[Numpy-discussion] Is this an indexing bug?
Sturla Molden
sturla at molden.no
Tue Jun 19 06:14:22 EDT 2007
On 6/19/2007 12:19 PM, Sven Schreiber wrote:
> To be more specific, I would expect shape==(4,14).
>>> h = numpy.zeros((1,4,14))
>>> h[0,:,numpy.arange(14)].shape
(14, 4)
>>> h[0,:,:].shape
(4, 14)
>>>
h[0,:,numpy.arange(14)] is a case of "sdvanced indexing". You can also
see that
>>> h[0,:,[0,1,2,3,4,5,6,7,8,9,10,11,12,13]].shape
(14, 4)
Citing from Travis' book, page 83:
"Example 2: Now let X.shape be (10,20,30,40,50) and suppose ind1 and
ind2 are broadcastable to the shape (2,3,4). Then X[:,ind1,ind2] has
shape (10,2,3,4,40,50) because the (20,30)-shaped subspace from X has
been replaced with the (2,3,4) subspace from the indices. However,
X[:,ind1,:,ind2,:] has shape (2,3,4,10,30,50) because there is no
unambiguous place to drop in the indexing subspace, thus it is tacked-on
to the beginning. It is always possible to use .transpose() to move the
sups pace anywhere desired. This example cannot be replicated using take."
So I think this strange behaviour is actually correct.
Sturla Molden
More information about the NumPy-Discussion
mailing list