scipy.fft.register_backend#

scipy.fft.register_backend(backend)[source]#

Register a backend for permanent use.

Registered backends have the lowest priority and will be tried after the global backend.

Parameters:
backend{object, ‘scipy’}

The backend to use. Can either be a str containing the name of a known backend {‘scipy’} or an object that implements the uarray protocol.

Raises:
ValueError: If the backend does not implement numpy.scipy.fft.

Examples

We can register a new fft backend:

>>> from scipy.fft import fft, register_backend, set_global_backend
>>> class NoopBackend:  # Define an invalid Backend
...     __ua_domain__ = "numpy.scipy.fft"
...     def __ua_function__(self, func, args, kwargs):
...          return NotImplemented
>>> set_global_backend(NoopBackend())  # Set the invalid backend as global
>>> register_backend("scipy")  # Register a new backend
# The registered backend is called because
# the global backend returns `NotImplemented`
>>> fft([1])
array([1.+0.j])
>>> set_global_backend("scipy")  # Restore global backend to default