zip code calculations

fnbcprog

Can someone point me in right direction for getting the all zip codes within "X" miles of another zip code?

jdorsch

First you need a database of zip codes that includes longitude and latitude, you should be able to find one for about $5 online and import it into your backend. Here is a function that will compute the distance, as the bird flies, between two longitudes and latitudes in miles.

	private static function calcDist($lon1,$lat1,$lon2,$lat2)
	{
		$r = 3963.0;
		$multiplier = 1;
		return $multiplier * $r * acos(sin($lat1/57.2958) *
 			sin($lat2/57.2958) +  cos($lat1/57.2958) *
 			cos($lat2/57.2958) * cos($lon2/57.2958 - $lon1/57.2958));
	}

So lets say you have 200 store locations, and you want to find the 5 closest to a given zip code. Select the longitude/latitude from your zip db for each of your store locations and put them in an array, then find the lat/longitude for the zip you are searching on.

Run each of the store locations through the function above with the search lat/lon, then store the results in another array with your storeID or whatever. Then reorder the array by the distance and strip out the elements that are too far or get the 5 closest or whatever you need.

etully

That method is fine for very small datasets. We found it inefficient when we had 2000+ stores and lots of people searching the database about the same time (more than one per second).

We did some preprocessing and built a sort of index table. For example, let's say that we have zip code 12345 and we already know that there are no stores within 500 miles of that zip code. We make an entry in the table for 12345, 0. Now we don't need to calculate the distance from 12345 -> 200 zip codes.

If the zip code has 1 store within 50 miles and two more that are 51-100 miles away, then we build the table like this:

12345, '55555', '66666 77777'

Now, when someone searches on 12345, we instantly know all the stores that are within 50 miles (55555) and all the stores that are within 100 miles (66666 + 77777). If we want to print actual distances, we can lookup to latlong and calculate as you said but we only have to calculate 3 distances, not 2000+. This technique might be overkill for smaller sites but it is crucial to any site that gets a decent amount of traffic.

jdorsch

Taking that a step further etully, couldn't you do this?

If you know that you are searching for all stores with 100 miles of a zip code, given that 1 degree of latitude is about 69 miles, you can limit your search on a new non-cached zip to only zip codes with a latitude that is plus or minus 1.45 degrees of the search zip latitude. You cannot do this with longitude since the distance narrows the further from the equator, but I assume this would eliminate maybe 80% of the stores you had to calculate distance for, of course this wouldn't help as much with a 500 mile radius, but the shorter the distance, the more efficient this becomes.