On Thu, Jul 22, 2010 at 6:52 PM, Terry Reedy <tjre...@udel.edu> wrote:
> Another sort of issue will be code maintainability. Two algorithms is
> potentially more problems than one. To me, that partly depends on how well
> factored the current code is.
>
Two algorithms is more code than one algorithm. No way around that. :-)
Radix sort is a relatively simple sorting algorithm, so the extra complexity
is not too bad. Once I have a patch ready, others can decide if the
additional complexity is worth the performance boost.
> It would be best is rsort were a switch replacement for timsort after all
> preparations (such as decorate) were done.
>
Yes, that's pretty much what I have in mind. In pseudo-code:
sort:
decorate()
if all_correct_type and n > 40:
radix_sort()
else
timsort()
undecorate()
> And that is about all I can think of.
Thank you for the thoughts. I appreciate them!
--
Daniel Stutzbach, Ph.D.
President, Stutzbach Enterprises, LLC <http://stutzbachenterprises.com>
--
http://mail.python.org/mailman/listinfo/python-list
