Building upon the work done by the chap that I forked this from, I have edited the mlabwrap project. I liked the look of the Python 3 support - though this is currently untested - which is why I chose to fork from this project. Unfortunately, my Anaconda environment does not seem to like building using Microsoft Visual Studio.

One of the major reasons for using this fork was also because it claimed to allow building on Windows. However, I haven't gotten it to work. I've maintained the logic that was apparent in the fork so that, hopefully, it should be possible to make small changes to setup.py to allow installation on Windows.

This site has an ongoing discussion about installing mlabwrap on Windows, though I've yet to get it to work. I'll happily discuss and fix this fork as appropriate if anyone comes up with any ideas.

NOTE: Matlab root refers to the main Matlab folder, e.g: /usr/local/MATLAB/R2013a/

To install using pip:

pip install git+https://github.com/patricksnape/mlabwrap.git

To manually set the Matlab root directory, use the following command:

pip install git+https://github.com/patricksnape/mlabwrap.git --install-option="--matlab-root=PATH_TO_MATLAB_ROOT"

To install using setup.py (with optional parameter --matlab-root):

python setup.py install --matlab-root=PATH_TO_MATLAB_ROOT

IMPORTANT: THE GLOBAL mlab OBJECT DOES NOT EXIST FROM VERSION 1.2 ONWARDS

To import mlabwrap we do the following (note the explict call to use of the instance management class MatlabInstance):

from mlabwrap import MatlabInstance
mlab = MatlabInstance.get_instance()
import numpy as np

Then we can call a function as follows:

mlab.svd(np.array([[1,2], [1,3]]))

If we want to pass in an explicit version of Matlab, we simply specify the root directory:

from mlabwrap import MatlabInstance
mlab = MatlabInstance.get_instance(matlab_root='/usr/local/MATLAB/R2013a')

• Fix setup.py to work properly when installing without source
• Move mlabwrap into it's own package
• Add concept of MatlabInstance management class to allow managing a single Matlab instance. This breaks previous behaiour of just importing mlab, now we need to explcitly get an instance.

• Fix setup.py so that it does correct comparisons for Matlab version numbers

• Change the interface to mlabwrap such that you must explicitly call the init function. This allows you to pass in the matlab version you want more easily. This is achieved by way of the matlab_root keyword arg that can be passed in to the init function. This function is simply the old constructor.

• Rewrite setup.py so that it tries to guess where your Matlab folder is
• Make setup.py work with pip
• Add flag --matlab-root to setup.py
• Fix casting bug in mlabwrap.cpp
• Remove the old code that tried to use the help command to guess the version as this doesn't work very well

• using system encoding scheme to decode string in mlabraw.eval
• fix memory leak in mlabraw.put caused by chained char2mx and PyUnicode_AsUTF8String