On Wed, Dec 21, 2016 at 1:23 PM, David Case <david.case.rutgers.edu> wrote:
> On Wed, Dec 21, 2016, Hai Nguyen wrote:
>
> > I actually updated configure2 to follow your option A (which I agree with
> > too).
>
> ...thx...that means we automatically advance to option D!
>
>
> A few questions:
>
> 1. do we know which programs require scipy? Can we try to get rid of that?
>
Pengfei Li's pymsmt package. He is using scipy for some minimization
protocols.
If he agrees, we can make scipy optional since user can install via
'amber.conda install scipy' in 5 seconds.
>
> 2. are we actually set up so that users don't need cython?
>
User don't need Cython (thanks to Jason suggestion to pre-compile pyx
files).
But developers still need to install Cython.
>
> 3. What does "optional" mean in your new configure2 script? I think we
> would
> be better off referring users to a web page/wiki for info about which
> python
> packages are needed for what, rather than to try to cram all that
> information
> into a configure2 usage statement that will quickly get out of date.
>
"Optional" means they are only useful for a small number of users (e.g: if
someone wants to use jupyter notebook for interactive computing or
visualization).
Hard requirement means "must have". numpy is hard requirement for both
ParmEd and pytraj.
> 4. Has anyone (e.g. Dan) tried a soft-link to have their
> $AMBERHOME/miniconda directory point to a common anaconda tree? This
> might be
> a simple way for developers to have many amber trees without having many
> miniconda installations. (Not sure this will work, but it seems worth
> trying...)
>
>
>
I myself say no to soft/hard/alias/... link. Just install Anaconda or
Miniconda in your $HOME folder and then use --with-python flag
bash configure --with-python $HOME/anaconda/bin/python gnu
Hai
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Wed Dec 21 2016 - 11:30:03 PST
