[Insight-developers] More thoughts on Interface directory...

[Lorensen, William E (CRD)]

I think the regression tests should not depend on any non-itk packages if at all possible. I'm not
sure how to do this yet.

I do believe the classes should have as a start hand crafted test programs that can be easily
validated.
In vtk, we had the problem that many of the tests produced interesting images, but they could not be
validated.

The next level should include more sophisticated tests(ala the vtk examples). Here we may need to
read and write non-image data. Not sure what the solution is right now,

Bill

--Original Message-----
From: Will Schroeder [mailto:will.schroeder at kitware.com]
Sent: Thursday, August 10, 2000 11:59 AM
To: Insight-Developers
Subject: [Insight-developers] More thoughts on Interface directory...


Just to add to what Vikram has started: my thinking is that the interfaces directory
would include interfaces to data files and such. As Vikram has pointed out, we are
not mandated to create a family of readers/writers or interfaces. But how do you
test something without data? The mesh class could be tested by hand crafting examples
with 5-10 cells, but the tests are a lot more convincing if we can throw in several
thousand triangles, quads, etc. which we have in abundance (vtk data files or
iso-surface generation). So we are testing the class by generating data with vtk and
then putting the data into itk. Then, we are visualizing the results by converting itk
into vtk and displaying it. And since this process has taken some effort, we don't want
to waste any work and are trying to collect it up for others to use.

Sorry for the soapbox,
Will


_______________________________________________
Insight-developers mailing list
Insight-developers at public.kitware.com
http://public.kitware.com/mailman/listinfo/insight-developers