Thanks Bill for the detailed description of your use case.
I think part of the issue is that today KStream does not allow non-keyed
aggregation, and hence users need to call a `map` function if they want to
do aggregations on other fields. We did this by design for KStream since if
users write the
Hi Guozhang,
Possibly, but the use case I'm working with is having a of collector
object , for aggregate statistics for example, that would output results
intermittently (via punctuate).
The issue for me is that 'transform(..)' returns a key-value pair for each
message, possibly of a different
Hello Bill,
We added transform() together with process() to support any user-customized
stateful processor that can still concatenate to another KStream.
So for your case, would `transform(...).to(topic)` provide the same
functionality as "processTo(topic, ...)"?
Guozhang
On Sat, Mar 12, 2016
Hi All,
While working with KStream/KStreamImp I discovered that there does not seem
to be any way to connect the results of the KStream.process method with a
sink node.
I'd like to propose an addition to the API a "processTo" method.
I've looked at and used the "transform", "reduceByKey" and "ag
