I don't know why the number of elements changes when the type is converted to dtype view

>>> arr = np.zeros(2, dtype=np.uint16)
>>> arr
array([0, 0], dtype=uint16)
>>> arr.view(np.uint8)
array([0, 0, 0, 0], dtype=uint8)
>>> arr
array([0, 0], dtype=uint16)


>>> arr[0] = 0xabcd
>>> arr
array([43981,     0], dtype=uint16)
>>> arr.view(np.uint8)
array([205, 171,   0,   0], dtype=uint8)
>>> hex(205)
'0xcd'
>>> hex(171)
'0xab'


>>> arr[0] = 0xabcd
>>> arr[1] = 0x1234
>>> arr
array([43981,  4660], dtype=uint16)
>>> viewhex = lambda l: list(map(hex, l))


>>> viewhex(arr)
['0xabcd', '0x1234']
>>> viewhex(arr.view(np.uint8))
['0xcd', '0xab', '0x34', '0x12']
>>> viewhex(arr.view(np.uint32))
['0x1234abcd']