Organization Problem

Question

I'm writing a portion of a program which is meant to do pathfinding using a series of connected convex polygons to denote the area on which the agent (the person trying to find a path) can walk. A* and some basic vector math make most of it pretty simple. However, I've rewritten the code a couple times because I can't decide on the best way to represent the data structures involved. Finally (and most difficult) I need to be able to save, load, and edit the map.
I basically just need a main class (call it Map) which contains Areas which are just convex polygons with geometry data and a set containing all polygons it is adjacent to. The points are simply double2 or double3 containing 2- or 3- dimensional coordinates.
My first idea was to simply have the Area store the points it's made of and adjacent areas. However, then I have to write a messy function to work backwards and figure out which edges the areas have in common. So then I associate a list of edges with the list of adjacent areas to store that information. But now I have the problem when I save, do I store two+ copies of each point (one associated with each edge it is an endpoint of and one associated with each polygon it is a part of) and multiple copies of edges that are part of multiple polygons? This requires messy functions which compare points (and edges) to determine if they are equal. So I thought about indexing the points, then having the edges simply storing two int's and the areas storing int's to reference the edges. But then I have a hard time drawing the polygons because the edges have to be in a specific order and the points in the edge have to be in the right order so I also need the areas to have int's to reference the points. But now with everything indexed, any insertions or deletions I make require me to go through and find all the dependencies and correct them so that the indices are the same.
Can anyone think of a way to represent this hierarchy that would avoid most of these issues and allow me to cleanly deal with the remaining problems?

Answer 1

I can use the Polygon class to represent a convex polygon, Line2D.Double to represent an edge, Point2D.Double etc, but this only creates more problems - I need to extend these classes to implement all the functionality I need, Polygon coordinates must be int instead of double, and I can no longer implement 3D coordinates easily. This also solves almost none of the issues I mentioned above.
EDIT: I think I might just have to index everything and simply tread carefully when I write the editing code.
Edited by: blevinstein on Aug 15, 2008 12:45 PM