[Nipy-devel] nipy.interfaces in nipy
Wed Nov 25 21:07:59 CST 2009
we actually have it as 1.3 and later. i know chris did some tests
initially. i just looked at the code and most of it that uses numpy
looks fairly basic and would be expected to have been there in numpy
for a while. the one place where we might run into trouble is
np.unique. the function signature has changed over the years and it's
going to change again with 1.4.
do you or anybody on the list know of a code analysis tool that can
inform us of this without actually running tests in different
On Wed, Nov 25, 2009 at 9:55 PM, Yaroslav Halchenko
> BTW, is numpy >= 1.2 requirement is very strict? or it is just for
> running unittests? the problem is that lenny has only 1.1...
> On Wed, 25 Nov 2009, Satrajit Ghosh wrote:
>> regarding dependencies for nipype, the interfaces depend only only on
>> scipy and numpy and the pipeline part requires networkx and ipython
>> components. so users wishing to interface with the external packages
>> through pythonic scripts can install and run nipype with only numpy
>> installed. (it's fairly lightweight). the scipy dependency will move
>> out to the nipype algorithms module (only SpecifyModel uses it) by the
>> next release. the algorithms in there will probably move out of nipype
>> and into nipy.algorithms or some such component in the future.
>> my suggestion would be to run both in parallel for a little while and
>> see how usage interacts and then rethink the shifting of components
>> depending on how dependencies/interaction play out.
>> dav: [a historical note] the nipype matlab and spm interfaces started
>> from matthew's implementation and the trunk interfaces branch and
>> evolved into what they are now.
> =------------------------------ /v\ ----------------------------=
> Keep in touch // \\ (yoh@|www.)onerussian.com
> Yaroslav Halchenko /( )\ ICQ#: 60653192
> Linux User ^^-^^ 
> Nipy-devel mailing list
More information about the Nipy-devel