On Thu, Jan 28, 2016 at 11:04 AM, Hannes Loeffler <
Hannes.Loeffler.stfc.ac.uk> wrote:
> Look at this from the bright side: at least it's not docker or, worse,
> a virtual machine :-) I agree that those virtualisation attempts seem
> to go out of hands (quite) a bit but when you have to maintain Python
> code, dependent on 3rd party libraries especially, you quickly learn why
> you want to do this.
I just want to stress in the context of AMBER, we only require numpy for
Python.
> On MacOSX this is just a nightmare and packaging
> your own Python distribution (or go miniconda or similar) seems to be
> the only sensible way to go. I learned that the hard way... And don't
> get me started on UCS2 vs UCS4 being a compile-time option in 2.7. I
> had to tell 3(!) sysadmins that they really ought to look into how
> (Linux) distributors build their stuff (hint: Linux distributions do
> not use the default of UCS2).
>
>
Yes, with miniconda installation, maintaining package and install new one
is super easy and superfast (takes only about 20-30 seconds).
For different OS system (Linux, Win, OSX), you only need to remember a
single command `*conda install*` rather ask if you need to use `apt-get`,
`yum install`, or `brew install`, ...
The easy installation with a single interface also makes the user issue
debugging much easier.
Hai
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Thu Jan 28 2016 - 11:30:03 PST
