### Author:  [@lorensen](https://discourse.vtk.org/u/lorensen)
#### Post date: [April 2, 2019, 8:43pm UTC](https://discourse.vtk.org/t/should-we-replace-safedowncast-with-dynamic-cast/669/1 "2019-04-02T20:43:11Z")
Folks,
_SafeDownCast_ was added to vtk way back when _dynamic\_cast_ was not available everywhere. Should we replace _SafeDownCast_ in vtk with _dynamic\_cast_? It is more natural for modern C++ and is well-documented. It may be more efficient although I doubt the current implementation is not a significant resource consumer.
See:
```auto
vtkKochanekSpline::SafeDownCast(this->ParametricSpline->GetXSpline())->SetDefaultTension(value);
```
versus
```auto
dynamic_cast(this->ParametricSpline->GetXSpline())->SetDefaultTension(value);
```
It is used in over 1500 vtk files. but the changes could be automated.
Bill
---
### Author:  [@lorensen](https://discourse.vtk.org/u/lorensen)
#### Post date: [April 3, 2019, 6:16pm UTC](https://discourse.vtk.org/t/should-we-replace-safedowncast-with-dynamic-cast/669/2 "2019-04-03T18:16:32Z")
I did an interesting experiment. I pushed an MR that replaced the implementation of SafeDownCast to use dynamic\_cast. The first problem was using vtkAbstractHyperTreeGridMapper in vtkCellPicker. The hyper class has no subclasses in VTK. Not sure why.
I’m also getting some failed tests. I’ll continue to investigate…
---
### Author:  [@ken-martin](https://discourse.vtk.org/u/ken-martin)
#### Post date: [April 3, 2019, 7:38pm UTC](https://discourse.vtk.org/t/should-we-replace-safedowncast-with-dynamic-cast/669/3 "2019-04-03T19:38:53Z")
I like this idea Bill, seems like if we can do it, we should.
---
### Author:  [@toddy](https://discourse.vtk.org/u/toddy)
#### Post date: [April 4, 2019, 1:50am UTC](https://discourse.vtk.org/t/should-we-replace-safedowncast-with-dynamic-cast/669/4 "2019-04-04T01:50:06Z")
`SafeDownCast` is redundant in wrapper code, so no issues there.
---
### Author:  [@finetjul](https://discourse.vtk.org/u/finetjul)
#### Post date: [April 9, 2019, 4:28pm UTC](https://discourse.vtk.org/t/should-we-replace-safedowncast-with-dynamic-cast/669/5 "2019-04-09T16:28:57Z")
> [@lorensen](#):
>
> It may be more efficient although I doubt the current implementation is not a significant resource consumer
I was under the impression that there are different options to improve performances of dynamic\_cast. (e.g. [http://www.stroustrup.com/fast\_dynamic\_casting.pdf](http://www.stroustrup.com/fast_dynamic_casting.pdf)). By keeping SafeDownCast, we let ourselves the option to speedup downcasting some day…
But maybe things have changed with the newests C++, I am not really up to speed here.
---
### Author:  [@lorensen](https://discourse.vtk.org/u/lorensen)
#### Post date: [April 16, 2019, 3:25am UTC](https://discourse.vtk.org/t/should-we-replace-safedowncast-with-dynamic-cast/669/6 "2019-04-16T03:25:19Z")
@jjomier I pushed an MR that replaced SafeDownCast with dynamic\_cast. As you suspected, there are performance issues. We are better off keeping DafeDownCast. I’m afraid I wasted some of my not-so-costly time. I am closing the MR.
See these results: [https://tinodidriksen.com/2010/04/cpp-dynamic-cast-performance/](https://tinodidriksen.com/2010/04/cpp-dynamic-cast-performance/)
Bill
---
### Author:  [@amaclean](https://discourse.vtk.org/u/amaclean)
#### Post date: [April 16, 2019, 5:53am UTC](https://discourse.vtk.org/t/should-we-replace-safedowncast-with-dynamic-cast/669/7 "2019-04-16T05:53:02Z")
Thanks for experimenting @lorensen we now know there are performance issues!
---
### Author:  [@lorensen](https://discourse.vtk.org/u/lorensen)
#### Post date: [April 18, 2019, 3:56pm UTC](https://discourse.vtk.org/t/should-we-replace-safedowncast-with-dynamic-cast/669/8 "2019-04-18T15:56:10Z")
Here’s a link to a dashboard that illustrates performance and failure related to my [dynamic\_cast experiment.](https://open.cdash.org/index.php?compare1=61&filtercount=1&field1=revision&project=VTK&showfilters=0&limit=100&value1=83dc26ad1b689795675f574272a46c755d843e06&showfeed=0)
---
### Author:  [@toddy](https://discourse.vtk.org/u/toddy)
#### Post date: [April 19, 2019, 12:41am UTC](https://discourse.vtk.org/t/should-we-replace-safedowncast-with-dynamic-cast/669/9 "2019-04-19T00:41:45Z")
Did you identify the cause of failure with the dynamic cast on some test machines? It seems odd.
---
### Author:  [@lorensen](https://discourse.vtk.org/u/lorensen)
#### Post date: [April 19, 2019, 1:44am UTC](https://discourse.vtk.org/t/should-we-replace-safedowncast-with-dynamic-cast/669/10 "2019-04-19T01:44:59Z")
There were lots of discussion about this at the hackathon. Way above my level of c++ knowledge.