[Insight-users] Python Wrapping Coverage for ITK

[Hua Qian]

Hi Luis,

Thank you every much for your response. I was about to
follow up the thread myself. I agree that an evolutionary
approach is more appropriate here.

By the way I've collected some wrappers from our local
users. What's the best way to contribute: just giving
the wish list, or sending the the wrapping files?
I understand that sometimes it is easier writing code
yourself and doing quality control.

Regards,
Hua

Luis Ibanez wrote:

>
> Hi Hua,
>
> Thanks a lot for your detailed information  about
> the classes needing wrapping for multi-resolution
> registration.
>
> Wrappers your the classes in your list have been
> added now.
>
> >   MultiResolutionImageRegistrationMethod
> >   RecursiveMultiResolutionPyramidImageFilter
> >   ChangeInformationImageFilte
> >   FlipImageFilter
> >   NormalizeImageFilter
> >   PermuteAxesImageFilter
>
>
> Massive wrapping is not impossible but it will be
> quite impractical. For the classes in BasicFilters
> it may not be that bad, but when you go to Algorithms
> and Statistics, there are way too many possible
> combinations.
>
> The "Evolutionary" method of simply wrapping the classes
> that are being used (and in the context of their use)
> seems to be the best approach to follow.
>
> For this year we are planning in increase a lot the
> coverage of classes wrapped. So, if you have more
> classes in you wrapping wish list, please let us know
> so we can make sure that they got included.
>
>
>
>
> Regards,
>
>
>
>    Luis
>
>
>
> -------------------
> Hua Qian wrote:
>
>> Hello All,
>>
>> As a test drive for my Python-ITK application,
>> I tried to re-implement the MultiResMIRegistration
>> application in Python recently. I didn't go very
>> far before I found that I have to wrap lots of
>> ITK classes myself, such as
>>
>>   MultiResolutionImageRegistrationMethod
>>   RecursiveMultiResolutionPyramidImageFilter
>>   ChangeInformationImageFilte
>>   FlipImageFilter
>>   NormalizeImageFilter
>>   PermuteAxesImageFilter
>>
>> While it is quick and easy to write a C++ wrapping
>> file for a class, it is slow to rebuild ITK
>> every time you add new wrapping files.
>> Now, I am wondering if I could get all the
>> building work done in one shot. My questions are:
>>
>> Is feasible to wrap all the ITK C++ class (at
>> least for US3, F3 and combinations) in python
>> (and Tcl)?
>>  
>> Does anyone know the coverage of python/tcl
>> wrapping in ITK 1.4 distribution? I am talking
>> about the class base, not all the template
>> parameter combinations.
>>
>> By the way, I attached a little python program
>> that generates a list of ITK python classes
>> (output: /usr/tmp/itklist.txt). I use the list
>> to check what classes are available in Python.
>>
>> Regards,
>> Hua
>>
>>
>> ------------------------------------------------------------------------
>>
>> #! /usr/bin/env python
>>
>> import InsightToolkit as itk
>>
>> file = open('/usr/tmp/itklist.txt', 'w')
>>
>> itklist = dir(itk)
>> for item in itklist:
>>     file.write(item)
>>     file.write('\n')
>>
>> file.close()
>