Jacob Jona Fahlenkamp created FLINK-36554:
---------------------------------------------
Summary: Support MapState in Operator State
Key: FLINK-36554
URL: https://issues.apache.org/jira/browse/FLINK-36554
Project: Flink
Issue Type: New Feature
Reporter: Jacob Jona Fahlenkamp
Since Flink currently does not support preaggregation datastream api, I would
like to do it myself using a ProcessFunction. But Operator State does not
support a MapState. I can use a hashmap, but it doesn't work if the map becomes
too large for memory. Here MapState in Operator state would help, as I could
use rocksdb and spill to disk. Is there any particular reason why Operator
State only supports list state?
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
