[Neuroimaging] Technical details managing Python versions and packages.

[Matthew Brett]

Yo,

On Mon, Aug 3, 2015 at 3:17 PM, Yaroslav Halchenko <lists at onerussian.com> wrote:
>
> On Mon, 03 Aug 2015, Matthew Brett wrote:
>
>> > If you're interested to do that, let me know, I'll write up a summary
>> > of what to do.  I guess you are on Debian / Ubuntu?
>
>> Despite deafening silence, I did write up the summary:
>
>> https://matthew-brett.github.io/pydagogue/installing_on_debian.html
>
> my 1c with the NeuroDebian/Debian developer hat on.
>
> If you are into neuroimaging, and just a researcher/user who wants to
> have work done, I would recommend to stick to using
> (Neuro)Debian-provided packages as long as they are provided from our
> repositories:  that is the point of having a distribution like Debian,
> that we are taking care about compatibility between versions of
> various packages you use and which inter-depend among  themselves.
> And then your python packages live coherently with the rest of the
> system,  so you can manage python and non-python packages with the same
> commands.  If you need more recent versions of neuroimaging (and
> related) python packages -- that is where NeuroDebian repository comes
> in.  If interested to try -- there is a virtualbox VM, and docker images
> available.
>
> And only then, when I hit the situation that some package is not yet
> available from (Neuro)Debian, I usually create a virtualenv which
> complements system-wide collection (virtualenv --system-site-packages)
> and use its pip to install additional packages ideally by specifying
> them within requirements.txt (pip install -r requirements.txt) so I
> could later reproduce the same virtualenv.

As a matter of interest - how do you set up fresh environments to test in?

One of the big advantages of virtualenv for this is you can test a
fresh install with no packages other than the stock pip / setuptools.

You can also easily junk a set of installs with:

rmvirtualenv my-env

How do you do that?

See you,

Matthew