[vtk-developers] New directory structured for VTK?

Michael Halle halazar at media.mit.edu
Thu Apr 5 10:37:53 EDT 2001


Ken,

I think there's good logic in your reconstruction plan.  Here are a few
issues that might be worth thinking about:

1. I already sometimes have problems guessing where a class is when I go
to look for it. (common?  graphics?  contrib?).  As the tree is
subdivided more, it would be nice to have some sort of quick index of
what's where in a single build (at least in the documentation, if not
also in a text file that gets built up from directory contents).

2. A tree restructuring might also be a good time to make is possible for
users to more easily add their own modules (not just classes) to local
and maybe contrib.  If the tree allowed several classes to be grouped
together by application domain in addition to structure/dependencies,
you might not have to restructure again as vtk grows.  

For an arbitrary example, Bill's subdivision surface stuff might have
been in its own subdirectory, and compiled to its own library.  These
subdirectories provide the breadth while your tree provides the depth.
Allowing subdirectories of code also allows a break from effective
"one file == one class" limitation of the current build, which in some
cases leads to very large class files with low-level implementation
code.

							--Mike






More information about the vtk-developers mailing list