views:

2158

answers:

7

We have a restaurant table that has lat-long data for each row.

We need to write a query that performs a search to find all restaurants within the provided radius e.g. 1 mile, 5 miles etc.

We have the following query for this purpose:

***Parameters***

Longitude: -74.008680
Latitude: 40.711676
Radius: 1 mile

***Query***

SELECT *
FROM restaurant
WHERE (
POW( ( 69.1 * ( Longitude - -74.008680 ) * cos( 40.711676 / 57.3 ) ) , 2 ) + POW( ( 69.1 * ( Latitude - 40.711676 ) ) , 2 )
) < ( 1 *1 );

The table has about 23k rows. The size of the result set is weird at times e.g. for a 5.4 mile search, it gives back 880 rows and for 5.5 miles, it gives back 21k rows.

This table contains restaurant data for nyc - so the real distribution is not as per the result set.

Question: IS THERE ANYTHING WRONG With this query?

DB: MQSQL, Longitude: DECIMAL(10,6), Latitude: DECIMAL(10,6)

+1  A: 

Check out the answer to this question. Similar problem.

Ramin
A: 

Hi James McNellis,

I am having the same problem. I have tried Pythagorean theorem also the Great circle distance calculation. But i am not getting the exact result. Can any one please enter those formula here?

Thanks.

A: 

You may want to create a SPATIAL index on your table to make the searches faster.

To do this, add a POINT column to your table:

ALTER TABLE restaurant ADD coords POINT NOT NULL;

CREATE SPATIAL INDEX sx_restaurant_coords ON restaurant (coords);

SELECT  *
FROM    restaurant
WHERE   MBRContains(coords, LineString(Point(583734 - 1609, 4507223 - 1609), Point(583734 + 1609, 4507223 + 1609))
        AND GLength(LineString(Point(583734, 4507223), coords)) <= 1609

You should store coords as UTM coordinates within a single zone.

Quassnoi
A: 

IS THERE ANYTHING WRONG With this query?

In my opinion the WHERE clause is going to be slow because of the maths involved, and the use of functions in the WHERE clause will prevent the database using an index to speed the query - so, in effect, you will examine every restaurant in the database, and perform the great-circle maths on every row, every time you make a query.

Personally I would calculate the TopLeft and BottomRight co-ordinates of a square (which only needs to be crudly calculated using pythagoras) with sides equal to the range you are looking for, and then perform the more complicated WHERE clause test on the smaller subset of records that are within that Lat/Long square.

With an Index on Lat & Long in the database the query

WHERE     MyLat >= @MinLat AND MyLat <= @MaxLat
      AND MyLong >= @MinLong AND MyLong <= @MaxLong

should be very efficient

(Please note that I have no knowledge of MySQL specifically, only of MS SQL)

Kristen
A: 

If your data is in SQL server database, you can use this:

CREATE PROC up_FindZipCodesWithinRadius

@ZipCode char(5) ,
@GivenMileRadius int

AS SET NOCOUNT ON

DECLARE @lat1 float, @long1 float

SELECT @lat1= latitude, @long1 = longitude FROM ZipSource WHERE zipcode = @ZipCode

SELECT ZipCode ,DistanceInMiles FROM ( SELECT ZipCode,3958.75 * ( Atan(Sqrt(1 - power(((Sin(@Lat1/57.2958) * Sin(latitude/57.2958)) + (Cos(@Lat1/57.2958) * Cos(latitude/57.2958) * Cos((longitude/57.2958) - (@Long1/57.2958)))), 2)) / ((Sin(@Lat1/57.2958) * Sin(latitude/57.2958)) + (Cos(@Lat1/57.2958) * Cos(latitude/57.2958) * Cos((longitude/57.2958) - (@Long1/57.2958)))))) as DistanceInMiles FROM ZipSource ) a WHERE a.DistanceInMiles <= @GivenMileRadius --AND ZipCode <> @ZipCode ORDER BY DistanceInMiles

GO

EXEC up_FindZipCodesWithinRadius '35085',20 GO

DROP PROC up_FindZipCodesWithinRadius

Dinci Garrone
A: 

Use a function, e.g. the one I posted here.

Then, query your restaurants, e.g. to get everything within a 5-mile radius

select * from restaurants 
  where dbo.udf_Haversine(latitude, longitude, @lat, @long) < 5

This performs fine with ZIP code data.

cdonner
A: 

You can check out this page http://www.zipcodeworld.com/developers.htm for the formula to calculate the distance for the radius.