[
https://issues.apache.org/jira/browse/QUARKS-230?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Dale LaBossiere updated QUARKS-230:
-----------------------------------
Description:
A recent use case involved the desire for "timer triggered" instead of
"partition content change triggered" window aggregations. E.g., "I want to
trigger an aggregation every second over the last 10 tuples in a window
partition" -- a count-based window with timer triggered aggregations.
I propose adding 3 methods to TWindow. Two in direct support of "timer
triggered" aggregations (this processing model seems like it could be common
enough to warrant making it conveniently available) and one to make it easier
to use the lower level Window API to define and use other processing models.
I'm submitting a PR with the details for review but the net is these additions
to TWindow:
```
<U> TStream<U> timedAggregate(long period, TimeUnit unit,
BiFunction<List<T>, K, U> aggregator);
<U> TStream<U> timedBatch(long period, TimeUnit unit, BiFunction<List<T>,
K, U> batcher);
<U, L extends List<T>> TStream<U> process(Window<T,K,L> window,
BiFunction<List<T>, K, U> aggregator);
```
See https://github.com/apache/incubator-quarks/pull/167
was:
A recent use case involved the desire for "timer triggered" instead of
"partition content change triggered" window aggregations. E.g., "I want to
trigger an aggregation every second over the last 10 tuples in a window
partition" -- a count-based window with timer triggered aggregations.
I propose adding 3 methods to TWindow. Two in direct support of "timer
triggered" aggregations (this processing model seems like it could be common
enough to warrant making it conveniently available) and one to make it easier
to use the lower level Window API to define and use other processing models.
I'm submitting a PR with the details for review but the net is these additions
to TWindow:
```
<U> TStream<U> timedAggregate(long period, TimeUnit unit,
BiFunction<List<T>, K, U> aggregator);
<U> TStream<U> timedBatch(long period, TimeUnit unit, BiFunction<List<T>,
K, U> batcher);
<U, L extends List<T>> TStream<U> process(Window<T,K,L> window,
BiFunction<List<T>, K, U> aggregator);
```
> Add timer triggered window aggregations
> ---------------------------------------
>
> Key: QUARKS-230
> URL: https://issues.apache.org/jira/browse/QUARKS-230
> Project: Quarks
> Issue Type: New Feature
> Reporter: Dale LaBossiere
> Assignee: Dale LaBossiere
>
> A recent use case involved the desire for "timer triggered" instead of
> "partition content change triggered" window aggregations. E.g., "I want to
> trigger an aggregation every second over the last 10 tuples in a window
> partition" -- a count-based window with timer triggered aggregations.
> I propose adding 3 methods to TWindow. Two in direct support of "timer
> triggered" aggregations (this processing model seems like it could be common
> enough to warrant making it conveniently available) and one to make it easier
> to use the lower level Window API to define and use other processing models.
> I'm submitting a PR with the details for review but the net is these
> additions to TWindow:
> ```
> <U> TStream<U> timedAggregate(long period, TimeUnit unit,
> BiFunction<List<T>, K, U> aggregator);
> <U> TStream<U> timedBatch(long period, TimeUnit unit, BiFunction<List<T>,
> K, U> batcher);
> <U, L extends List<T>> TStream<U> process(Window<T,K,L> window,
> BiFunction<List<T>, K, U> aggregator);
> ```
> See https://github.com/apache/incubator-quarks/pull/167
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
