Measure time to complete individual streaming steps?

Question

Is there a library that allows measuring Java Stream steps?

As in "when consuming this stream, X amount of time is spent in that filter, Y amount of time is spent in this map, etc..."?

Or do I have to tinker around a bit?

Answer 1

By definition, there is no 'between' time for steps. The method chain that you define only defines the steps to take (the methods build an internal structure containing the steps).

Once you call the terminal operator, all steps are executed in indeterminate order. If you never call a terminal operator, no steps are executed either.

What I mean by indeterminate order is this:

intstream.peek(operation1).peek(operation2).sum();

You don't know if operation1 is called, then operation2 called, then operation1 again, or if operation1 is called a bunch of times, then operation2 a bunch of times, or something else yet. (I hope I'm correct in saying this. [edit] correct) It's probably up to the implementation of the VM to trade off storage with time complexity.

The best you can to is measure the time each operation takes as you have full control over every operation. The time the stream engine takes is hard to gauge. However, know that when you do this, you're breaking the stream API requirement that operations should be side-effect-free. So don't use it in production code.

Answer 2

The only tool I can think of is JMH, which is a general benchmarking tool for Java. However note that it might be hard to understand the effect that different Stream operators have on the benchmark due to JIT. The fact that your code contains a Stream with a bunch of different operators doesn't mean that's exactly the way JIT compiles it.