[Numpy-discussion] Overriding numpy.ndarray.__getitem__?

[Keith Hughitt]

Okay, I found something that seems to do the trick for this particular
problem. Instead of just checking whether the input to __getitem__ is an
int, I also check the number of dimensions to make sure we are indexing
within the full cube, and not some sub-index of the cube:

if self.ndim is 3 and isinstance(key, int):
...

I think what was happening is that when repr() is called on the map, it
recursively walks through displaying one dimension at a time, and this is
what was causing my code to choke; the instantiation of a subclass only
makes sense for one of the three dimensions.

Keith

On Wed, Aug 17, 2011 at 3:25 PM, Keith Hughitt <keith.hughitt at gmail.com>wrote:

> Hi all,
>
> I have a subclass of ndarray which is built using using a stack of images.
> Rather than store the image header information separately, I overrode
> __getitem__ so that when the user indexes into the image cube a single image
> a different object type (which includes the header information) is returned
> :
>
> class ImageCube(np.ndarray):
> .
> .
> .
> def __getitem__(self, key):
>         """Overiding indexing operation"""
>         if isinstance(key, int):
>             data = np.ndarray.__getitem__(self, key)
>             header = self._headers[key]
>             return SingleImage(data, header)
>         else:
>             return np.ndarray.__getitem__(self, key)
>
>
> Everything seems to work well, however, now when I try to combine that with
> indexing into the other dimensions of a single image, errors relating to
> numpy's array printing arise, e.g.:
>
>
> >>> print imagecube[0,0:256,0:256]
> .
> .
> .
>
> /usr/lib/pymodules/python2.7/numpy/core/arrayprint.pyc in _formatArray(a,
> format_function, rank, max_line_len, next_line_prefix, separator,
> edge_items, summary_insert)
>     371             if leading_items or i != trailing_items:
>     372                 s += next_line_prefix
> --> 373             s += _formatArray(a[-i], format_function, rank-1,
> max_line_len,
>     374                               " " + next_line_prefix, separator,
> edge_items,
>     375                               summary_insert)
>
>
> I think the problem has to do with how I am overriding __getitem__: I check
> to see if the input is a single integer, and if it is, I return the new
> object instance. This should only occur when something like "imagecube[n]"
> is called, however, array2str ends up calling imagecube[x], even if the
> original thing you are trying to print is something like
> imagecube[0,1:256,1:256].
>
> Any ideas? I apologize if the explanation is not very clear; I'm still
> trying to figure out exactly what is going on.
>
> Thanks,
> Keith
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/numpy-discussion/attachments/20110817/65f2c90e/attachment.html>