BRG blog

Running Matlab as a Python subprocess

Posted on August 3rd, 2016, by tom in Code

Although NumPy and Scipy usually provide all the tools you need (and more) for scientific computing, sometimes you may want to use a bit of Matlab in the background. In my case, for example, to use rref. Conveniently, Matlab can be run as a Python subprocess, provided of course that Matlab is installed on your system.

import os

from subprocess import Popen
from scipy.io import loadmat
from scipy.io import savemat

import numpy as np

infile = os.path.join(os.path.dirname(__file__), 'infile.mat')
outfile = os.path.join(os.path.dirname(__file__), 'outfile.mat')

with open(infile, 'wb+') as fp: pass
with open(outfile, 'wb+') as fp: pass

A = np.array([
[1, 0, 0, 0],
[0, 1, 0, 0],
[0, 0, 1, 0],
[0, 0, 0, 1]], dtype=float)

indict = {'A': A}

savemat(infile, indict)

matlab  = ['matlab']
options = ['-nosplash', '-r']
command = ["load('{0}');[R, jb]=rref(A);save('{1}');exit;".format(infile, outfile)]

# on mac use the full path to matlab or add the path to your .profile
# matlab = ['/Applications/MATLAB_R2016a.app/bin/matlab']

# on windows use:
# options = ['-nosplash', '-wait', '-r']

p = Popen(matlab + options + command)

stdout, stderr = p.communicate()

outdict = loadmat(outfile)
jb = outdict['jb'].tolist()[0]

print jb

# should be [1, 2, 3, 4]
# don't forget Matlab is 1-based
# so convert to [0, 1, 2, 3]