As I understand your code, you first create a wrapper struct whose members look like the ParameterReader members.  Then you have functions to copy from the ParameterReader to the wrapper and back, all in c++.  Then in params.py  you have a python version of the wrapper, which I think is perhaps hard coded to directly access the wrapper.  

The wrappers add a bit of risk of bugs since changes have to be made in multiple places.  Is there any way this could be avoided?  Is it possible to use SWIG? 

Yes SWIG is probably a good solution. 

In the meantime, update share and check genCtypesStructs.py. I fixed it to make it generate proper wrappers, but only for ParameterReader.

Usage: python share/genCtypesStructs.py include/ParameterReader.h include/ParameterReader_wrapper.h share/pyMMB/params.py 

(you have to be at the root of the project).

In the past, I used this script for the classes and structs in utils.h. But it doesn't produce ready to use wrappers. This is probably broken now anyway.

Alternatively, would it not be better, if the params.py were linked hard to ParameterReader rather than its wrapper?  I don’t see that there is any validation happening, which would be one reason to have ParameterReader_wrapper.h.  Is this possible? Bear in mind I’m a python amateur.

No, unfortunately you can't do that. Python can execute C functions, but not C++. Indeed, once compiled, C++ method names are modified (see C++ name mangling for more info), making it unreliable to be wrapped. 

This is why we have these wrappers. Notice that the code of python_wrapper.cpp is wrapped inside an extern "C" declaration, marking it as C code for programs using the library. Note that only the signature of the functions has to follow the C rules. Their guts can totally use c++, and this is what we do here.

And actually, python_wrapper.cpp is quite ill named as it is really a C wrapper. The python wrappers are in wrappers.py.

SWIG is doing the same thing, except that it's automatized.