It is awkward to check whether an obj is of a particular heap type (defined in an extension module)

[iritkatriel]

I wanted to check whether an object is of type array and ended up doing this:

    PyObject *exc = PyErr_GetRaisedException();
    PyObject *module = PyType_GetModuleByDef(Py_TYPE(lhs), &arraymodule);
    if (module == NULL) {
        if (!PyErr_Occurred() || PyErr_ExceptionMatches(PyExc_TypeError)) {
            /* lhs is not an array instance - ignore the TypeError (if any) */
            PyErr_SetRaisedException(exc);
            return 0;
        }
        else {
            _PyErr_ChainExceptions1(exc);
            return -1;
        } 
    }   
    PyErr_SetRaisedException(exc);
    return array_Check(lhs, get_array_state(module));

CC @ericsnowcurrently @encukou

[encukou]

In 3.14, add {Py_tp_token, Py_TP_USE_SPEC} to array_slots, then use PyType_GetBaseByToken(lhs, &array_spec, NULL)

[mdboom]

Is this something that should be documented in the "isolating extension modules" migration guide? Seems like it would be easy for extension authors to miss.

[iritkatriel]

In 3.14, add {Py_tp_token, Py_TP_USE_SPEC} to array_slots, then use PyType_GetBaseByToken(lhs, &array_spec, NULL)

This is better. Would be nicer if it just returned a Boolean so I won't need to deal with exception state.