1. 技术文章
  2. How to get a list of suburbs surrounding a location then repeat for other locations using MySql?

How to get a list of suburbs surrounding a location then repeat for other locations using MySql?

由 邮差的信 提交于 2019-11-28 02:18:00

You simply need to perform a self-join. Joining tables is a very fundamental part of SQL—you really should read up on it before trying to understand this answer further.

SELECT   poi.asciiname,
         suburb.asciiname,
         suburb.country,
         DEGREES(
           ACOS(
             SIN(RADIANS(   poi.latitude))
           * SIN(RADIANS(suburb.latitude))
           + COS(RADIANS(   poi.latitude))
           * COS(RADIANS(suburb.latitude))
           * COS(RADIANS(poi.longitude - suburb.longitude))
           )
         ) * 60 * 1.852 AS distance
FROM     geoname AS poi
    JOIN geoname AS suburb
WHERE    poi.asciiname IN ('Tamworth', 'Birmingham', 'Roanoke')
     AND poi.population > 0
     AND poi.fcode = 'PPL'
     AND suburb.fcode IN ('PPLX', 'PPPL')
HAVING   distance <= 60
ORDER BY poi.asciiname, distance

See it on sqlfiddle.

You'll have noticed that I've used MySQL's IN() operator as a shorthand for value = A OR value = B OR ....

You'll also have noticed that I've used MySQL's DEGREES() and RADIANS() functions rather than trying to perform such conversions explicitly.

You were then multiplying minutes of latitude by a factor of 1.851999999962112, which was rather strange: it's extremely close to 1.852, which is the precise number of kilometres in a nautical mile (historically defined as a minute of latitude), but yet bizarrely slightly different—I've assumed you meant to use that instead.

Finally, you had the literal value by which you were filtering the distances in the resultset as a string, i.e. '60', whereas obviously this is a numeric value and should be unquoted.

Using Spatial Data Types.

Well first of all if you have a lot of geospatial data, you should be using mysql's geospatial extensions rather than calculations like this. You can then create spatial indexes that would speed up many queries and you don't have to write long drawn out queries like the one above.

Using a comparision with ST_Distance or creating a geometry with the radius of interest along with ST_within might give you good results and could be a lot faster than the current. However the best and fastest way to achieve this, ST_Dwithin isn't implemented yet in mysql.

These data types are available mysql 5.7 onwards but it's totally worth the effort to upgrade your DB if you are in an older version.

The new table structure.

CREATE TABLE `geoname2` (
    `geonameid` INT(11) NOT NULL,
    `asciiname` VARCHAR(200) NULL DEFAULT NULL COLLATE 'utf8_unicode_ci',
    `country` VARCHAR(2) NULL DEFAULT NULL COLLATE 'utf8_unicode_ci',
    `pt` POINT,
    `fcode` VARCHAR(10) NULL DEFAULT NULL COLLATE 'utf8_unicode_ci',
    `population` INT(11) NULL DEFAULT NULL,
    `area` INT(11) NULL DEFAULT NULL,
    PRIMARY KEY (`geonameid`),
    INDEX `asciiname` (`asciiname`),
    INDEX `country` (`country`),
    INDEX `fcode` (`fcode`),
    INDEX `population` (`population`),
    INDEX `area` (`area`),
    SPATIAL INDEX `pt` (`pt`)
)COLLATE='utf8_unicode_ci'
ENGINE=InnoDB;

Notice that the latitude and longitude fields have been replaced by pt and their indexes have been replaced by a single index.

The new query A

SELECT asciiname as suburb, 'Tamworth' as point_of_interest, country,  
  ST_DISTANCE(`pt`, POINT(@lat,@lng)) as distance 
FROM geoname2     
WHERE (fcode='PPLX' OR fcode='PPPL') AND ST_DISTANCE(`pt`, POINT(@lat,@lng))  <= 1
ORDER BY distance ASC;

Clearly it's a lot simpler. It's probably faster too but with only 14 records to test on it's hard to reach any sort of conclusion, no index will be used for such small tables.

Note that ST_DISTANCE results are returned in degrees it's conventionally assumed that 1 degree is about 60 miles or 111 km (you have done so in your calculation)

BTW, In the existing setup, you do have an index on latitude and longitude but please note that mysql can use only one index per table so if you don't adopt geospatial queries you might want to convert that into a single composite index on latitude,longitude.

The full query.

Now the above query can be modified as follows to give the 'query B' in it's new form.

SELECT DISTINCT  g1.asciiname, g2.asciiname ,ST_DISTANCE(g1.pt, g2.pt) *111 as distance FROM geoname2 g1 
INNER JOIN (SELECT `pt`, asciiname  
    FROM geoname2 
     WHERE (fcode='PPLX' OR fcode='PPPL') AND 
       ST_DISTANCE(`pt`, POINT(@lat,@lng))  <= 1) as g2
WHERE ST_DISTANCE(g1.pt,g2.pt) < 1 
AND g1.asciiname != g2.asciiname ORDER BY distance ASC;

Note again that I am assuming 1 degree (approximately 111 km to be close to one another)

标签
mysql
geolocation
geonames
user-variables
易学教程内所有资源均来自网络或用户发布的内容，如有违反法律规定的内容欢迎反馈！
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!