Running Matlab as a Python subprocess
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.
01.import os02. 03.from subprocess import Popen04.from scipy.io import loadmat05.from scipy.io import savemat06. 07.import numpy as np08. 09.infile = os.path.join(os.path.dirname(__file__), 'infile.mat')10.outfile = os.path.join(os.path.dirname(__file__), 'outfile.mat')11. 12.with open(infile, 'wb+') as fp: pass13.with open(outfile, 'wb+') as fp: pass14. 15.A = np.array([16.[1, 0, 0, 0],17.[0, 1, 0, 0],18.[0, 0, 1, 0],19.[0, 0, 0, 1]], dtype=float)20. 21.indict = {'A': A}22. 23.savemat(infile, indict)24. 25.matlab = ['matlab']26.options = ['-nosplash', '-r']27.command = ["load('{0}');[R, jb]=rref(A);save('{1}');exit;".format(infile, outfile)]28. 29.# on mac use the full path to matlab or add the path to your .profile30.# matlab = ['/Applications/MATLAB_R2016a.app/bin/matlab']31. 32.# on windows use:33.# options = ['-nosplash', '-wait', '-r']34. 35.p = Popen(matlab + options + command)36. 37.stdout, stderr = p.communicate()38. 39.outdict = loadmat(outfile)40.jb = outdict['jb'].tolist()[0]41. 42.print jb43. 44.# should be [1, 2, 3, 4]45.# don't forget Matlab is 1-based46.# so convert to [0, 1, 2, 3]