[Numpy] where does not support python scalar as the input

[sxjscience]

import mxnet as mx
mx.npx.set_np()
a = mx.np.sym.var('a')
a = mx.sym.var('a').as_np_ndarray()
mx.sym.np.where(a, a, 0)

Error message:

---------------------------------------------------------------------------
AssertionError                            Traceback (most recent call last)
<ipython-input-5-e7a6fe89b743> in <module>
----> 1 mx.sym.np.where(a, a, 0)

~/mxnet/python/mxnet/symbol/numpy/_symbol.py in where(condition, x, y)
   5501 
   5502     """
-> 5503     return _npi.where(condition, x, y, out=None)
   5504 
   5505 

~/mxnet/python/mxnet/symbol/register.py in where(condition, x, y, name, attr, out, **kwargs)

AssertionError: Argument y must be Symbol instances, but got 0

Also, the imperative case:

import mxnet as mx
mx.npx.set_np()
mx.np.where(mx.np.ones((10, )), mx.np.ones((10, )), 0)

Error message:

AssertionError                            Traceback (most recent call last)
<ipython-input-14-bdddf3065582> in <module>
----> 1 mx.np.where(mx.np.ones((10, )), mx.np.ones((10, )), 0)

~/mxnet/python/mxnet/numpy/multiarray.py in where(condition, x, y)
   7996            [ 0.,  3., -1.]])
   7997     """
-> 7998     return _mx_nd_np.where(condition, x, y)
   7999 
   8000 

~/mxnet/python/mxnet/ndarray/numpy/_op.py in where(condition, x, y)
   6035         return nonzero(condition)
   6036     else:
-> 6037         return _npi.where(condition, x, y, out=None)
   6038 
   6039 

~/mxnet/python/mxnet/ndarray/register.py in where(condition, x, y, out, name, **kwargs)

AssertionError: Argument y must have NDArray type, but got 0

[xidulu]

I actually mentioned this problem in the PR, #16829 (comment)
The author suggested that you should wrap the scalar withnp.array.

[samskalicky]

@apeforest assign [@reminisce ]

[sxjscience]

@xidulu @hgt312 In my use case, I'm using the symbolic interface so I cannot call mx.np.array.

[reminisce]

@hgt312

[hgt312]

I will add the scalar version soon.

[hgt312]

If both x and y are scalars, dtype of the output should be what? In official numpy, it may be int64 or float64 according to the inputs' type.

[sxjscience]

Are we able to make it numpy compatible?

[yzhliu]

Closed by #17249