Git archives for data sharing

Posted October 26, 2013 at 06:49 PM | categories: data | tags:

1. A molecule
2. A bulk calculation
3. Analysis via the JSON files
4. Create the archive file
5. Summary

in some past posts we have looked at constructing JSON files for data sharing. While functional, that approach requires some extra work to create the data files for sharing, and may not be useful for all sorts of data. For instance you may not want to store electron density in a JSON file.

Here we consider using git archives for packaging exactly the data you used in the same hierarchy as on your file system. The idea is to store your work in a git repository. You commit any data files you would want to share, and then create an archive of that data. This enables you to control what gets shared, while keeping the data that should not be shared out of the archive.

We will run a few VASP calculations, and summarize each one in a JSON file. We will commit those JSON files to the git repository, and finally make a small archive that contains them.

1 A molecule

Calculate the total energy of a CO molecule.

from ase import Atoms, Atom
from jasp import *
import numpy as np
import json
np.set_printoptions(precision=3, suppress=True)

co = Atoms([Atom('C',[0,   0, 0]),
            Atom('O',[1.2, 0, 0])],
            cell=(6., 6., 6.))

with jasp('molecules/simple-co', #output dir
          xc='PBE',  # the exchange-correlation functional
          nbands=6,  # number of bands
          encut=350, # planewave cutoff
          ismear=1,  # Methfessel-Paxton smearing
          sigma=0.01,# very small smearing factor for a molecule
          atoms=co) as calc:
    print 'energy = {0} eV'.format(co.get_potential_energy())
    print co.get_forces()
    with open('JSON', 'wb') as f:
        f.write(json.dumps(calc.dict))
    os.system('git add JSON')

energy = -14.687906 eV
[[ 5.095  0.     0.   ]
 [-5.095  0.     0.   ]]

2 A bulk calculation

Now we run a bulk calculation

from jasp import *

from ase import Atom, Atoms

atoms = Atoms([Atom('Cu',  [0.000,      0.000,      0.000])],
              cell=  [[ 1.818,  0.000,  1.818],
                      [ 1.818,  1.818,  0.000],
                      [ 0.000,  1.818,  1.818]])

with jasp('bulk/alloy/cu',
          xc='PBE',
          encut=350,
          kpts=(13,13,13),
          nbands=9,
          ibrion=2,
          isif=4,
          nsw=10,
          atoms=atoms) as calc:
    print atoms.get_potential_energy()
    with open('JSON', 'wb') as f:
        f.write(json.dumps(calc.dict))
    os.system('git add JSON')

-3.723306

3 Analysis via the JSON files

This analysis is independent of jasp and therefore is portable

We can retrieve the bulk data

import json

with open('bulk/alloy/cu/JSON', 'rb') as f:
    d = json.loads(f.read())
    print d['data']['total_energy']

-3.723306

Or the molecule data:

import json

with open('molecules/simple-co/JSON', 'rb') as f:
    d = json.loads(f.read())
    print d['data']['total_energy']

-14.687906

4 Create the archive file

As you do your work, you add and commit files as needed. For this project all that needs to be shared are the JSON files, and the scripts (which are in this document) that we used to run the calculations and do the analysis. If we are satisfied with the state of the git repository, we create an archive like this:

git archive --format zip HEAD -o archive.zip

Here is the result: archive.zip .

You can download the zip file, unzip it, and rerun the analysis to extract the total energies on any system with a modern Python installation.

5 Summary

This seems to be an easy way to share data from a single project, i.e. a single git repository. It isn't obvious how you would package data from multiple projects, or how you would run multiple projects in a single directory.

org-mode source

The Kitchin Research Group

Chemical Engineering at Carnegie Mellon University