Date: Fri, 11 Dec 1998 11:44:44 -0500 (EST)
From: David Nash <dnash@umich.edu>
X-Sender: dnash@pacman.rs.itd.umich.edu
To: beale@BEST.COM
Subject: Problem with orb (spherical heightfield generator)

I think I've found a bug in the Unix (C source only) version of orb.  I've
been using it to make planets in POV-Ray, using actual altitude data from
space missions.  No problems compiling it, but I found that what I got out
didn't resemble the image map used as input.  The x-dimension (horizontal)
of the image map maps correctly, but only the central part of the
y-dimension (vertical) gets mapped.

I've found what seems to be the problem:  gsphere.c has the lines:


    r2d = sqrt(r.x * r.x + r.y * r.y); /* radius in the x-y plane */
    phi = atan2(r.z,r2d);  /* range of atan2 is [-pi..pi] */
    theta = atan2(r.y, r.x);
   
    xmap = (int)((0.5 + pi2inv * theta) * (double)xsize);
    ymap = (int)((0.5 + pi2inv * phi) * (double)ysize);

In gsphere.c phi will never range over more than two quadrants, since r2d is
always positive.  I.e., its range will be -pi/2 to pi/2.  However, ymap assigns
the top of the map to phi=-pi (instead of -pi/2) and the bottom (ysize) to phi=+pi
-- meaning that only the central portion of the image map gets processed. 
Changing this line to

    ymap = (int)((0.5 + (pi2inv * phi * 2)) * (double)ysize);

solves the problem.

Sorry to bother you if you're already aware of this.

Sincerely,
David Nash
dnash@umich.edu