[SciPy-User] Reshaping Question
Wed Nov 4 21:56:10 CST 2009
On Wed, Nov 4, 2009 at 9:05 PM, Anne Archibald
> 2009/11/4 David Warde-Farley <firstname.lastname@example.org>:
>> Hi Skipper,
>> No, I don't believe so. The reason is that NumPy arrays have to obey constant stride along each dimension. Assuming
>> dtype is int32, the reshaping you describe (assuming you want to reshape c into d) would require the stride along dim
>> 2 to be 4 bytes to get from 0 to 1, and then 12 bytes to get to 4, and then 4 bytes again to get to 5. This isn't
>> legal, you'd have to do a copy to construct this matrix.
Ah ok. This makes sense, and is kind of why I thought I couldn't do
what I wanted as easily, as I'd like.
> Reshape sometimes creates copies. It tries hard not to, and if you
> assign the shape attribute rather than calling reshape it won't ever
> make a copy, but if necessary reshape will copy the input array:
> In : np.transpose(c.reshape(2,2,2,2),(0,2,1,3)).reshape(4,4)Out:
> array([[ 0, 1, 4, 5],
> [ 2, 3, 6, 7],
> [ 8, 9, 12, 13],
> [10, 11, 14, 15]])
> The trick is to use transpose to do an arbitrary permutation of the
> input axes, and also to rearrange the first axis with an additional
This makes sense as well. This is kind of what I was looking for I
just couldn't figure out the permutation. I was trying to roll the
axes, though I guess this could still work if you add the extra axis.
I don't know if I'd use this in the end though, as it might sacrifice
too much readability in the code, but maybe that's just me...
What if I had the outermost container as a list? Say,
c = [np.arange(4).reshape(2,2),np.arange(4,8).reshape(2,2),np.arange(8,12).reshape(2,2),np.arange(12,16).reshape(2,2)]
I seem to be running into much the same problems trying to use list
comprehension to end up with d.
It seems like I'm going to need a copy anyway, so maybe I'd be better
off just allocating a new array and filling it up transparently?
>> On Wed, Nov 04, 2009 at 08:25:12PM -0500, Skipper Seabold wrote:
>>> My brain is failing me. Is there a clean way to reshape an array like
>>> the following?
>>> import numpy as np
>>> c = np.arange(16).reshape(4, 2, 2)
>>> In : c
>>> array([[[ 0, 1],
>>> [ 2, 3]],
>>> [[ 4, 5],
>>> [ 6, 7]],
>>> [[ 8, 9],
>>> [10, 11]],
>>> [[12, 13],
>>> [14, 15]]])
>>> So that c == d where
>>> d = np.array(([0, 1, 4, 5], [2,3,6,7], [8,9,12,13], [10, 11, 14, 15]))
>>> In : d
>>> array([[ 0, 1, 4, 5],
>>> [ 2, 3, 6, 7],
>>> [ 8, 9, 12, 13],
>>> [10, 11, 14, 15]])
>>> SciPy-User mailing list
>> SciPy-User mailing list
> SciPy-User mailing list
More information about the SciPy-User