Changing behavior of `__getitem__` and `__setitem__` in Numpy array subclass

Question

Numpy arrays can be efficiently subclassed, but I want to modify the behavior of __getitem__ and __setitem__ so they can take in a datetime range, while preserving the maximum amount of built-in machinery like operations, cumsum, etc. Can this be done with __array_ufunc__?

It appears that in their example, the numpy.ufunc.at method is overridden.

Could this be used to modify the get/set behavior of numpy arrays?

Answer 1

You can implement __getitem__ and __setitem__ to handle your specific cases (with datetime objects) and dispatch to super().__{get|set}item__ in other cases. That way the remaining functionality of ndarray remains preserved. For example:

from datetime import date
import numpy as np

class A(np.ndarray):
    def __array_finalize__(self, obj):
        if obj is not None:
            obj.start_date = date.today()

    def __getitem__(self, item):
        if isinstance(item, slice) and isinstance(item.start, date) and isinstance(item.stop, date):
            return super().__getitem__(slice((item.start - self.start_date).days,
                                             (item.stop - self.start_date).days,
                                             item.step))
        return super().__getitem__(item)

a = A((10,), buffer=np.arange(10), dtype=int)
print(a[1:8])
print(a[date(2019, 1, 22):date(2019, 1, 29):2])
print(np.cumsum(a))
print(np.add.outer(a, a))

Which outputs:

[1 2 3 4 5 6 7]
[1 3 5 7]
[ 0  1  3  6 10 15 21 28 36 45]
[[ 0  1  2  3  4  5  6  7  8  9]
 [ 1  2  3  4  5  6  7  8  9 10]
 [ 2  3  4  5  6  7  8  9 10 11]
 [ 3  4  5  6  7  8  9 10 11 12]
 [ 4  5  6  7  8  9 10 11 12 13]
 [ 5  6  7  8  9 10 11 12 13 14]
 [ 6  7  8  9 10 11 12 13 14 15]
 [ 7  8  9 10 11 12 13 14 15 16]
 [ 8  9 10 11 12 13 14 15 16 17]
 [ 9 10 11 12 13 14 15 16 17 18]]