[Insight-developers] compiling tests, include problems

Damion Shelton dmsst59+@pitt.edu
Wed, 25 Jul 2001 11:45:10 -0400


Ok... I'll move the files. A few comments:

> Afterall, when you put in your "true" test of the BloxImage, it is just
testing your Blox code? Or is
> it also testing readers/writers, the gaussian filter, etc.

The latter; the Blox algorithm is reasonably compartmentalized, and it's
hard to just start in the middle of things. For instance, to test the
itkBloxCoreAtomAnalyzer class, you need an itkCoreAtomImage.
itkCoreAtomImage needs an itkBloxBoundaryPointImage as input, which in turn
needs an image of CovariantVectors representing gradient as input.

> Now you could construct a Blox test that was real that didn't use some of
these other classes.  For
> instance, if need gaussian blurred data, you could use an input dataset
that was already blurred.
> Thus, your test wouldn't test the blurring code and would focus on testing
your algorithm.

That's a good idea (reading data from a file). If I started with a gradient
image from disk I could test just the blox portion of the algorithm. Right
now I have to generate that data manually.

Are any of the file reader formats planned for the beta capable of storing
non-scalar (i.e. vector/gradient) formats?

-Damion-