Wrong result when using parallelStream().reduce()

Question

I want to multiply all elements of a list together and then multiply that result by 5 using streams in Java 8. This is my code:

List<Integer> numbers = Arrays.asList(1, 2, 3, 4);
int mul = numbers.stream().reduce(5, (acc, x) -> x * acc);
System.out.println(mul);

This gives me the correct result, 120. But if we change it to parallelStream() then it generates the wrong value. Why? Why does parallelization produce the wrong result in this case? And what is the fix?

See also http://stackoverflow.com/q/28724850/2711488 and http://stackoverflow.com/q/32866581/2711488 — Holger, Feb 20 '17 at 12:29

score 14 · Accepted Answer · answered Feb 19 '17 at 19:41

14

Because you're not respecting the preconditions of the reduce() method:

The identity value must be an identity for the accumulator function. This means that for all t, accumulator.apply(identity, t) is equal to t.

5 * t is not equal to t for all values of t.

You should use 1 as the identity, and multiply the result by 5.

answered Feb 19 '17 at 19:41

JB Nizet

678,734
91
1,224
1,255

But why it is producing correct result in case of iterative solution ? – Soumya Kanti Naskar Feb 19 '17 at 19:49
8

Because the identity is only used once in that case, but is used several times in a parallel stream – JB Nizet Feb 19 '17 at 19:50

Wrong result when using parallelStream().reduce()

1 Answers1