Github user gallenvara commented on the pull request:
https://github.com/apache/flink/pull/1838#issuecomment-212300584
@fhueske PR updated.
I am a little confused when i wrote the tests. The original dataset handled
by a `map` operator to ensure that the type of partition key is same with the
boundary in the supplied distribution. In the `DataDistribution` interface, the
type of `getBucketBoundary` method returned is `Object[]`. My doubt is whether
this can be changed to type of `Tuple`. I mean that when range partition by one
field, it return `Tuple1` and two fields return `Tuple2`. Also in the
`OutputEmmiter`, change the type of keys from `Object[]` to `Tuple` and
comparing the key with boundary using `Tuple` comparator. If this is possible,
the boundaries in the distribution for rangePartition test will be:
`Tuple2<Integer, Long>[] boundaries = new Tuple2[]{
new Tuple2(1, 1L),
new Tuple2(3, 2L),
....
}`
This can make the test more succinct and direct.
Another confusing is that why partitionByHash and partitionByRange do not
support some KeySelectors returned Tuple type such as:
```
public static class KeySelector3 implements
KeySelector<Tuple3<Integer,Long,String>, Tuple2<Integer,Long>> {
private static final long serialVersionUID = 1L;
@Override
public Tuple2<Integer,Long> getKey(Tuple3<Integer,Long,String>
in) {
return new Tuple2<>(in.f0,in.f1);
}
}
```
and can not run the following codes:
```
DataSet<Tuple3<Integer,Long,String>> dataSet = ...;
dataSet.partitionByRange(new KeySelector3());
```
Can you explain it to me?Thanks!
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
