[
https://issues.apache.org/jira/browse/QUARKS-80?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15212010#comment-15212010
]
Daniel John Debrunner commented on QUARKS-80:
---------------------------------------------
Not sure that batch() implies an aggregation, most likely it does, but it could
perform an operation that enriches each tuple more efficiently than doing it a
tuple at a time.
The natural use case for `aggregate()` is to aggregate a window in a streaming
paradigm.
The natural use case for `batch()` is to process tuples as a batch.
Don't get me started on `flatMap`. :-)
> The TWindow.aggregate name could be imporoved
> ---------------------------------------------
>
> Key: QUARKS-80
> URL: https://issues.apache.org/jira/browse/QUARKS-80
> Project: Quarks
> Issue Type: Improvement
> Reporter: Will Marshall
> Assignee: Will Marshall
> Priority: Minor
>
> For example, TWindow.batch clearly provides batch functionality for tuples in
> the partition -- tuples are processed in bulk, and then evicted.
> The contract for TWindow.aggregate is less clear given the name. The way it
> currently works, the partition is aggregated any time its contents change (a
> tuple is inserted or evicted).
> Perhaps something like TWindow.aggregateOnChange would be better?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
