[Paraview] camera values range 2

[Davide Detomi]

Moreland, Kenneth ha scritto:
>> Hi, I'm a novice of paraview so I beg your pardon if I submit an already
>> discussed problem. I had many problems playing with the camera because
>> it seems the min-max range of the possible values spans the interval
>> [-10000 ,+10000]. This limits sound strange to me, I wondered why it
>> hasn't been set a much higher value. For example, in my case, I had a
>> model in mm, thus it was easy to reach the upper and lower bound,
>> causing a crazy behavior of the final view. Two questions about this:
>>
>> 1) Has anybody already faced such a problem?
>>     
>
> There are no set limits, but multiple readers have run into problems with camera movement when the geometry bounds of the data are very large or very small (see bug #5576 http://www.paraview.org/Bug/view.php?id=5576).  I'm pretty sure they are due to some rounding issues with a 32-bit float somewhere.  However, they usually don't occur unless the bounds exceed +/-1e13.  I've never seen it on bounds as small as yours.  Can you describe your data further?
>
>   

Hi Kenneth, thanks for your reply. What I saw is that the min-max values 
of the camera spinbuttons are, as I said, [-10000,+10000]. Even trying 
to modify them by hands by clicking on the spinbutton arrows I couldn't 
exceed those values. Then I guessed they were strong bounds of the 
camera settings. My version is the 3.3.0, thus the last one, I guess. 
I've found a workaround by rescaling my mesh by a factor 1000, by doing 
this the camera works fine. However, this is clearly a temporary 
solution, just a trick to fix the problem. Can you change the camera 
bounds beyond those limits? I think my problem is not due to any 
rounding issue, but just to (strange) bounds on the min-max allowed 
camera values. And if the mesh is in mm is not that unusual to reach the 
bounds while playing with the camera.


>> 2) Is it possible to somehow alter these bounds?
>>     
>
> Try using the Transform filter and apply a scale to make the bounds smaller.
>
>   

scaling by means of the transform filter works fine. It is the same I 
did before to fix the problem, even if at that time I did it outside 
Paraview within the mesh generator.

>> Another non-related issue:
>>
>> 3) If I load a huge mesh (around 8 millions of elements) Paraview
>> suddenly crashes because a failed memory allocation. Would it be
>> possible in the future to have a smoother failure, maybe through a popup
>> warning that the memory limits have been reached? This would be
>> extremely helpful both for a cleaner behavior and to avoid a waste of
>> time because of a not yet saved state.
>>     
>
> I think everyone agrees that would be ideal, but unfortunately requires a huge level of effort.  Also, any solution would be difficult to test and maintain.
>
> -Ken
>
>
>
>   

I understand that a clean, fully satisfactory solution it's not 
straightforward but all the crashes I got have been due to failed 
mallocs, then a first step for future releases might be a simple check 
on the allocated memory, that now I guess is sometimes missing, 
eventually popping up a warning when the allocation fails. I can imagine 
that there are many calls to mallocs all over the code, thus this may 
require some time. However, I think this effort is worthwhile and, at 
this first level, I don't think it would be too complicate to test and 
maintain.

Anyway, thanks for all your hints

Davide


-- 
--------------------------------------------------------------

Davide Detomi, Post-Doc Assistant

Institute of Mathematics
Chair of Modelling and Scientific Computing
Ecole Polytechnique Fédérale de Lausanne
Address: EPFL, IACS-CMCS, CH-1015 Lausanne (Switzerland)

Tel. Work: +41 21 69 32909
Fax:       +41 21 69 34303
E-mail:    davide.detomi at epfl.ch
Office:    MA C2 567
--------------------------------------------------------------