Is the normalize-by-median.py script deterministic?

[zmunro]

Trying to recreate some previous;y computed results and there are a few places where nondeterminism could be present and I was wondering if it could be present in this script.

[standage]

I'll let @ctb correct me if I'm wrong. But if nothing else has changed in the environment, running normalize-by-median.py multiple times with the same arguments and the same inputs should produce identical output in single-threaded mode. There can be some non-deterministic behavior when counting k-mers in multi-threaded mode, so I guess there's a chance you might get slightly different results there. But I wouldn't expect the differences to be very much.

[ctb]

On Fri, Jun 12, 2020 at 12:41:31PM -0700, Daniel Standage wrote:

I'll let @ctb correct me if I'm wrong. But if nothing else has changed in the environment, running normalize-by-median.py multiple times with the same arguments and the same inputs should produce identical output in single-threaded mode. There can be some non-deterministic behavior when counting k-mers in multi-threaded mode, so I guess there's a chance you might get slightly different results there. But I wouldn't expect the differences to be very much.

Agreed!