[Scipy-tickets] [SciPy] #999: Fast (O(n log(n)) ) implementation of Kendall's Tau

SciPy Trac scipy-tickets@scipy....
Sat Nov 13 21:35:19 CST 2010


#999: Fast (O(n log(n)) ) implementation of Kendall's Tau
-------------------------+--------------------------------------------------
 Reporter:  enzo         |       Owner:  somebody    
     Type:  enhancement  |      Status:  needs_review
 Priority:  normal       |   Milestone:  0.9.0       
Component:  scipy.stats  |     Version:  0.7.0       
 Keywords:  kendall tau  |  
-------------------------+--------------------------------------------------

Comment(by rgommers):

 I'm not so sure about the license being OK if we don't hear from Enzo. His
 file explicitly says "inspired by Java implementation", which is GPL. Even
 if the Python looks different now, if he studied the Java one first it
 should be GPL'd. I think. Of course, IANAL.

 On the speed, I did a quick comparison, and dsimcha's C version is at
 least two orders of magnitude faster (C: ~1sec for 1e6 elements, Python:
 6.5sec for 5e4 elements). Which is not surprising, looking at the Python
 code - that really could benefit from being Cythonized I think.

 I also don't have experience binding C to Python, but I guess it shouldn't
 be all that hard. Otherwise a clean Cython version may be a good option
 too. Even without considering the license issue, it would probably be good
 to go with the fastest version seeing how much work several people have
 already put into it.

-- 
Ticket URL: <http://projects.scipy.org/scipy/ticket/999#comment:9>
SciPy <http://www.scipy.org>
SciPy is open-source software for mathematics, science, and engineering.


More information about the Scipy-tickets mailing list