Github user tdas commented on the pull request:
https://github.com/apache/spark/pull/2940#issuecomment-60877622
I mentioned this in the earlier in the original thread. This is a tradeoff
between generality and type checking. I want the code in
ReceiverSupervisorImpl, and ReceivedBlockInfo to be agnostic to the
implementation of the ReceivedBlockHandler that is in use. Otherwise no point
of making this a pluggable interface, where we could plug in implementation
like CassandraBasedBlockHandler in future.
To achieve that these classes have to be agnostic to the exact return type
of the ReceivedBlockHandler. In which case there are three options as far as I
think.
1. Refer to the handler as ReceivedBlockHandler[Any]
2. Refer to the handler as ReceivedBlockHandler[_]
3. Give the return type a trait name (say ReceivedBlockHandlerStoreResult)
and refer to the handler as
ReceivedBlockHandler[ReceivedBlockHandlerStoreResult]
I dont see much advantage in 2 over 1. I thought of 3 as an option. But it
gets weird for the default BlockManagerBasedBlockHandler, as it always returns
None, and so does not really need the type. But 3 will force a dummy type (that
extends ReceivedBlockHandlerStoreResult) to be defined.
Thoughts?
---
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.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
