Hi,
Let's assume the following case. - a stream processor that uses the Processor API - context.schedule(1000) is called in the init() - the processor reads only one topic that has one partition - using custom timestamp extractor, but that timestamp is just a wall clock time Image the following events: 1., for 10 seconds I send in 5 messages / second 2., does not send any messages for 3 seconds 3., starts the 5 messages / second again I see that punctuate() is not called during the 3 seconds when I do not send any messages. This is ok according to the documentation, because there is not any new messages to trigger the punctuate() call. When the first few messages arrives after a restart the sending (point 3. above) I see the following sequence of method calls: 1., process() on the 1st message 2., punctuate() is called 3 times 3., process() on the 2nd message 4., process() on each following message What I would expect instead is that punctuate() is called first and then process() is called on the messages, because the first message's timestamp is already 3 seconds older then the last punctuate() was called, so the first message belongs after the 3 punctuate() calls. Please let me know if this is a bug or intentional, in this case what is the reason for processing one message before punctuate() is called? Thanks, Peter Péter Sinóros-Szabó Software Engineer Ustream, an IBM Company Andrassy ut 39, H-1061 Budapest Mobile: +36203693050 Email: peter.sinoros-sz...@hu.ibm.com
