[Numpy-discussion] Another flags question

[Albert Strasheim]

Hello all

Me vs the flags again. I found another case where the flags aren't what I 
would expect:

In [118]: x = N.array(N.arange(24.0).reshape(6,4), order='F')

In [119]: x
Out[119]:
array([[  0.,   1.,   2.,   3.],
       [  4.,   5.,   6.,   7.],
       [  8.,   9.,  10.,  11.],
       [ 12.,  13.,  14.,  15.],
       [ 16.,  17.,  18.,  19.],
       [ 20.,  21.,  22.,  23.]])

In [120]: x[:,0:1]
Out[120]:
array([[  0.],
       [  4.],
       [  8.],
       [ 12.],
       [ 16.],
       [ 20.]])

# start=0, stop=1, step=2
In [121]: x[:,0:1:2]
Out[121]:
array([[  0.],
       [  4.],
       [  8.],
       [ 12.],
       [ 16.],
       [ 20.]])

In [122]: x[:,0:1].flags
Out[122]:
  C_CONTIGUOUS : False
  F_CONTIGUOUS : True
  OWNDATA : False
  WRITEABLE : True
  ALIGNED : True
  UPDATEIFCOPY : False

In [123]: x[:,0:1:2].flags
Out[123]:
  C_CONTIGUOUS : False
  F_CONTIGUOUS : False
  OWNDATA : False
  WRITEABLE : True
  ALIGNED : True
  UPDATEIFCOPY : False

In [124]: x[:,0:1].strides
Out[124]: (8, 48)

In [125]: x[:,0:1:2].strides
Out[125]: (8, 96)

The views are slightly different (as can be seen from at least the strides), 
but I'd expect F_CONTIGUOUS to be true in both cases. I'm guessing that 
somewhere this special case isn't being checked for, which translates into a 
"missed opportunity" for marking the view as contiguous. Probably not a bug 
per se, but I thought I'd mention it here.

Cheers,

Albert