Inconsistent indexing of arrays with no elements, depending on shape (Trac #1171) #1769
Description
Original ticket http://projects.scipy.org/numpy/ticket/1171 on 2009-07-14 by trac user jguyer, assigned to unknown.
An empty, rank-0 array can be indexed with an empty list:
#!python
>>> import numpy as np
>>> np.__version__
'1.3.0.dev6304'
>>> a = np.zeros((0,))
>>> a[..., []]
array([], dtype=float64)
but an empty array of higher rank throws an index error
#!python
>>> b = np.zeros((1, 0))
>>> b[..., []]
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
IndexError: invalid index
We have a not-completely-insane reason for wanting the second case to return array([], shape=(1, 0), dtype=float64). At a minimum, it seems the two cases should be consistent.
I wonder if this is related to ticket:1169?
I see this on a 10.5.7 Mac, with Python 2.6.2 and NumPy 1.3.0dev and on a Debian etch machine running Python 2.4.4 and NumPy 1.2.1.