Github user StephanEwen commented on the pull request:
https://github.com/apache/incubator-flink/pull/203#issuecomment-64698910
I like the direction of this pull request. Here are some thought what we
could improve:
- Keeping the UnknownTypeInfo all the way and actually trying to work
with it will shadow problems. At much later points, when someone tries to use a
POJO field on the type info, an error will occur, at a point that is
disconnected from the source of the problem.
- How about we rename the `UnknownTypeInfo` to `MissingTypeInfo` and use
it only as a place-holder for the `returns()` hint method, which exchanges that
type info by the proper one.
- We could make sure that type extraction failures are recognized early,
by having the `getType()` method on the DataSet throw an error if only a
`MissingTypeInfo` exists.
- Getting Generic Lambdas to run with this may not be a good idea. This
goes against my original suggestion to keep this for the purpose of local
development with OracleJDK, but in hind-sight, I think things will get quite
intransparent to reason about and hard to maintain/debug/fix.
---
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.
---
