[SciPy-dev] Deprecating packages

Robert Kern robert.kern@gmail....
Fri Jul 27 03:49:54 CDT 2007


David Cournapeau wrote:
> Hi,
> 
>     In the context of releasing soon a new scipy tarball, I was 
> wondering what is the most appropriate approach to deprecate toolboxes 
> (in sandbox, not standard packages of course). 4 packages in 
> scipy.sandbox are now in scikits.learn (pyem, svm, ann and ga). What I 
> did for pyem is to throw an import error in the __init__ with 
> instructions on where to get the new code. Can I do the same for other 
> packages (those were not "mine", so I didn't want to change anything 
> before feedback) ?

Sandbox packages have no guarantees of stability, buildability, or continued
existence. You can just move them without bothering with any kind of deprecation
step. That said, it would be a good idea to talk to the authors of the packages
that aren't yours.

As I haven't heard from Fred Mailhot since the end of the SoC project for ann,
I'll stand in and say go for it. I'd kinda like to see ga go back to scipy.ga if
it's working again instead of moving into scikits.learn.machine (I'm sorry I
missed this aspect of the previous discussion). I don't think a general-purpose
function optimizer should be "hidden" in a machine learning package.

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless enigma
 that is made terrible by our own mad attempt to interpret it as though it had
 an underlying truth."
  -- Umberto Eco


More information about the Scipy-dev mailing list