[Insight-users] itk::SmartPointer - problem making code const-correct

[Stuart Golodetz]

  Frederic Perez wrote:
> Hello Stuart,
>
> I would like to be able to do those kinds of assignments too!
>
> For example, in our group we have functions like this one:
>
>      template<typename TITKImgType>
>      bool
>      WriteFile(
>          const std::string& a_fullFilename,
>          typename TITKImgType::ConstPointer);
>
> declared this way to promise that WriteFile won't modify the input image.
> (By the way, the template stuff is not relevant to this thread.)
>
> In order to workaround the const-vs-nonconst problem, we do calls like 
> this:
>
> typedef itk::Image<int, 3> ImgType;
> ImgType::Pointer image = ImgType::New();
> ...
> WriteFile<ImgType>("out.mhd", image.GetPointer());
>
> Unfortunately, this populates the code with lots of GetPointer() that 
> we would like to get rid of.
> Not a real solution, but maybe you can use it.
Seems better than my "solution" at least, cheers :) The other 
alternative I'd been contemplating was to write a wrapper that can be 
(implicitly) constructed from either a Pointer or a ConstPointer. You 
then use parameters of the wrapper type (or const& to the wrapper), and 
unwrap them within the function. That might make the called code 
marginally longer, but cleans up the code at the call sites.

I guess if you wrote a full wrapper with a conversion operator back to 
itk::SmartPointer then you might be able to make the whole thing 
relatively seamless. It would be a lot better if they just changed 
itk::SmartPointer to allow this, though - unless there's some compelling 
reason not to that I've missed of course. I wonder how much bureaucracy 
would be required for anyone to even contemplate that sort of change 
though(!)

Cheers,
Stu
> I would like to hear more on this issue.
>
> Cheers,
>
> Frederic Perez
>
> On Thu, May 27, 2010 at 7:21 PM, Stuart Golodetz <itk at gxstudios.net 
> <mailto:itk at gxstudios.net>> wrote:
>
>     Hi,
>
>     Hope this is the right place to post this. I was just wondering if 
>     there's a reason why itk::SmartPointer was designed so as not to allow e.g.
>
>     itk::Image<int,3>::Pointer image;
>     const itk::Image<int,3>::ConstPointer& cimage = image;
>
>     ?
>
>     The equivalent with boost::shared_ptr is allowed, e.g.
>
>     boost::shared_ptr<int> p(new int);
>     const boost::shared_ptr<const int>& cp = p;
>
>     This doesn't seem like a major problem, until you start writing 
>     functions taking const itk::Image<...>::ConstPointer& parameters - at 
>     which point it won't let you pass a normal Pointer in without explicitly 
>     constructing a ConstPointer from it. Now the types are often quite long, 
>     and it's annoying to have to add extra typedefs in the calling code just 
>     for that purpose. Duplicating the functions with const 
>     itk::Image<...>::Pointer& parameters doesn't work either, because you 
>     get a combinatorial explosion when you have multiple such parameters. 
>     For instance, with 3 parameters, you have to create functions with 
>     combinations:
>
>     const Pointer&, const Pointer&, const Pointer&
>     const Pointer&, const Pointer&, const constPointer&
>     const Pointer&, const ConstPointer& const Pointer&
>     // more here
>     const ConstPointer&, const ConstPointer& const ConstPointer&
>
>     This seems like an unproductive way to spend one's time, to say the 
>     least. The only other "reasonable" alternative I've managed to come up 
>     with that doesn't either (a) clutter up the call site or (b) cause the 
>     combinatorial explosion just outlined, is to just use the non-const 
>     Pointers everywhere and abandon the idea of making the code 
>     const-correct. But that seems defeatist to me  :)  Please could you tell 
>     me if there's something I'm missing? (And if so, what?)
>
>     Cheers,
>     Stuart
>         
>
>
>     _____________________________________
>     Powered by www.kitware.com <http://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://www.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-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20100528/9259166d/attachment-0001.htm>