[SciPy-user] Running SciPy SVN without installing Was: Matlab IO -- can others with matlab test this bug?

Emanuele Olivetti emanuele@relativita....
Tue Jul 29 09:04:03 CDT 2008

Insted of stow why not using just PYTHONPATH?
(in bash, continuing the Zachary's example on numpy1)
export PYTHONPATH="SOMEPATH/stow/numpy1"
then run python/emacs/ipython/whatever and use numpy1.


Jonathan Hunt wrote:
> Hi,
> Thanks for the advice. But all of these seem like hacks. Is there any
> clean way to use SVN SciPy without installing it. This is particularly
> important since I'm hoping to modify SciPy (I want to add a mio
> wrapper for MATLAB 7/HDF5 files) and obviously I would like to avoid
> having the build/install everytime I make a modification to the
> library that I want to test.
> Thanks,
> Jonny
> On Tue, Jul 29, 2008 at 10:37 PM, David Cournapeau
> <david@ar.media.kyoto-u.ac.jp> wrote:
>> Zachary Pincus wrote:
>>> Perhaps some others can chime in with more "official" methods, but I
>>> typically do one of several things to test SVN versions:
>> On Unix, a nice and easy way to easily switch between different numpy
>> installs without touching at all the environment is stow. Basically what
>> you do once stow is installed is:
>> python setup.py install --prefix=SOMEPATH/stow/numpy1
>> python setup.py install --prefix=SOMEPATH/stow/numpy2
>> And then stow makes link between the 'real' numpy and the one you want:
>> stow numpy1 -> numpy1 is used
>> stow -R numpy1 -> numpy1 is disabled
>> stow numpy2 -> numpy2 is enabled
>> I switch between many numpy configurations (one which only uses released
>> versions for 'real' work, many other with various compilers combination
>> to test the build system). It also enables true uninstallation, since
>> any installed version is self-contained in one directory (you can rm -rf
>> the directory without any chance of removing something unrelated).
>> cheers,
>> David

More information about the SciPy-user mailing list