Support taccumulate

[anicusan]

Most of the higher-order iterator functions in Base (sum, prod, any, all, count, minimum, maximum, etc.) can be implemented on top of tmapreduce. The only other fundamental building block that would be useful is taccumulate for cumsum and the like - are there any plans for this?

[MasonProtter]

Parallelizing accumulate is typically very difficult, since the later results require the earlier results. It's possible with lots of re-associations and some clever two-pass algorithms.

If you'd like to take a shot at implementing it, I'd try and take a look and give advice.