Matching closest longitude/latitude

Question

Using Maxmind's GeoIP software, we can narrow down the LONG/LAT of an IP address to relative accuracy within 25 miles around 80% of the time.

Now, we don't want to use any of the other information provided by MaxMind, because there are a lot of discrepancies between feature names, i.e., cities, to perform a look up. We plan on attempting such a look up if other methods fail to locate a feature, but for performance reasons, look ups on floats are much faster than strings.

Now, I'm a little clueless on how we can find the closest matching LAT/LONG given from Maxmind to our database. The problem is, our datbase features has a much higher precision compared to that of Maxmind, therefore a straight comparison might not be effective. If we try applying a ROUND() to the column during query, that will obviously be really slow.

Given the following data, would the fastest way simply be something like

LONG 79.93213 LAT 39.13111

SELECT `feature_name` FROM `geo_features`
WHERE long BETWEEN 79.93 AND 79.79.94
AND lat BETWEEN 39.13 AND 39.14

Can anyone thing of an elegant solution that will be blazing fast? I know there are some new spatial storage types in MySQL 5, perhaps anyone can provide a solution beyond the blinders I've seem to put up on myself.

Answer 1

The elegant (more accurate) way of doing this (but not blazing fast)

// Closest within radius of 25 Miles
// 37, -122 are your current coordinates
// To search by kilometers instead of miles, replace 3959 with 6371
SELECT feature_name, 
 ( 3959 * acos( cos( radians(37) ) * cos( radians( lat ) ) 
  * cos( radians( long ) - radians(-122) ) + sin( radians(37) ) 
  * sin( radians( lat ) ) ) ) AS distance 
FROM geo_features HAVING distance < 25 
ORDER BY distance LIMIT 1;

Edit

This is Haversine formula for calculating circular distance from geo-coordinates. Here are some implementation of this formula in different platforms

R = earth’s radius (mean radius = 6,371km)
Δlat = lat2− lat1
Δlong = long2− long1
a = sin²(Δlat/2) + cos(lat1).cos(lat2).sin²(Δlong/2)
c = 2.atan2(√a, √(1−a))
d = R.c
// Note that angles need to be in radians to pass to Trigonometric functions