functions of ndimage

[Juan Nunez-Iglesias]

On Wed, Apr 24, 2013 at 4:54 PM, Nelle Varoquaux
<nelle.varoquaux at gmail.com>wrote:

> I don't agree with this point: if you take the case of scikit-learn, there
> are some methods from scipy wrapped in scikit-learn, where the API changed
> to match the scikits (all classifiers are classes, with a fit method, a
> predict method, etc). The API of methods in skimage should match skimage's
> API, not scipy's.
>

Hi Nelle! =) This is a fair point, but scikit-learn has a very stable, very
well-defined interface for its estimators. There are two differences in the
case of scikit-image: 1) most of the API is similar to the scipy ndimage
API in the first place, implemented as functions that take an image and
return an image of the same shape.
2) it's not as consistent overall (e.g. graph.MCP is an object,
graph.shortest_path is a function).

So, I would recommend aiming for new interfaces in scikit-image to be as
close as possible to the scipy.ndimage interface. If a few kwargs are lost
in translation, well, that's not a big deal. =)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/scikit-image/attachments/20130424/e6225ee6/attachment.html>