Whether or not it is a good idea (and it probably is considered bad practice), I sometimes return an Error in python instead of raising it. def fnA( some_param ): # use some_param in complicated thing if some_param is None: return ValueError( "something went wrong" ) def fnB( param_a ): val = fnA( param_a ) if isinstance( val, Exception ): # error path I have been trying to create a python C extension from a regular module that I have written both as practice and because I think it could benefit from it. To do that, I have been using pybind11. It covers most of my needs, except that. I cannot seem to create a ValueError and return it from C (or C++ in this case). I have looked at the documentation, at python's own source code and I have not found a way to create an exception without raising it. What I have "found"/tried is this: auto some_fn() -> PyExc_ValueError* { PyObject* type; PyObject* value; PyObject* traceback; PyErr_SetString( PyExc_ValueError, "error in my function !!!" ); PyErr_Fetch( &type, &value, &traceback ); PyErr_NormalizeException( &type, &value, &traceback ); if( traceback != NULL ) { PyException_SetTraceback( value, traceback ); } PyErr_Clear(); return value; } which sets the exception, fetches it, normalizes it, clears it and returns the concrete error. I thought it would work but it does not even compile. Anyone can point me in the right direction for creating and returning an exception in the C API? It might not really be possible, but then again, it is so easy in python code, I have a hard time believing it can't be done. Thanks in any case. Continue reading...
