Wessi - 6 months ago 23

SQL Question

I have a table containing the lat/lon of cafés. I would like to make an SQL query that gives me the closest café to each café. Can someone help with advice on how to do this?

The table basically looks like this:

Cafe ID

`+-------------------------+----------------------+----------------------+`

| cafe_id | gps_latitude | gps_longitude |

+-------------------------+----------------------+----------------------+

| 011-1003 | 55.86649500000000000 | 8.16856200000000000 |

| 192-143 | 57.04419159749860000 | 10.36447024359820000 |

| 037-0233 | 55.08773849210000000 | 8.56101036070000000 |

| 121-934 | 56.89120900000000000 | 9.16818100000000000 |

+-------------------------+----------------------+----------------------+

Any help is much appreciated!

Answer

You could use the Spherical Law of Cosines to get the distance expressed in earth-radii.

Some prefer the Haversine formula because it gives better precision, but given that MySql's floating point precision is high enough, the difference between the two is negligible. The first one is simpler to implement:

```
select c1.cafe_id,
substring_index (
group_concat( C2.cafe_id order by
acos( sin(radians(c1.gps_latitude)) * sin(radians(c2.gps_latitude))
+ cos(radians(c1.gps_latitude)) * cos(radians(c2.gps_latitude))
* cos(radians(c2.gps_longitude-c1.gps_longitude)) ) ),
',', 1) nearest
from cafe c1
inner join cafe c2 on c1.cafe_id <> c2.cafe_id
group by c1.cafe_id
```

Output for the sample data is:

```
| cafe_id | nearest |
|----------|----------|
| 011-1003 | 037-0233 |
| 037-0233 | 011-1003 |
| 121-934 | 192-143 |
| 192-143 | 121-934 |
```

Here is a MySql fiddle.

The distance calculation is used in the `order by`

clause of the `group_concat`

aggregation function, which results in a comma-separated list of *cafe_id* values in order of their distance to the cafe you group by. The `substring_index`

function extracts the first item from that list.

The join condition is important as without it you'd get the cafe itself as the closest neighbor (its distance is obviously 0 then).