[Numpy-discussion] Wrong behavior of atleast_3d?

[Yannick Copin]

Hi,

I wonder if there's a "behavior bug" in atleast_3d, with respect to 
atleast_2d. In atleast_2d, the results on a 1D-array or on a single-element 
list of 1D-arrays are the same:

In [22]: atleast_2d(arange(3)).shape
Out[22]: (1, 3)

In [23]: atleast_2d([arange(3),]).shape
Out[23]: (1, 3)

On the contrary, with atleast_3d, the results differ between a 2D-array and a 
single-element list of 2D-arrays:

In [24]: atleast_3d(randn(3,3)).shape
Out[24]: (3, 3, 1)

In [25]: atleast_3d([randn(3,3),]).shape
Out[25]: (1, 3, 3)

Wouldn't it be more logical wrt atleast_2d that atleast_3d(randn(3,3)) return 
an array of shape (1,3,3)? (even though the docstring makes it clear that it 
*will* return a (3,3,1)-array).

I mostly use atleast_nd to be able to iterate over input arrays without having 
to know if it's a single array, or a list of arrays. For that, at_least2d has 
the correct behavior, but not atleast_3d.

Cheers.
-- 
    .~.   Yannick COPIN  (o:>*  Doctus cum libro
    /V\   Institut de physique nucleaire de Lyon (IN2P3 - France)
   // \\  Tel: (33/0) 472 431 968     AIM: YnCopin ICQ: 236931013
  /(   )\ http://snovae.in2p3.fr/ycopin/
   ^`~'^