g_relation?

[Brett Viren <bviren superk physics sunysb edu>]

Hello.

I have a GLIB question (I think).

I have 2 types of data which have links or connections between them.
I have at most 10,000 type `A' data elements and at most 100,000 type
`B' elements.  However, typically there are only some 1000s of `A's
and some 1000s of `B's.

What I need is some way to create a list of `A's, each of which have a
list of `B's associated and also create a list of `B's each of which
has a list of `A's associated (that is the first list of lists turned
inside out).

This can just be done with a lot of list manipulation, but it may not
be done in the most optimized or simple to use way.  I think what I
want is called a `map' and I think that this is what GLIB's
g_relation_* routines are for.  Is this so?  Would g_relation be
suitable for 1000s or more relationships?

If so, I need some docs on how to learn about using the g_relation_*
routines.  Can anyone help me out here?  So far all I can find is the
test code in testglib.c and that isn't enough for me, unfortunately.


-Brett.

PS: Specifically what I am trying to do is a circular Hough transform
from a real space set of at most 10,000 points on a sphere to a Hough
space of about 10,000 points (also on a sphere) using about 10
different assumed Hough angles (that is, Hough radii if this were done
on a flat space).  It is desireable to know not only which points in
real space contributed to each point in Hough space, but also vice
versa.