Github user ChengXiangLi commented on the pull request:
https://github.com/apache/flink/pull/1161#issuecomment-161501441
Hi, @StephanEwen , is there any progress of Managed Memory Allocation
abstractions for UDF? Not only about TopK operator, i think it's also very
important for users or DSLs to build more robust and efficient applications.
For example, in Table API queries, as the data schema is predictable during
each phase of processing, we does not need to create real `Row` object, just
store the binary data in self managed memory, and use the offset to read `Row`
fields. So all the intermediate data is store as binary on self managed memory,
no need to create lots of `Row` object and its fields object anymore, which
should be more robust, memory-efficient, and with better performance.
---
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.
---
