[SciPy-Dev] Recent changes to scipy stats

Travis Oliphant oliphant@enthought....
Tue Jun 1 01:38:35 CDT 2010


My recent changes to trunk certainly started a controversy.   I'm not exactly sure why.    I do not mean to give the impression that people should "clean" up after me which has been implied by some.   Please let me know if there is something specific that you would like me to do. 

I appreciate the speific concerns that Ralf raised as opposed to "generalizations" and metaphors that are open to interpretation.   All of his concerns have been addressed, I think, except the addition of all tests that some would like to see. 

Some of the added methods are so simple, that I do not think they require tests to verify their accuracy --- you can look at the code and understand it.    In cases like this I get somewhat frustrated with a naive fixed rule like "no check-ins" without "tests".  

There can always be more tests, but tests cost and should be part of a general improvement strategy and not just trotted out as a weapon when there is disagreement about something else.   

Is there a disagreement about other changes that have been made?    The only one I can think of that could be controversial is perhaps pulling in Josef's expect methods from his file when he did not want the "API" methods finalized.   I'm fine with removing them if he wants to do that. 

Perhaps, the interface I chose to fix certain parameters for the fit methods is also in question.    I really don't know as I have received no specific communication about the concerns.   I welcome any review or comment on what has been done. 

As I am not able to follow all threads on SciPy-User and SciPy-Dev, I did not know that Ralf was going to create the 0.8.x branch when he did.  Perhaps I should have known, but I did not know.

-Travis



More information about the SciPy-Dev mailing list