[
https://issues.apache.org/jira/browse/FLINK-986?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14239712#comment-14239712
]
ASF GitHub Bot commented on FLINK-986:
--------------------------------------
Github user StephanEwen commented on the pull request:
https://github.com/apache/incubator-flink/pull/254#issuecomment-66321956
This will take a bit to double-check. All in all: nice code!
You serialize Strings in most places by writing their byte length and their
bytes: `out.writeInt(msg.getBytes().length);` and
`out.writeBytes(msg.getBytes());`. In most other places, we used either
`out.writeUtf8()`, or preferable `StringUtils.writeNullableString()` or
`StingValue.writeString()`. Also, the above code frequently goes through the
charset encoding code paths multiple, which is a bit unnecessary.
The encoding logic in the `NettyMessage` does a few unnecessary or at least
clumsy steps. The reason seems to be that every message must compute its exact
size before actually putting its contents into the netty buffer. We can
simplify that by allowing the *write()* call to allocate and return the target
buffer by itself. That way, we can also implement code that does not really
copy data into netty buffers, but only wraps existing NIO buffers.
> Add intermediate results to distributed runtime
> -----------------------------------------------
>
> Key: FLINK-986
> URL: https://issues.apache.org/jira/browse/FLINK-986
> Project: Flink
> Issue Type: New Feature
> Components: Distributed Runtime
> Reporter: Ufuk Celebi
> Assignee: Ufuk Celebi
> Priority: Blocker
>
> Support for intermediate results in the runtime is currently blocking
> different efforts like fault tolerance or result collection at the client.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
