[
https://issues.apache.org/jira/browse/FLINK-1396?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14306878#comment-14306878
]
ASF GitHub Bot commented on FLINK-1396:
---------------------------------------
Github user StephanEwen commented on the pull request:
https://github.com/apache/flink/pull/363#issuecomment-73014361
Looks good.
I have one suggestion concerning the Hadoop dependencies: The `flink-java`
project depends on the Hadoop API, for the `Writable` interface, the
`NullValue` and the InputFormat classes.
We should be able to exclude all transitive dependencies from the Hadoop
dependency in `flink-java`, making the project more lightweight, so that
someone that only writes against the flink API does not have the long tail of
transitive hadoop dependencies.
Whenever we really execute Hadoop code, we have the flink runtime involved,
which then has the necessary dependencies.
To exclude all transitive dependencies, use
```
<exclusions>
<exclusion>
<groupId>*</groupId>
<artifactId>*</artifactId>
</exclusion>
</exclusions>
```
> Add hadoop input formats directly to the user API.
> --------------------------------------------------
>
> Key: FLINK-1396
> URL: https://issues.apache.org/jira/browse/FLINK-1396
> Project: Flink
> Issue Type: Bug
> Reporter: Robert Metzger
> Assignee: Aljoscha Krettek
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
