A Matlab package for wrapping Python classes to make them act just like Matlab classes.
To use it, subclass the light_python_wrapper.light_python_wrapper abstract base class, and define your own constructor.
In particular, the subclass must define the pyobj and classname properties.
pyobj is a reference to the constructed Python object and classname is a string being the full class name as returned by type(obj) in Python.
For example, to wrap a class Bar in module foo.py which is on the Python path,
the Matlab constructor could specify obj.pyobj = py.foo.Bar() and classname = 'foo.Bar'.
The constructor should then call the obj.populate_props() method to create dummy Matlab properties so that tab-completion will work.
An example is given by the light_python_wrapper.generic_python_wrapper class.
The derived Matlab class can access all properties and call all methods of the wrapped Python class with the standard dot notation.
Hidden Python properties (which starts with _) can be accessed using brace notation, e.g. obj{'__str__'}()
The wrapper also performs some simple type conversions between Matlab and Python, but this is one area where many isses arise as Matlab and Python data types do not exactly match. Thus, you should carefully check the data transfered to Python and returned to Matlab are what you expect.
In particular the wrapper assumes numpy is installed and will convert Matlab numeric arrays to numpy arrays,
in order to take advantage of features which ensure that fewer data copies are made (at least for an "in-process" Python environment).
However, the conversion from Matlab to numpy array is done internally by matlab and can sometimes result in
numpy arrays which are automatically "squeezed" (missing singleton dimensions) or transposed.
If you find issues with the type conversion, please create an issue here.
Finally, the wrapper provides a help method so that help(obj) will return the Python (pydoc) help information for that object, method or property.
There are also overloaded help and doc functions in the helputils folder which will work with the help class / doc class syntax
(e.g. without the brackets and without needing to construct an object),
but these are not part of the light_python_wrapper class and so the helputils folder needs to be added to the path separately.
The saveobj and loadobj methods of the wrapper are also overloaded to serialise dependent Python objects using pickle,
so they can be saved as to a Matlab mat file.
Note that objects which cannot be pickled (which do not have a __dict__ or __[get/set]state__ methods) cannot be saved correctly.
In these cases a warning will be given and a mat file will still be generated but it will not correctly reload.