Basically I want this chained slicing to overwrite the value (show the first element as a 2 instead of 1)
tst = np.array([1,2,3,4])
msk1 = [True, False, True, False]
msk2 = [True, False]
tst[msk1][msk2] = 2
tst
> array([1, 2, 3, 4])
>Solution :
This issue is that tst[msk1] is a copy by slicing, and not a view.
You can check it using tst[msk1].base that returns nothing (it would return the original array if a view). So when you further slice and modify tst[msk1][mask2] = 2, you actually modify the in memory copy, not tst.
To my knowledge, you cannot force a view with fancy indexing, so what you want is not directly achievable.
One workaround might be to get the indices of the first mask, and then to slice them with the second mask:
idx = np.where(msk1)[0][msk2]
tst[idx] = 2
# print(tst)
# [2 2 3 4]