SWIG 'cstring.i' Python return byte type instead of string type

Question

I have a C function like

int foo(void ** buf, int * buf_size)

And I use cstring.i to wrap it for use in Python 3. The result of the wrapped Python function is of type string.

Is there a way to get the result as type binary?

Background: The data buf gets filled with is msgpack encoded data, so using str.decode in Python is not an option. The msgpack implementations for Python only accept binary data.

Answer 1

If you use %cstring_output_allocate_size the wrapper function _wrap_foo calls SWIG_FromCharPtrAndSize() which has the following decoding logic:

#if PY_VERSION_HEX >= 0x03000000
#if defined(SWIG_PYTHON_STRICT_BYTE_CHAR)
      return PyBytes_FromStringAndSize(carray, (Py_ssize_t)(size));
#else
#if PY_VERSION_HEX >= 0x03010000
      return PyUnicode_DecodeUTF8(carray, (Py_ssize_t)(size), "surrogateescape");
#else
      return PyUnicode_FromStringAndSize(carray, (Py_ssize_t)(size));
#endif
#endif
#else
  return PyString_FromStringAndSize(carray, (Py_ssize_t)(size));
#endif

So you can get bytes instead of unicode string by #defining SWIG_PYTHON_STRICT_BYTE_CHAR. This is documented in http://www.swig.org/Doc3.0/Python.html so it's an official feature. But since it's a global switch, it's only useful if you want all string parameters to be mapped to bytes. If you need a mix of str and bytes in your API the only solution I can see is a custom typemap.