Hi Philipp, am I correct to assume that your tuples do not arrive in the order of the timestamp that you extract. Unfortunately, for that case the current windowing implementation does not work correctly. We are working hard on fixing this for the upcoming 0.10 release, though. If you are interested in what we are doing there you can check out the design documents in the Flink wiki: https://cwiki.apache.org/confluence/display/FLINK/Streams+and+Operations+on+Streams https://cwiki.apache.org/confluence/display/FLINK/Time+and+Order+in+Streams
Cheers, Aljoscha On Fri, 18 Sep 2015 at 13:19 Philipp Goetze <philipp.goe...@tu-ilmenau.de> wrote: > Hello again, > > another question from me =). Could you provide an example on how to > correctly use windows based on timestamps on the tuples (i.e. non-realtime)? > > As a simple example I tried something like this: > > val w4 = rdfSource.window(Time.of(200000L, customTimestamp, > 1219625100000L)).mapWindow(customMap _).flatten().print() > > In the customMap I just collect all tuples in one list for each window. > However, this does not work correctly as sometimes all the records are > collected in one list, although there should be two windows for the example > data. > > Best Regards, > Philipp >
