Confusing ndarray repr: always shows dtype as if in native byteorder

[Zac-HD]

>>> from numpy import array, dtype

# The repr of a dtype object distinguishes non-native byteorder.
>>> dtype('<u2')
dtype('uint16')
>>> dtype('>u2')
dtype('>u2')

# But arrays always use the repr of the native endianness
>>> array([0], dtype='<u2')
array([0], dtype=uint16)
>>> array([0], dtype='>u2')
array([0], dtype=uint16)  #   <-- this is confusing

This isn't strictly speaking a bug, but using the byteorder-sensitive repr of the dtype instead would make it easier to tell when you have an endianness issue in e.g. this StackOverflow question.

[Zac-HD]

@mattip - any tips on where I should be looking to fix this?

Twenty minutes with grep and examining various __repr__-related code snippets didn't get me anywhere, though I can see from https://github.com/numpy/numpy/blob/main/numpy/core/_dtype.py that it might be tricky if I can't simply have the array repr defer to the dtype repr...

[seberg]

Closed by gh-23295