[ITK-users] [ITK] auto keyword and New() smart pointer

Tue Mar 21 10:59:39 EDT 2017

Hello,

Your code and description illustrates the problem with the implicit auto type.

Based on this simple test I think that it is probably safe to use auto to store the output from filters. It does not look like it interferes with the reference counting or leads to memory leaks, as Yann was worried about.

This can be problematic:
  auto testOutput = testFilter->GetOutput();

But that is not what you are doing in your example. When you use the “auto” type you need to be careful to determine if the type is a raw pointer or a smart pointer. The GetOutput methods returns an raw pointer without reference counting. This is the type auto is above.

However in your example the return type for your method is ImageType::Pointer, which is an ITK smart pointer. Your code adds an implicit cast for the raw pointer to ITK’s smart pointer. And this is not made clear by the usage of auto.

HTH,
Brad

On Mar 21, 2017, at 10:49 AM, Kiran Joshi <kiran.j88 at gmail.com<mailto:kiran.j88 at gmail.com>> wrote:

Hi Francois,

Maybe I misunderstood, but I disagree with your statement that "you cannot use the keyword auto with these other methods"

Attached is a screenshot showing a small ITK test program in a Visual Studio debugging session.

The program is very simple; an AddImages() function is called, which itself calls a CreateImage() function twice before adding the created images and returning the result.
Note that I have used auto when storing the image pointers returned by the CreateImage and AddImage functions.

On the right of the image you can see snapshots of the memory usage as the program executes, and I have also annotated which lines of the program were executed immediately before taking the snapshot.

The first snapshot show that on line 22, very little memory has been allocated, since we have not yet created any images.
On line 23 ~16MB of memory is used when the first image is created.
On line 24 ~32MB is used after the 2nd image is created.
On line 30 the two images are added together and the result is returned from the function. At this point ~48MB is used because all 3 image still exist in memory.
On line 31 we exit from the AddImages() function and can see that the memory for two temporary images has been correctly deallocated, since the im1 and im2 pointers have gone out of scope.
Finally on line 38 we exit the main() function and see that the memory usage returns to ~0MB, since the image variable also goes out of scope and the memory associated with it is deallocated.

Let me know if you see any mistakes in my tests!
Kiran

<ITKPointerTest.png>

On Tue, 21 Mar 2017 at 12:38 Francois Budin <francois.budin at kitware.com<mailto:francois.budin at kitware.com>> wrote:
Hello Yann,

Indeed, you are correct. New() will return a smart pointer, but other methods will return a pointer, so you cannot use the keyword "auto" with these other methods.
The two following links give some insight in why ITK is implemented this way:
https://itk.org/Wiki/ITK/Examples/Utilities/ReturnObjectFromFunction
https://itk.org/pipermail/insight-developers/2011-April/018011.html

Hope this helps,
Francois

On Tue, Mar 21, 2017 at 6:45 AM, asertyuio via Insight-users <insight-users at itk.org<mailto:insight-users at itk.org>> wrote:

Hi Hans,

Thanks for your answer !

I was asking that because when I use auto keyword with a filter output, it seems like is result in a normal point type, not a smart pointer, so a code with

auto testOutput = testFilter->GetOutput();

has not the same behavior as

 FilterOutputType::Pointer testOutput = testFilter->GetOutput();

testOutput is a smart pointer in the second case, and a normal pointer in the first one.

I was wondering if it could interfere with the smart pointer role (for example causing memory leak, or wrong reference count).

After a deeper look, New() return a smart pointer (so with auto =, no problem) and GetOutput() a pointer. So the second code snipset just creates another instance of a smart pointer.

I hope what was my question is now clearer !
Yann

Le 21/03/2017 à 01:43, Johnson, Hans J a écrit :

Yann,

It will work only if you are compiling with C++11, and you have no intention of sharing your code with others who may not use C++11 syntax.

ITK proper currently maintains backwards compatibility with older versions of C++.

Hans

_____________________________________
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.php

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://public.kitware.com/mailman/listinfo/insight-users

_____________________________________
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.php

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://public.kitware.com/mailman/listinfo/insight-users
_____________________________________
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.php

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://public.kitware.com/mailman/listinfo/insight-users
_______________________________________________
Community mailing list
Community at itk.org
http://public.kitware.com/mailman/listinfo/community

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/insight-users/attachments/20170321/0ea60c0c/attachment.html>