banmoy commented on a change in pull request #8611: [FLINK-12693][state] Store
state per key-group in CopyOnWriteStateTable
URL: https://github.com/apache/flink/pull/8611#discussion_r300222482
##########
File path:
flink-runtime/src/main/java/org/apache/flink/runtime/state/heap/StateTable.java
##########
@@ -39,7 +47,8 @@
* @param <N> type of namespace
* @param <S> type of state
*/
-public abstract class StateTable<K, N, S> implements StateSnapshotRestore {
+public abstract class StateTable<K, N, S>
Review comment:
I prefer to use the inheritance for the following reasons:
1. in the followup spill implementation, we need some memory statistics
about the `StateTable` to decide whether to build a on-heap map or a on-disk
map. I think it's not convenient for `MapAndSnapshotFactory ` to communicate
with `StateTable`
2. implementation of `StateTable` may need it's own method, such as
`getStateMapSnapshotArray` in `CopyOnWriteStateTable` and more custom methods
in the spill implementation
3. if we introduce `MapAndSnapshotFactory`, we should also change the way to
create `StateTable` using `SnapshotStrategySynchronicityBehavior` currently,
but I think this is not in the scope of this refactor
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
