Meg: a MATLAB engine connector in Python

Meg is a Python module to interface with the MATLAB engine, allowing to transfer data between Python and MATLAB, and to call MATLAB code from Python.

Assuming you can run matlab from the command line (i.e. the main MATLAB executable is in your PATH), using Meg is as simple as:

import meg
import numpy

with meg.Engine() as engine:
    data = numpy.empty((4,3))
    
    # Copy Python data to the engine
    engine["data"] = data
    
    # Execute MATLAB instructions
    engine("count = numel(data)")
    
    # Get data from MATLAB
    print(engine["count"])

Meg can be installed as any Python package: get the latest stable version from PyPi using pip (e.g. python3 -m pip install meg) or clone the source code and add it to your Python path.

Connecting to MATLAB

The main workhorse of Meg is the Engine object: it must be started before sending data between Python and MATLAB. An engine can be started (and automatically stopped) using the following syntax:

import meg

with meg.Engine() as engine:
    # Interact with MATLAB
    pass
# Once we reach this point, the engine has been automatically stopped.

Note that once an engine is stopped, all data stored inside which has not been transfered to Python is lost.

It is also possible to manually start and stop the engine:

import meg

# Create the engine, but do not start it
engine = meg.Engine()

engine.open()
# Interact with MATLAB
engine.close()

For more complicated environment (e.g. multiple versions of MATLAB, executable not in PATH, etc.), it is possible to specify the root directory of your target installation of MATLAB, and the explicit command to run when starting the engine:

import meg

# Target a specific installation of MATLAB
meg.setup("/opt/MATLAB/R2012b")

# Start the engine with a non-default command
with meg.Engine("/opt/MATLAB/R2012b/bin/matlab -nosplash") as engine:
    pass

Getting data to and from MATLAB

Data can be exchanged between Python and MATLAB using the Engine object: to store the content of the Python object name foo in the MATLAB object called bar, simply write engine["bar"] = foo. The reverse operation (storing the content of the MATLAB object called bar to a Python object called foo), write foo = engine["bar"].

Numbers (and number containers) are translated as-is between MATLAB and Python. From Python, list, tuples and numpy arrays will be converted to MATLAB numeric arrays; from MATLAB, numeric arrays will be converted either to Python scalars or to numpy arrays.

From Python, heterogeneous containers (lists, tuples or numpy arrays containing object of different types) will be converted to MATLAB cell arrays; from MATLAB, cell arrays will be converted to numpy arrays.

From Python, dictionaries and structured numpy arrays are converted to MATLAB struct arrays, and the same applies to the reverse conversion.

Calling MATLAB code

MATLAB statements are run by calling the engine object: assuming you have stored an object called x in MATLAB, computing the number of elements in it is done through engine("count = numel(x)"). Note that objects are not automatically exchanged between Python and MATLAB: they must be explicitely stored in the MATLAB engine before using them in MATLAB code.