https://github.com/pv/pythoncall
Python embedded inside Matlab
https://github.com/pv/pythoncall
Last synced: 3 months ago
JSON representation
Python embedded inside Matlab
- Host: GitHub
- URL: https://github.com/pv/pythoncall
- Owner: pv
- License: bsd-3-clause
- Created: 2009-11-13T21:14:20.000Z (over 15 years ago)
- Default Branch: master
- Last Pushed: 2009-11-13T21:15:35.000Z (over 15 years ago)
- Last Synced: 2024-12-31T13:44:05.126Z (5 months ago)
- Language: C
- Homepage: http://www.iki.fi/pav/software/pythoncall/
- Size: 91.8 KB
- Stars: 15
- Watchers: 4
- Forks: 9
- Open Issues: 0
-
Metadata Files:
- Readme: README.txt
- License: LICENSE.txt
Awesome Lists containing this project
README
==========
Pythoncall
==========
:author: Pauli Virtanen
:version: 0.3.1
Python embedded inside Matlab.
Conversion of basic types + numeric array types between Python and Matlab
is provided.
Usage
-----
- Evaluate python string::
pythoncall('eval', command)
``command`` is a string containing the Python program to run.
It is run in the __main__ module.
- Get variable values from the Python space::
[value1, value2, ...] = pythoncall('get', varname1, varname2, ...)
This gets variables from the top-level of __main__ module.
- Set variable values to the Python space::
pythoncall('set', varname1, value1, varname2, value2, ...)
This sets variables to the top-level of __main__ module.
Compilation (important!)
------------------------
.. warning::
This is apparently outdated! Pythoncall may require some modifications
before it will work with recent Matlab versions.
Note that Matlab apparently ``dlopens`` its extensions with RTLD_GLOBAL
disabled. This means that if you load the Python interpreter inside
Matlab without extra tricks, any extensions modules loaded by Python
will not be able to see Python symbols. Which means that they do not
work, which is especially annoying since you surely would like to use
for example numpy_.
To work around this nasty limitation, ``pythoncall.c`` has an equally nasty
hack: it explicitly loads symbols in the Python library to its own
"namespace", by ``dlopening`` the Python library file. To do this,
it needs to know the location of Python's ``libpython2.X.so`` file.
Have a look at the ``Makefile`` and fix the location of ``LIBPYTHON``
to match your version of Python.
While you're at it, also change ``-DNUMPY`` to ``-DNUMARRAY`` or
``-DNUMERIC`` if you use one of the earlier array libraries of Python.
.. _numpy: http://www.scipy.org/
Type conversion
---------------
The whole point of having Python inside Matlab is to have easy
conversion of variables. In ``pythoncall`` this goes as follows:
=================== ========================================================
From To
=================== ========================================================
Python int Matlab 1x1 int32
Python float Matlab 1x1 double
Python complex Matlab 1x1 double complex
Python bool Matlab 1x1 logical
Python dict Matlab 1x1 struct array. Field names: repr(key)
Python object array Matlab cell array
Python array Matlab array
Python sequence Matlab 1xN cell array
Python None Matlab empty cell array
Python other Matlab empty cell array + warning
------------------- --------------------------------------------------------
Matlab cell array Python list if one-dim, otherwise python object array
Matlab char array Python string if one-dim, otherwise python numeric array
Matlab struct array Python dict. If 1x1, values are converted directly.
If not scalar, values are python arrays/obj-arrays,
same size as the struct.
Matlab array Python array
Matlab other Python None + warning
=================== ========================================================