[SciPy-dev] Should ndimage.measurements.* should return lists if index is a list?

[Thouis (Ray) Jones]

My last patch did not do the right thing for ndimage.extrema(), which
needs to return 4 values with the same semantics.  I've attached an
addendum.

Ray Jones


On Sat, Apr 18, 2009 at 16:47, Thouis (Ray) Jones <thouis at broad.mit.edu> wrote:
> Here's a first attempt at a patch.  It fixes the corner case of a
> single index to match the index type (scalar for scalar, list for
> single element sequence), also the case of index=[] to return [] for
> any measurement.  Finally, it improves type checking of the index
> type, which was only being checked for sum() and not any of the other
> measurements (?).
>
> Zach, would you mind reviewing?  If it looks good, I'll add some tests
> for the new functionality and submit a patch via trac.
>
> Ray
>
> On Fri, Apr 17, 2009 at 22:06, Zachary Pincus <zachary.pincus at yale.edu> wrote:
>>> Current behavior:
>>>>>> from scipy import ndimage
>>>>>> ndimage.maximum([[1,2], [3,4]], [[1,0],[0,2]], index=[1,2])
>>> [1.0, 4.0]
>>>>>> ndimage.maximum([[1,2], [3,4]], [[1,0],[0,2]], index=[1])
>>> 1.0
>>>>>> ndimage.maximum([[1,2], [3,4]], [[1,0],[0,2]], index=1)
>>> 1.0
>>>
>>> I think the second result should be [1.0].  We're using the ndimage
>>> code for image processing, and have to wrap the measurement functions
>>> to make the code general for the case where there is only one object
>>> in the image.
>>
>> I've run into this too, and had the same thought. It's such a corner-
>> case that it doesn't seem like gratuitous API-breakage to make this
>> fix...
>>
>> Zach
>> _______________________________________________
>> Scipy-dev mailing list
>> Scipy-dev at scipy.org
>> http://mail.scipy.org/mailman/listinfo/scipy-dev
>>
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0002-special-case-for-extrema.patch
Type: application/octet-stream
Size: 1216 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/scipy-dev/attachments/20090418/3efe0ca4/attachment.obj>