[vtk-developers] NaN in VTK

[Philippe Pébay]

Hello,

I am wondering what is the status of things regarding support of NaN in VTK. I see it implemented, albeit with caveats 
in the documentation, in a few classes.

Just to give some context: I would like to improve the correlative statistics by generalizing to the case where the
discriminant of the covariance matrix (a.k.a. the "generalized variance") is not positive: instead of treating this 
degenerate case as such, and as a result having to resort to invalid values when assessing data, I would like to simply 
use NaNs which would thus automatically to NaN assessment values.

There would be additional benefits, too, such as the fact that one linear regression might be valid when the other is 
not, when one variable has non-zero variance while the other has not.

So, if anyone could summarize what to expect, or point me to a document discussing this issue of NaNs in VTK, it would 
be most useful.

Thanks
Philippe


-- 
Philippe Pébay
Sandia National Laboratories