On Thu, Jul 22, 2010 at 5:25 PM, Terry Reedy <tjre...@udel.edu> wrote:
> Looks good so far. I would like to see that repeated all the way down to > range(10) to make sure people doing millions of small sorts were not getting > screwed. > I only use the radix sort for n > 40. :-) > Have you run a patched version against test_sort.py? I believe it mostly > tests lists of small ints, so radix methods would mostly be switched in. > For testing purposes, I maintain a private fork of Python where I've replaced the built-in list with blist. I then run Python's test suite as a way of testing blist. So, yes, all of the tests have been run. :-) However, since I only use radix for n > 40, many of the tests have not actually tested the radix sort. > If it were added and the switching were internal, new test cases would be > needed to test test timsort. That's a good point. It's tempting to add an undocumented parameter to blist.sort that selects the sorting algorithm to use, to make it make it easier to test multiple algorithms. There are probably several different ways to achieve a similar effect. Do you mind if we table that discussion until I actually have a patch? -- Daniel Stutzbach, Ph.D. President, Stutzbach Enterprises, LLC <http://stutzbachenterprises.com>
-- http://mail.python.org/mailman/listinfo/python-list
