[ase-users] ase-gui no longer displays energy plots by default?

[Eric Hermes]

On 10/10/2014 3:23 AM, Jens Jørgen Mortensen wrote:
> Den 09-10-2014 kl. 16:26 skrev Eric Hermes:
>> Hello,
>>
>> I recently updated the global version of ASE on our cluster from 
>> 3.7.0 to 3.8.1, and I noticed that ase-gui (nee ag) no longer 
>> displays energy plots for trajectories by default. I spent a lot of 
>> time troubleshooting my matplotlib and pygtk installations before I 
>> realized that this was due to line 106 in ase/gui/graphs.py being 
>> commented out. Additionally, chuan (sorry, I don't know your name) 
>> mentioned in the gpaw IRC channel that invoking Tools -> NEB would 
>> cause not only the splined NEB plot to show up, but also any other 
>> plots that had been suppressed. This appears to happen because 
>> ase/gui/neb.py invokes pylab.show(), which is the command that is 
>> commented out on line 106 in ase/gui/graphs.py. Uncommenting this 
>> line caused the plots to start showing up again with no ill effects. 
>> Is this functionality disabled for a reason?
>
> That line 106 has been commented for 7 years now!  So it must be 
> something else that has changed from 3.7.0 to 3.8.1.  I think show() 
> should not be needed when ion() has been called (interactive mode). 
> What architecture are you on?  Do you have some personal matplotlib 
> settings?
Indeed, it would appear that ASE 3.7.0 also had that line commented out, 
so I misdiagnosed the issue. Regardless, uncommenting pylab.show() 
appears to have brought back the old behavior.

Our cluster runs Rocks Cluster Distribution 5.1, which I believe is 
based on CentOS 5.1. Our cluster is a traditional Intel x86_64 
architecture with 8 cores per node. I do not have any custom matplotlib 
settings. The version of Python in the repositories is 2.4, so I have 
installed Python 2.7.3 manually in /usr/local. This also required that I 
install several other Python packages manually:

MatPlotLib: 1.4.0
NumPy: 1.9.0
SciPy: 0.14.0
pygtk: 2.10.1

libpng is version 1.2.10 installed from the repositories. I'm not sure 
whether it matters, but NumPy and SciPy are dynamically linked against 
the Intel MKL from Composer XE 2013.5.192.

Eric
>
> Jens Jørgen
>
>>
>> Thanks,
>> Eric
>>
>

-- 
Eric Hermes
J.R. Schmidt Group
Chemistry Department
University of Wisconsin - Madison