Kurt Young created FLINK-11959:
----------------------------------
Summary: Introduce window operator for blink streaming runtime
Key: FLINK-11959
URL:
https://issues.apache.org/jira/browse/FLINK-11959 Project: Flink
Issue Type: New Feature
Components: Runtime / Operators
Reporter: Kurt Young
Assignee: Kurt Young
We introduced a new window operator in blink streaming runtime, the differences between blink's window operator and the one used in DataStream API are:
# The blink's window operator is mainly used by window aggregate. It work closely with SQL's aggregate function, hence we didn't provide the flexibility to apply arbitrary `WindowFunction` like DataStream did. Instead, we only need to save the intermediate accumulate state for aggregate functions. There is no need for us to save original input rows into state, which will be much more efficient.
# This new window operator can deal with retract messages.
# We did some pane based optimization within sliding window operator, similar with [FLINK-7001|
https://issues.apache.org/jira/browse/FLINK-7001].
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)