 -> BiFunction
> * VoidFunction -> Consumer
> * FlatMapFunction etc -> extends Function<T,Iterable<R>> etc
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
