[Insight-developers] orientation

[Stephen Aylward]

For an interesting discussion on backward compatibility, read this
month's issue of The Source (starting on page 5):
http://www.kitware.com/products/thesource.html
and
http://www.kitware.com/products/archive/kitware_quarterly0708.pdf

In that article, Bill L, Simon W, I, and others have given our
opinions on what it means to be backward compatible (what is required,
how important it is,  when it should be broken, etc.).

Short answer (if you don't want to read the article) backward
compatibility concerns operation as well as compilation.   With the
suggested change, people's code may still compile, but its
operation/output would be quite different.  Furthermore, the way
itkImage works now is not actually "wrong" it is just very misleading.
  Albeit, the way various other filters use it is wrong because it is
misleading.  Additionally, there is probably code other there that
also depends on its current operation - so even if our tests pass, it
probably is more of an indication of the inadequacy of our tests
rather than the backward-compatibility of the change.

Stephen



On Wed, Aug 6, 2008 at 2:05 PM, kent williams
<norman-k-williams at uiowa.edu> wrote:
> In looking back through this thread I think I'm beginning to see where the
> problem is: SpatialObject is using methods that are not implemented
> consistently between itk::Image and itk::OrientedImage.
>
> It seems that OrientedImage performs as expected, and that Image does not,
> which suggests a possible solution -- rename itk::OrientedImage as
> itk::Image, and make itk::OrientedImage simply a derived class of itk::Image
> with no new behavior, for backwards compatibility.
>
> Would this break any existing code -- more to the point, is there code that
> depends on itk::Image behaving in an apparently incorrect manner?
>
>
> On 8/6/08 12:47 PM, "Rupert Brooks" <rupe.brooks at gmail.com> wrote:
>
>> Hi Simon,
>>
>> I would be very interested in the test you mention.   Maybe the wiki
>> is the right place to post it?
>>
>> Due to the problems with gradients on non-axis-aligned Oriented
>> images, i created my own class, FastOrientedImage.  Its in the insight
>> Journal somewhere.   I was not aware of oriented images having
>> problems in 2D and I thought I had used them successfully in the past.
>>  I've been planning to switch back to the more mainstream ITK approach
>> and retire my custom classes - thus - I'd be very interested in this
>> bug.
>>
>> Thanks
>> Rupert
>>
>
> _______________________________________________
> Insight-developers mailing list
> Insight-developers at itk.org
> http://www.itk.org/mailman/listinfo/insight-developers
>



-- 
Stephen R. Aylward, Ph.D.
Chief Medical Scientist
Kitware, Inc. - Chapel Hill Office
http://www.kitware.com
(518) 371-3971 x300