从 MATLAB 调用 Python 函数

我需要从 MATLAB 调用一个 Python 函数。我该怎么做呢?

136602 次浏览

You could embed your Python script in a C program and then MEX the C program with MATLAB but that might be a lot of work compared dumping the results to a file.

You can call MATLAB functions in Python using PyMat. Apart from that, SciPy has several MATLAB duplicate functions.

But if you need to run Python scripts from MATLAB, you can try running system commands to run the script and store the results in a file and read it later in MATLAB.

I had a similar requirement on my system and this was my solution:

In MATLAB there is a function called perl.m, which allows you to call perl scripts from MATLAB. Depending on which version you are using it will be located somewhere like

C:\Program Files\MATLAB\R2008a\toolbox\matlab\general\perl.m

Create a copy called python.m, a quick search and replace of perl with python, double check the command path it sets up to point to your installation of python. You should now be able to run python scripts from MATLAB.

Example

A simple squared function in python saved as "sqd.py", naturally if I was doing this properly I'd have a few checks in testing input arguments, valid numbers etc.

import sys


def squared(x):
y = x * x
return y


if __name__ == '__main__':
x = float(sys.argv[1])
sys.stdout.write(str(squared(x)))

Then in MATLAB

>> r=python('sqd.py','3.5')
r =
12.25
>> r=python('sqd.py','5')
r =
25.0
>>

Since Python is a better glue language, it may be easier to call the MATLAB part of your program from Python instead of vice-versa.

Check out Mlabwrap.

Since MATLAB seamlessly integrates with Java, you can use Jython to write your script and call that from MATLAB (you may have to add a thin pure JKava wrapper to actually call the Jython code). I never tried it, but I can't see why it won't work.

I've adapted the perl.m to python.m and attached this for reference for others, but I can't seem to get any output from the Python scripts to be returned to the MATLAB variable :(

Here is my M-file; note I point directly to the Python folder, C:\python27_64, in my code, and this would change on your system.

function [result status] = python(varargin)
cmdString = '';
for i = 1:nargin
thisArg = varargin{i};
if isempty(thisArg) || ~ischar(thisArg)
error('MATLAB:python:InputsMustBeStrings', 'All input arguments must be valid strings.');
end
if i==1
if exist(thisArg, 'file')==2
if isempty(dir(thisArg))
thisArg = which(thisArg);
end
else
error('MATLAB:python:FileNotFound', 'Unable to find Python file: %s', thisArg);
end
end
if any(thisArg == ' ')
thisArg = ['"', thisArg, '"'];
end
cmdString = [cmdString, ' ', thisArg];
end
errTxtNoPython = 'Unable to find Python executable.';
if isempty(cmdString)
error('MATLAB:python:NoPythonCommand', 'No python command specified');
elseif ispc
pythonCmd = 'C:\python27_64';
cmdString = ['python' cmdString];
pythonCmd = ['set PATH=',pythonCmd, ';%PATH%&' cmdString];
[status, result] = dos(pythonCmd)
else
[status ignore] = unix('which python'); %#ok
if (status == 0)
cmdString = ['python', cmdString];
[status, result] = unix(cmdString);
else
error('MATLAB:python:NoExecutable', errTxtNoPython);
end
end
if nargout < 2 && status~=0
error('MATLAB:python:ExecutionError', ...
'System error: %sCommand executed: %s', result, cmdString);
end

EDIT :

Worked out my problem the original perl.m points to a Perl installation in the MATLAB folder by updating PATH then calling Perl. The function above points to my Python install. When I called my function.py file, it was in a different directory and called other files in that directory. These where not reflected in the PATH, and I had to easy_install my Python files into my Python distribution.

The simplest way to do this is to use MATLAB's system function.

So basically, you would execute a Python function on MATLAB as you would do on the command prompt (Windows), or shell (Linux):

system('python pythonfile.py')

The above is for simply running a Python file. If you wanted to run a Python function (and give it some arguments), then you would need something like:

system('python pythonfile.py argument')

For a concrete example, take the Python code in Adrian's answer to this question, and save it to a Python file, that is test.py. Then place this file in your MATLAB directory and run the following command on MATLAB:

system('python test.py 2')

And you will get as your output 4 or 2^2.

Note: MATLAB looks in the current MATLAB directory for whatever Python file you specify with the system command.

This is probably the simplest way to solve your problem, as you simply use an existing function in MATLAB to do your bidding.

This seems to be a suitable method to "tunnel" functions from Python to MATLAB:

http://code.google.com/p/python-matlab-wormholes/

The big advantage is that you can handle ndarrays with it, which is not possible by the standard output of programs, as suggested before. (Please correct me, if you think this is wrong - it would save me a lot of trouble :-) )

Try this MEX file for ACTUALLY calling Python from MATLAB not the other way around as others suggest. It provides fairly decent integration : http://algoholic.eu/matpy/

You can do something like this easily:

[X,Y]=meshgrid(-10:0.1:10,-10:0.1:10);
Z=sin(X)+cos(Y);
py_export('X','Y','Z')
stmt = sprintf(['import matplotlib\n' ...
'matplotlib.use(''Qt4Agg'')\n' ...
'import matplotlib.pyplot as plt\n' ...
'from mpl_toolkits.mplot3d import axes3d\n' ...
'f=plt.figure()\n' ...
'ax=f.gca(projection=''3d'')\n' ...
'cset=ax.plot_surface(X,Y,Z)\n' ...
'ax.clabel(cset,fontsize=9,inline=1)\n' ...
'plt.show()']);
py('eval', stmt);

As @dgorissen said, Jython is the easiest solution.

Just install Jython from the homepage.

Then:

javaaddpath('/path-to-your-jython-installation/jython.jar')


import org.python.util.PythonInterpreter;


python = PythonInterpreter; %# takes a long time to load!
python.exec('import some_module');
python.exec('result = some_module.run_something()');
result = python.get('result');

See the documentation for some examples.

Beware: I never actually worked with Jython and it seems that the standard library one may know from CPython is not fully implemented in Jython!

Small examples I tested worked just fine, but you may find that you have to prepend your Python code directory to sys.path.

A little known (and little documented) fact about MATLAB's system() function: On unixoid systems it uses whatever interpreter is given in the environment variable SHELL or MATLAB_SHELL at the time of starting MATLAB. So if you start MATLAB with

SHELL='/usr/bin/python' matlab

any subsequent system() calls will use Python instead of your default shell as an interpreter.

With Matlab 2014b python libraries can be called directly from matlab. A prefix py. is added to all packet names:

>> wrapped = py.textwrap.wrap("example")


wrapped =


Python list with no properties.


['example']

Starting from Matlab 2014b Python functions can be called directly. Use prefix py, then module name, and finally function name like so:

result = py.module_name.function_name(parameter1);

Make sure to add the script to the Python search path when calling from Matlab if you are in a different working directory than that of the Python script.

See more details here.

Like Daniel said you can run python commands directly from Matlab using the py. command. To run any of the libraries you just have to make sure Malab is running the python environment where you installed the libraries:

On a Mac:

  • Open a new terminal window;

  • type: which python (to find out where the default version of python is installed);

  • Restart Matlab;

  • type: pyversion('/anaconda2/bin/python'), in the command line (obviously replace with your path).
  • You can now run all the libraries in your default python installation.

For example:

py.sys.version;

py.sklearn.cluster.dbscan