setParallelism changes test outcome about flink-spector

[lofifnc]

I have been successful in recreating your problem and found a fix for it. #93 I'm going to release it and wait for your confirmation.

[lofifnc]

Any parallelism greater than one will remove determinism from the test run. Because the events are now processed in parallel by multiple instances of the operators. For example the order of the events is not predefined any more. If I interpret your test correctly no expectation is checking the order of events. It could be the case, that the logic you're testing depends on order, or that a certain instance of a function sees certain events.
Without more information it's hard to say what's the case. I hope this helps.

[haf]

@lofifnc Thank you for your reply. I have nothing that requires a deterministic ordering, but I have a keyBy operator in there as well as a BoundedOutOfOrderTimestampExtractor. In fact, all of the code fits in this screengrab:

When setParallelism(2) is run, the matcher gets zero output events, but like you can see I have a println in a map, which gets run with parallelism enabled; and it outputs the correct value... I'm at a bit of a loss as to why the asserter would get zero values...

[lofifnc]

Version 0.9.3 has been released. Can you confirm this resolves your issue?

[haf]

Thank you very much, I'm going to try it tomorrow morning first thing!

[haf]

@lofifnc I didn't make it "first thing" but I've tried it now, and it works perfectly! Thank you @lofifnc