Calculating 'nearest' 10 locations to a postcode

Hi Dan,

Thanks for your comment.

I think i perhaps wasnt clear enough. When the system is in use, I would like to be able to calculate the closest 10 locations to any defined location. Therefore given the location could change, it would need calculating at runtime.

Regards
Dean

Thanks to everyone for their comments, apologies for the massive delay responding. I have been meaning too, but wanted to run some test before commenting, and never found the time...

I suspect one of the reasons for the poor performance is that the Db is stored on a shared host, so im competing for processing power. I have ran some tests and it looks like the main reason for the delay is simply querying the large postcode table (1.7 million records).

I performed the haversine calc on table of 50,100,150 and 200 records only and it was pretty much instant.

SELECT post4.latitude, post4.longitude, 111.045*haversine(latitude,longitude,latpoint, longpoint) AS distance_in_km FROM post4 INNER JOIN ( SELECT 57.144165160000000 AS latpoint, -2.114847768000000 longpoint ) AS p ORDER BY distance_in_km

Select all Open in new window

Yet, simply joining my place table to the postcode table to get the lat/lng info took 6 -7 secs...
 

SELECT * FROM type_place AS p JOIN postcodelatlng AS pc ON p.postcode=pc.postcode

Select all Open in new window

That increased to 14 secs when i added a WHERE clause

SELECT * FROM type_place AS p JOIN postcodelatlng AS pc ON p.postcode=pc.postcode WHERE pc.latitude>54 AND pc.latitude<56 AND pc.longitude<-1.5 AND pc.longitude>-1.7

Select all Open in new window

Finally joining all records and performing the haversine calc took 180secs.....

SELECT 111.045*haversine(latitude,longitude,latpoint, longpoint) AS distance_in_km FROM type_place AS p JOIN postcodelatlng AS pc ON p.postcode=pc.postcode JOIN ( SELECT 57.144165160000000 AS latpoint, -2.114847768000000 longpoint ) AS c ORDER BY distance_in_km

Select all Open in new window

I realised that i dont actually need to access the large table, as i only need to know what places are near other places in my smaller table of 1000 rows. Therefore i plan to just add the lat/lng information to the place table and perform the select and calc on that instead.

Thanks again
Dean

Hi Ray,

Yes i had read you article, but to be honest i kind of discounted the temporary table to begin with, because i thought if the initial JOIN was taking 6+ seconds then it would surely have similar results... lazy i know.

I have been trying this morning to see if that would have indeed been the case, but the script im using doesnt seem to like creating the temporary table with the JOIN.

Im using the PHP below:

<?php
ini_set("display_errors",1<wbr ></wbr>);
error_reporting(E_ALL);

// A SCRIPT TIMER
$alpha_time = microtime(TRUE);

/* Open a connection */
$mysqli = new mysqli("localhost", "xxx", "xxx", "xxx");

if (mysqli_connect_errno()) {
    printf("Connect failed: %s\n", mysqli_connect_error());
    exit();
}

$mysqli->query("
CREATE TEMPORARY TABLE IF NOT EXISTS 
  temp_table ( INDEX(place_id) ) 
ENGINE=MEMORY
AS (
SELECT * FROM type_place 
)
");

if ($result = $mysqli->query("SELECT * FROM temp_table")) {
    printf("Select returned %d rows.\n", $result->num_rows);
    /* free result set */
    $result->close();
}

$mysqli->close();

    // A SCRIPT TIMER
	$omega_time = microtime(TRUE);
    $lapse_time = $omega_time - $alpha_time;
    $lapse_msec = $lapse_time * 1000.0;
    $lapse_echo = number_format($lapse_msec,<wbr ></wbr> 1);
    echo "<br/>SCRIPT TIME: $lapse_echo MILLISECONDS <br/>";
?>

Select all Open in new window

The script works fine when simply selecting all from the type_place table(returns 750 rows). However, when i swap that for the following sql:

SELECT p.place_title, pc.latitude, pc.longitude FROM type_place AS p JOIN postcodelatlng AS pc ON p.postcode=pc.postcode WHERE pc.latitude>55 AND pc.latitude<58 AND pc.longitude<-1.5 AND pc.longitude>-2.2

Select all Open in new window

It doesnt return a result, even though the same sql returns 26 records when i run it directly in phpmyadmin.

Finally, not sure if it works in the same way as when called from script, but when i run the create temporary table script in phpmyadmin (making the join to the large table) it also takes 18secs:

CREATE TEMPORARY TABLE nearby ( distance DECIMAL(6,1) ) ENGINE=MEMORY SELECT p.place_title, pc.latitude, pc.longitude FROM type_place AS p JOIN postcodelatlng AS pc ON p.postcode=pc.postcode WHERE pc.latitude>55 AND pc.latitude<58 AND pc.longitude<-1.5 AND pc.longitude>-2.2

Select all Open in new window

Regards
Dean