use_cython

tenpy.tools.optimization.use_cython(func=None, replacement=None, check_doc=True)[source]

Decorator to replace a function with a Cython-equivalent from _npc_helper.pyx.

This is a decorator, which is supposed to be used in front of function definitions with an @ sign, for example:

@use_cython
def my_slow_function(a):
    "some example function with slow python loops"
    result = 0.
    for i in range(a.shape[0]):
        for j in range(a.shape[1]):
            #... heavy calculations ...
            result += np.cos(a[i, j]**2) * (i + j)
    return result

This decorator indicates that there is a Cython implementation in the file tenpy/linalg/_npc_helper.pyx, which should have the same signature (i.e. same arguments and return values) as the decorated function, and can be used as a replacement for the decorated function. However, if the cython code could not be compiled on your system (or if the environment variable TENPY_OPTIMIZE is set to negative values, or the environment variable TENPY_NO_CYTHON is “true”), we just pass the previous function.

Note

In case that the decorator is used for a class method, the corresponding Cython version needs to have an @cython.binding(True).

Parameters:
  • func (function) – The defined function

  • replacement (string | None) – The name of the function defined in tenpy/linalg/_npc_helper.pyx which should replace the decorated function. None defaults to the name of the decorated function, e.g., in the above example my_slow_function.

  • check_doc (bool) – If True, we check that the cython version of the function has the exact same doc string (up to a possible first line containing the function signature) to exclude typos and inconsistent versions.

Returns:

replacement_func – The function replacing the decorated function func. If the cython code can not be loaded, this is just func, otherwise it’s the cython version specified by replacement.

Return type:

function