Type casting python int to long in scipy.weave.inline

Question

Given the following code:

from scipy.weave import inline
code = "return_val = input_val + 1;"
inline(code, ["input_val"], local_dict=dict(input_val=-9223372036854775808))

The resulting C code internally converts the long int into int and happily returns 1. Trying to specify input_val=np.int64(-9....) or input_val=long(-9...) gives a compilation error, as the numpy int type is not converted at all and gets treated as a PyObject. I saw some mentions of a type_converters keyword argument for inline , but the documentation unfortunately doesn't detail its handling further. Any ideas, how to force type conversion to long instead of int?

Answer 1

This finally helped. My compiler didn't like the inbuilt long_converter. So this overrides some type settings for long and prepends the default converter list:

from scipy.weave import inline, converters, c_spec

class long_converter(c_spec.scalar_converter):
    def init_info(self):
        c_spec.scalar_converter.init_info(self)
        # !! long to int conversion isn't safe!
        self.type_name = 'long'
        self.check_func = 'PyLong_Check'
        self.c_type = 'long'
        self.return_type = 'long'
        self.to_c_return = "(long) PyLong_AsLongLong(py_obj)"
        self.matching_types = [types.LongType]

type_conversion = [long_converter()] + converters.default

inline("return_val = input_val + 1;", ["input_val"], 
        local_dict=dict(input_val=long(-9223372036854775808)), 
        type_converters=type_conversion)

>>> -9223372036854775807