How to optimize SQL query with calculating distance by longitude and latitude?

Question

I have a table with structure like that:

table name: shop

id_shop      int(10)
name         varchar(200)
latitude     double
longitude    double


        

Answer 1

Here's a few ideas, some of which may not apply depending on your exact situation.

1. You could do the conversion of latitude and longitude to radians and store that in the row as well. This would save the cost of those calculations (actually the cost would be accrued once when storing the data).
2. If your table is very large, you could use a simple linear distance calculation rather than the Haversince formula to limit the results to which you apply the Haversince formula.
3. If you have other data in the table that would serve as a good first filter (country/region/etc.), you could apply it first.
4. You could reorder your joins so that they are applied after the distance filter so that you don't incur the cost of the join on data that doesn't qualify.

Answer 2

Well, for starters, you could store pre-calculated stuff in the database when you store the latitude and longitude. If you pre-store the latitude and longitude as radians, for example, you'll only need to calculate RADIANS(latitude) and RADIANS(longitude) once, when you store each location, not every time you need to do your distance calculation (presumably far more than once.)

Maybe shave off a bit more by storing the SIN(RADIANS(latitude)) and COS(RADIANS(latitude)) when you first populate the row, too...

I'm guessing you're doing many, many "nearest thing to X" calculations over time -- that's what people are generally doing when faced with this calculation -- and pre-calculating what you can is normally the first thing to try.