import numpy as np
from fastats.core.decorator import fs
def value(x):
return x[0]
@fs
def windowed_pass(x, win):
"""
Performs a rolling (windowed) iteration
over the first dimension of `x`.
This is a simplistic rolling window which
does not take into account NaN values,
and maintains the same fixed size window
over all data points.
The leading values will be nan up until
the window size.
Example
-------
>>> import numpy as np
>>> def mean(x):
... return np.sum(x) / x.size
>>> x = np.array(range(10), dtype='float')
>>> result = windowed_pass(x, 5, value=mean)
>>> result[:6]
array([nan, nan, nan, nan, 2., 3.])
>>> result[6:10]
array([4., 5., 6., 7.])
"""
result = np.full_like(x, np.nan)
for i in range(win, x.shape[0]+1):
result[i-1] = value(x[i-win:i])
return result
@fs
def windowed_pass_2d(x, win):
"""
The same as windowed pass, but explicitly
iterates over the `value()` return array
and allocates it in the `result`.
This allows 2-dimensional arrays to be returned
from `value()` functions, before we support
the behaviour properly using AST transforms.
This allows for extremely fast iteration
for items such as OLS, and at the same time
calculating t-stats / r^2.
"""
result = np.full_like(x, np.nan)
for i in range(win, x.shape[0]+1):
res = value(x[i-win:i])
for j, j_val in enumerate(res):
result[i-1, j] = j_val
return result
if __name__ == '__main__':
import pytest
pytest.main([__file__])
