skimage.io.Image??

[Emmanuelle Gouillart]

Oops sorry Fernando, and thanks Stéfan for sharing a healthy reminder
that the deprecation policy is very useful, and should be enforced with
care. 

On Mon, Apr 18, 2016 at 01:03:38PM -0700, Stéfan van der Walt wrote:
> Hey Fernando (& skimage list, CC'd)

> On 18 April 2016 at 01:43, Fernando Perez <fperez at lbl.gov> wrote:
> > Quick question, where did this go?

> > http://scikit-image.org/docs/0.11.x/api/skimage.io.html#image

> > It's not there in 0.12 and newer, and nowhere do the docs indicate that it
> > was removed, moved or otherwise tell me what I should do:

> > http://scikit-image.org/docs/dev/api_changes.html#version-0-12

> This was an oversight; we have a two release deprecation policy (with
> inline warnings + mentions in the docs) which was not followed in this
> case.  The internal API change was documented in the v0.9 api_changes
> document, but unfortunately not the rest.

> > We have an example in IPython that uses it:
> > https://github.com/ipython/ipywidgets/blob/master/docs/source/examples/Image%20Processing.ipynb

> I spoke to Brian several months ago to fix the issue after this came
> to our attention, but I guess the changes have not made their way
> through to ipywidgets.

> > Any help appreciated (and you shouldn't remove classes like this from the
> > API without deprecation period, warning or mention in the docs :)

> Of course, you are right, this type of thing should never, ever
> happen--I apologize for the confusion and wasted time.

> The Image class, just like most other numpy subclasses, turned out to
> be a mistake.  It wouldn't carry through the whole pipeline, and ended
> up causing more confusion than what it helped.

> I've factored out the source and attached it here, in case that is of some help:

> from image import Image
> from skimage import data
> cat = data.chelsea()
> Image(cat)

> Stéfan