I wrap a lot of C++ using the Python 2 API (I can't use things like swig or boost.python for various technical reasons). When I have to pass a string (usually a path, always ASCII) into C/C++, I use something like this:
std::string file_name = PyString_AsString(py_file_name);
if (PyErr_Occurred()) return NULL;
Now I'm considering updating to Python 3, where PyString_* methods don't exist. I found one solution that says I should do something like this:
PyObject* bytes = PyUnicode_AsUTF8String(py_file_name);
std::string file_name = PyBytes_AsString(bytes);
if (PyErr_Occurred()) return NULL;
Py_DECREF(bytes);
However this is twice as many lines and seems a bit ugly (not to mention that it could introduce a memory leak if I forget the last line).
The other option is to redefine the python functions to operate on bytes objects, and to call them like this
def some_function(path_name):
_some_function(path_name.encode('utf8'))
This isn't terrible, but it does require a python-side wrapper for each function.
Is there some cleaner way to deal with this?
解决方案
Looks like the solution exists in python 3.3, with char* PyUnicode_AsUTF8(PyObject* unicode). This should be exactly the same behavior as the PyString_AsString() function from python 2.