Github user fhueske commented on the pull request:
https://github.com/apache/flink/pull/1553#issuecomment-177997324
The `GroupReduceWithCombineProperties.instanciate()` method checks the
shipping strategy of the input channel. In case of the `WordCount` example
*without* explicit hash combiner, the shipping strategy is `PARTITION_HASH` and
the `else` branch will inject a combiner. If you add an explicit partition
operator, the input shipping strategy of the Reduce operator is `FORWARD` and
the `if` branch is executed and does not add a combiner.
Hence the logic has to into the `if` branch and not into the `else` branch.
Or even better add an additional condition to the `if` case
`!(in.getSource().getOptimizerNode() instanceof PartitionNode)` and add an `if
else` branch to handle the special case of the explicit partition operator.
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
