[Java] replaces synchronized Stack with Deque and refactors some othe…

[ingvarc]

…r methods

[ingvarc]

It would be better to use Java Microbenchmark Harness (JMH) instead. It is used for building, running, and analyzing nano/micro/milli/macro benchmarks and produces reliable results.

[NeilFraser]

Can you explain why a double-ended queue is superior to a stack? The only operations used are at one end of the data structure, there are no operations at the other end.

[ingvarc]

@NeilFraser
First of all, we definitely don't want synchronized behavior here as we use it locally to a method.
Moreover, it's sensible in terms of inheritance and the javadoc of Stack says: "A more complete and consistent set of LIFO stack operations is provided by the Deque interface and its implementations, which should be used in preference to this class". Additionally, the javadoc of ArrayDeque states: "This class is likely to be faster than Stack when used as a stack, ...".
Apart from Stack class itself there is no class in JDK that could represent the stack data structure and the implementations of the Deque interface therefore is considered the best option for stacks.

[NeilFraser]

Understood, makes sense. A side effect of this change is that we'd no longer support Java 5, but I guess that's probably ok these days. I'll revert if we get any complaints.

I'll follow up with another change that mirrors the "lastequality" -> "lastEquality" rename on all the other languages. Keeping cross-language compatibility makes maintaining DMP possible.