[Insight-developers] itk::Index, itk::Offset, ... as subclass of itk::FixedArray?

[Gaëtan Lehmann]

Ok, back to the itk::Index, itk::Offset and itk::Size.
Those classes are real plain old data, while itk::FixedArray and  
subclasses are not.
As a consequence, itk::Index, itk::Offset and itk::Size have several  
properties which make them non implementable as a subclass of  
itk::FixedArray:

* they can be initialized with this kind of syntax:

    Size<3> size = {{1,2,3}};

I think this behavior can't be reproduced with non POD without using C+ 
+0x.

* their internal members are public, so they can be used without  
accessor.
I don't see how we can keep the following valid use case while  
implementing them as subclasses of itk::FixedArray:

    Index<3> idx = {{1,2,3}};
    long * array = idx.m_Index;

So it doesn't seem possible to change the way those classes are  
implemented without introducing a lot of backward compatibility  
problems, which are certainly not justified by the problem this is  
supposed to fix.

Le 16 nov. 10 à 16:56, Tom Vercauteren a écrit :

> Hi all,
>
> On Tue, Nov 16, 2010 at 16:35, Luis Ibanez <luis.ibanez at kitware.com>  
> wrote:
>>
>> Hi Gaetan,
>>
>> Probably only historical reasons...
>>
>> I don't see why we couldn't make the itk::Index and
>> the itk::Offset be derived classes of the itk::FixedArray.
>>
>> Note however that Dave Cole is working hard on
>> fixing the Windows 64 bits issues, so you may want
>> to coordinate those changes (since the itk::Index and
>> the itk::Offset are at the heart of the 64bits problem).
>>
>> --
>>
>> Regarding the intentional lack of a default constructor,
>> we may have to run these experiments again.  I would
>> suspect that it is still a valid issue, given that this relates
>> to making the itkFixedArray appear as plain-old-data,
>> instead of as a class. (there was also an issue of padding
>> if I remember correctly...)
>
> Indeed, currently it is still easy to access the memory help by an
>  itk::Image< itk::Vector<float,2>, 2 >
> by a simple
>  float * floatptr = image->GetBufferPointer()[0].GetDataPointer()
>
> Not sure if it is standard conformant since itk::Vector seems to be
> only POD like and not strictly POD but at least it works on the
> systems I have tested so far. This is important to avoid copying
> memory between different libraries.
>
> This works because itk::Vector inherits from itk::FixedArray and is
> also POD-like. The less POD it becomes the more fragile my hypothesis
> becomes. Remember that we had to remove an empty destructor to make
> itk::FixedArray closer to being POD. This led to a better memory
> alignment and improved the performance of some of our code by a factor
> 2:
> http://www.itk.org/mailman/private/insight-developers/2008-June/010444.html
>
> Note that I am not the only one caring about this:
> http://www.itk.org/pipermail/insight-users/2010-October/038525.html
>
> Anyhow, we may want to improve the documentation of itk::FixedArray
> with respect to that point.
>
> My two cents,
> Tom
>
>>
>> What would be the argument in favor of adding the
>> default constructor ?
>>
>> (initialization ?,
>> note that we don't know what T is, in advance...
>> I'm not sure that we can provide a reasonable
>> initialization anyways.  T could be a std::string,
>> or a highly complex class...)
>>
>>
>>      Luis
>>
>>
>> ------------------------------------------------------------------------
>> 2010/11/16 Gaëtan Lehmann <gaetan.lehmann at jouy.inra.fr>
>>>
>>> Hi,
>>>
>>> Is there a reason to have the basic array types itk::Index and
>>> itk::Offset, and maybe others, not implemented as a subclass of
>>> itk::FixedArray?
>>> Also, I see in the doc that no constructor are provided for  
>>> performance
>>> reasons. Is it still true with the compilers supported by itk 4?
>>>
>>> Thanks,
>>>
>>> Gaëtan
>>>
>>>
>>> --
>>> Gaëtan Lehmann
>>> Biologie du Développement et de la Reproduction
>>> INRA de Jouy-en-Josas (France)
>>> tel: +33 1 34 65 29 66    fax: 01 34 65 29 09
>>> http://voxel.jouy.inra.fr  http://www.itk.org
>>> http://www.mandriva.org  http://www.bepo.fr
>>>
>>>
>>> _______________________________________________
>>> Powered by www.kitware.com
>>>
>>> Visit other Kitware open-source projects at
>>> http://www.kitware.com/opensource/opensource.html
>>>
>>> Kitware offers ITK Training Courses, for more information visit:
>>> http://kitware.com/products/protraining.html
>>>
>>> Please keep messages on-topic and check the ITK FAQ at:
>>> http://www.itk.org/Wiki/ITK_FAQ
>>>
>>> Follow this link to subscribe/unsubscribe:
>>> http://www.itk.org/mailman/listinfo/insight-developers
>>>
>>
>>
>> _______________________________________________
>> Powered by www.kitware.com
>>
>> Visit other Kitware open-source projects at
>> http://www.kitware.com/opensource/opensource.html
>>
>> Kitware offers ITK Training Courses, for more information visit:
>> http://kitware.com/products/protraining.html
>>
>> Please keep messages on-topic and check the ITK FAQ at:
>> http://www.itk.org/Wiki/ITK_FAQ
>>
>> Follow this link to subscribe/unsubscribe:
>> http://www.itk.org/mailman/listinfo/insight-developers
>>
>>

-- 
Gaëtan Lehmann
Biologie du Développement et de la Reproduction
INRA de Jouy-en-Josas (France)
tel: +33 1 34 65 29 66    fax: 01 34 65 29 09
http://voxel.jouy.inra.fr  http://www.itk.org
http://www.mandriva.org  http://www.bepo.fr

-------------- next part --------------
A non-text attachment was scrubbed...
Name: PGP.sig
Type: application/pgp-signature
Size: 203 bytes
Desc: Ceci est une signature ?lectronique PGP
URL: <http://www.itk.org/mailman/private/insight-developers/attachments/20101117/8a3b8ce9/attachment.pgp>