How to create a Numpy ufunc over user-defined data types?

Question

The Numpy C API contains a function PyUFunc_FromFuncAndData for creating a PyUFuncObject data structure. This function takes a char[] types argument that specifies the function's built-in argument types (int, float, etc.). But, the type number for a user-defined data type is an int, not a char. If I use the C-API to create a custom (user-defined) data type, how do I create a ufunc that operates on this data type?

Steve Mitchell · Accepted Answer · 2016-03-28 23:35:04Z

1

I don't see this explained in the Numpy documentation, but there are examples around.

int my_type_num = PyArray_RegisterDataType(&my_custom_type_descr);

PyUFuncObject* ufunc = (PyUFuncObject*)PyUFunc_FromFuncAndData(
    NULL, NULL, NULL, 0, nin, nout, identity, name, doc, 0);
PyUFunc_RegisterLoopForType(ufunc, my_type_num, my_loop_func, NULL, NULL);

This was discussed relative to an older version of Numpy on the Numpy mailing list. The source code for these API functions is here.

answered Mar 28, 2016 at 23:35

Steve Mitchell

2,0981 gold badge16 silver badges17 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

How to create a Numpy ufunc over user-defined data types?

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related