Github user jkovacs commented on the pull request:
https://github.com/apache/flink/pull/1138#issuecomment-141497981
Thanks @fhueske, that's a good point I haven't considered.
Another idea that occurred to me was to convert the result tuple types to
`GenericTypeInfo<T>` (instead of `GenericTypeInfo<Object>`), where `T` is the
original type of the tuple field (e.g. `String` or `Integer`). This would be
null safe _and_ would allow the user to group by those fields, assuming of
course they are sure that the fields are non-null (e.g. on a left or right
outer join).
Although I'm not sure of all the consequences of using, say,
`GenericTypeInfo<String>` instead of `BasicTypeInfo<String>` for serialization
and comparison.
I pushed this change as
https://github.com/jkovacs/flink/commit/f682baa50137e0a54bae091ba60ba85fdb8f4c1b
to a different branch to test it
Also rebased branch onto current master and resolved conflicts (Failing
test is some YARN integration test).
---
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.
---
