[Insight-users] Cswig_java/Wrap_ITK generate different java APIs

[Gaëtan Lehmann]

Le Thu, 01 Feb 2007 21:02:26 +0100, Blair Cruz <blair.cruz at gmail.com> a  
écrit:

> Quick question.  I noticed that when I build ITK with the cswig_java  
> option
> I get a jar that has a different api then when I use the wrap_itk  
> option.  I
> particular, the class names change, an example is:
>
> itkMutualInformationImageToImageMetricIF3IF3_Pointer is generated with  
> the
> wrap_itk option in cmake, and:
> itkMutualInformationImageToImageMetricF3F3_Pointer is generated with the
> cswig_java option in cmake.
>
> There are a few other differences in the API as well, for instance, with  
> the
> wrap_itk option I cant set a
> itkRecursiveMultiResolutionPyramidImageFilterIF3IF3_Pointer object to a
> itkMultiResolutionImageRegistrationMethodIF3IF3_Pointer object.

The name of the template instantiations have been changed to be more  
consistent in all the instantiations, and more descriptive.

> I can get
> by without this but I do notice a performance difference.

I doubt you can see any difference: cswig and WrapITK both use the same  
system (cable swig), and wrap the same code (ITK).

> There is also a
> very noticeable size difference in the compiled libraries (~60mb).

WrapITK produce huge binaries because it instantiate a lot of templated  
classes. On one side you get a lot much classes, but on the other side,  
you have bigger binaries.
Note that you can easily customize your WrapITK to fit your needs, by  
removing the wrap_*.cmake files of the class you don't use, or by turning  
to OFF the WRAP_ options of the modules or the types you don't use. All  
those methods will highly impact the size of the binary.
You may also want to build WrapITK in MinSizeRel mode, and to turn on  
explicit instantiations in cmake.

>  I would
> use the cswig_java option however it wont build with this on os x.

cswig has open the way, but should be replaced by WrapITK in the next  
releases. I don't think cswig has ever worked properly out of the box with  
java on mac os x.

>
> This is not an issue right now but in the future I will have to address  
> some
> performance issues and I would like to know why this is happening.
>

I hope to have replied above. Please let us know if you have any problem,  
or if you think to things which could be enhanced.

Regards,

Gaetan



-- 
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