[Numpy-discussion] Inplace reshape

[Charles R Harris]

On 4/23/07, Christopher Barker <Chris.Barker@noaa.gov> wrote:
>
> Gael Varoquaux wrote:
> > Unless I miss something obvious "a.reshape()" doesn't modify a, which is
> > somewhat missleading, IMHO.
>
> quite correct. .reshape() creates a new array that shared data with the
> original:


Sometimes it do, sometimes it don't:

In [8]: x = ones((8,8))

In [9]: y = x[:3,:6]

In [10]: z = y.reshape(2,9)

In [11]: z[...] = 0

In [12]: x
Out[12]:
array([[ 1.,  1.,  1.,  1.,  1.,  1.,  1.,  1.],
       [ 1.,  1.,  1.,  1.,  1.,  1.,  1.,  1.],
       [ 1.,  1.,  1.,  1.,  1.,  1.,  1.,  1.],
       [ 1.,  1.,  1.,  1.,  1.,  1.,  1.,  1.],
       [ 1.,  1.,  1.,  1.,  1.,  1.,  1.,  1.],
       [ 1.,  1.,  1.,  1.,  1.,  1.,  1.,  1.],
       [ 1.,  1.,  1.,  1.,  1.,  1.,  1.,  1.],
       [ 1.,  1.,  1.,  1.,  1.,  1.,  1.,  1.]])

In [13]: z
Out[13]:
array([[ 0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.],
       [ 0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.]

As Ann pointed out, sometimes reusing the same array isn't possible. This
makes the use of reshaped arrays as l-values subject to subtle errors, so
caution is advised.

Chuck
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://projects.scipy.org/pipermail/numpy-discussion/attachments/20070423/ebcd14c2/attachment.html