re: Dealer/zip locator constipated

This is a computationally intense problem if you approach it using a least-squares-mean approach to find the distance between two arbitrary 2-D points on a sphere as large as the Earth. Several years ago, I wrote some code for a client that computed shortest-distance routes between FBOs for general aviation (small aircraft) travelling cross-country. Using this approach took tens-of-seconds per pair of locations, and that was in C code running on 486-class machines. The problem is that you've got an x-y grid and you have no localization cues. If you used a sparse matrix and "spaced" things out in ten-second (of arc) increments, you might be able to narrow down your search, but your database would become terribly bloated. You MUST put some pre-filtering criteria in your database to localize your searches in order to reduce the search time in a 2-D space. Fundamentally, it's the wrong approach to this kind of a problem. As that project was completing, another client came to me who wanted a moving-map display that updates itself in real-time. He proposed the least-squares-mean approach on a database with 10**3 more data points and a requirement that the granularity vary from city blocks to hundreds of miles. A couple of test cases showed this approach wasn't going to work at all. About that time, I found this article article in DDJ's 12/92 issue: SPATIAL DATA AND THE VORONOI TESSELLATION by Hrvoje Lukatela and John Russell http://www.ddj.com/articles/1992/9212/9212toc.htm I just did a search on Yahoo! for "voronoi" and came up with lots of hits. The the guys who wrote the article and have the libraries available are still around at: http://www.geodyssey.com/ They put out C libraries. I worked with their code on the latter project, and it's a TERRIFIC model to use! I haven't looked around, but I suspect you could get you hands on some code that would be helpful. The DDJ article had simplified presentations of the Voronoi algorithms in C that I played with and you wouldn't believe how much FASTER this approach is than the approach you're trying! Back then, Lukatela and Russell were just about the only folks who were doing anything with Voronoi nets, but judging from the results of the Yahoo! search, it's really blossomed. The code in the original DDJ article would probably be sufficient for your needs -- you'd get some distance variations of a few miles for coast-to-coast distances, which might be tolerable. Depending on how much detail you want to provide to your voronoi nets, you could reduce the errors by several orders of magnitude. Good luck! -David Schwartz