前言:項目中用到了postgreSQL中的earthdistance()函數功能計算地球上兩點之間的距離,中文的資料太少了,我找到了一篇英文的、講的很好的文章,特此翻譯,希望能夠幫助到以后用到earthdistance的同學。
一、兩種可用的選擇
當我們想用Postgres作為GEO函數使用時,我們通常有2中選擇(據我所知):
1.PostGIS: 為postgreSQL提供了高級GEO函數功能。我用了它一段時間,但是它對於我的需求來說太笨重了。
2.Cube和Earthdistance: 這兩個拓展為輕量級的Geo關系實體提供了簡單、快速的實現方法。
二、為什么在數據庫服務器端做計算
這是件非常明顯的事。服務器存儲了所有的數據,服務器拓展是用C/C++實現的,非常快。為數據表做索引也能加快計算速度。
三、使用我的選擇--Cube and EarthDistance
作為開始,你應該先建一個數據庫(我想你知道該怎么做),然后使它們能用我們的架構。 執行:
- CREATE EXTENSION cube;
- CREATE EXTENSION earthdistance;
在我們的例子中,我創建了名為events的表,字段有:id(serial), name(varchar 255), lat(double), lng(double)。(別忘了~~)
四、計算2個坐標之間的距離
計算2個坐標之間的距離,我們要用到earth_distance(ll_to_earth($latlngcube), ll_to_earth($latlng_cube))這個函數。 earth_distance()函數接受2組坐標值,返回值一個以米為單位的的數值。這能用於很多場景,比如根據某一位置找到離其最近的發生的新聞事件的列表。
【譯者注】這里要提下幾個重要的函數:(參考:http://www.postgresql.org/docs/8.3/static/earthdistance.html)
Table F-3. Cube-based earthdistance functions
Function | Returns | Description |
---|---|---|
earth() |
float8 | Returns the assumed radius of the Earth. |
sec_to_gc(float8) |
float8 | Converts the normal straight line (secant) distance between between two points on the surface of the Earth to the great circle distance between them. |
gc_to_sec(float8) |
float8 | Converts the great circle distance between two points on the surface of the Earth to the normal straight line (secant) distance between them. |
ll_to_earth(float8, float8) |
earth | Returns the location of a point on the surface of the Earth given its latitude (argument 1) and longitude (argument 2) in degrees. |
latitude(earth) |
float8 | Returns the latitude in degrees of a point on the surface of the Earth. |
longitude(earth) |
float8 | Returns the longitude in degrees of a point on the surface of the Earth. |
earth_distance(earth, earth) |
float8 | Returns the great circle distance between two points on the surface of the Earth. |
earth_box(earth, float8) |
cube | Returns a box suitable for an indexed search using the cube @> operator for points within a given great circle distance of a location. Some points in this box are further than the specified great circle distance from the location, so a second check using earth_distance should be included in the query. |
數據庫的操作可能就像下面這樣:
- SELECT events.id events.name, eaerthdiatance(ll_to_earth({currentuserlat}, {currentuserlng}), llto_earth(events.lat, events.lng))
- as distancefromcurrentlocation FROM events
- ORDER BY distancefromcurretnlocation ASC;
這將給我們一個很nice的新聞事件列表,按他們的離我們當前位置的距離由近到遠排序。第一個是離我們最近的。
五、找到某個半徑范圍內的記錄
Cube和Earthdiatance拓展提供的另一個偉大的函數是earth_box(ll_to_earch($latlngcub), $radiusinmetres)。 這個函數通過簡單的比較就能到找到某個半徑范圍內的所有記錄。它是靠返回2點之間的“大圓距離”實現的。
【譯者注】大圓距離(Great circle disstance)指的是從球面的一點A出發到達球面上另一點B,所經過的最短路徑的長度。一般說來,球面上任意兩點A和B都可以與球心確定唯一的大圓,這個大圓被稱為黎曼圓,而在大圓上連接這兩點的較短的一條弧的長度就是大圓距離。如果想了解更多,請看wiki:大圓距離
它能用於查詢我們城市中所有的新聞事件:
- SELECT events.id, events.name FROM events
- WHERE earth_box({currentuserlat}, {currentuserlng}, {radiusinmetres}) @> ll_to_earth(events.lat, events.lng);
這條查詢語句僅僅會返回在radius_in_metres指定的半徑范圍內的記錄,非常簡單吧!
六、提高查詢速度
你可能會發現上面的查詢有不小的開銷。以我的經驗,最好對一些字段建立索引。 (下面這條語句假定你又events表, 同時events表有字段lat和lng)
- CREATE INDEX ${nameofindex} on events USING gits(lltoearth(lat, lng));
七、數據類型
我的應用比較簡單,所以我把經緯度(lat和lng)都設成了double類型。這使得我用Node.js開發起來更加快速,而不用再去自己定制針對GIST類型的解決方案。
八、就這些!
很神奇,對么?!?我們僅僅用常用的數據類型(double)就足以去用一些GEO函數創建基於地理位置的社交app(【譯者注】 知乎上的一個回答 )!
---------------------------
英語水平有限,如有翻譯不周之處,請您指點!
------------------------------
update:
我使用的postgreSQL語句總結:
- /*
- * postgreSQL之earthdistance學習筆記
- * author: wusuopubupt
- * date: 2013-03-31
- */
- /*創建表*/
- CREATE TABLE picture (
- id serial PRIMARY KEY ,
- p_uid char(12) NOT NULL,
- p_key char(23) NOT NULL,
- lat real not null,
- lng real NOT NULL,
- up int NOT NULL,
- down int NOT NULL,
- ip varchar(15) DEFAULT NULL,
- address varchar(256) DEFAULT NULL
- );
- /*插入記錄*/
- INSERT INTO picture(p_uid, p_key, lat, lng, up, down, ip, address)
- VALUES('aaaabbbbcccc', '2014032008164023279.png', 40.043945, 116.413668, 0, 0, '', '');
- /*插入記錄*/
- INSERT INTO picture(p_uid, p_key, lat, lng, up, down, ip, address)
- VALUES('xxxxccccmmmm', '2014032008164023111.png', 40.067183, 116.415230, 0, 0, '', '');
- /*選擇記錄*/
- SELECT * FROM picture;
- /*更新記錄*/
- UPDATE picture SET address='LiShuiqiao' WHERE id=1;
- UPDATE picture SET address='TianTongyuan' WHERE id=2;
- /*對經緯度列創建索引*/
- CREATE INDEX ll_idx on picture USING gist(ll_to_earth(lat, lng));
- /*根據半徑(1000米)選擇記錄*/
- SELECT * FROM picture where earth_box(ll_to_earth(40.059286,116.418773),1000) @> ll_to_earth(picture.lat, picture.lng);
- /*選擇距離當前用戶的距離*/
- SELECT picture.id, earth_distance(ll_to_earth(picture.lat, picture.lng), ll_to_earth(40.059286,116.418773))
- AS dis FROM picture
- ORDER BY dis ASC;
- /*
- * 以下內容是網上的一篇教程
- * 地址:http://www.cse.iitb.ac.in/dbms/Data/Courses/CS631/PostgreSQL-Resources/postgresql-9.2.4/contrib/earthdistance/expected/earthdistance.out
- */
- --
- -- Test earthdistance extension
- --
- -- In this file we also do some testing of extension create/drop scenarios.
- -- That's really exercising the core database's dependency logic, so ideally
- -- we'd do it in the core regression tests, but we can't for lack of suitable
- -- guaranteed-available extensions. earthdistance is a good test case because
- -- it has a dependency on the cube extension.
- --
- CREATE EXTENSION earthdistance; -- fail, must install cube first
- ERROR: required extension "cube" is not installed
- CREATE EXTENSION cube;
- CREATE EXTENSION earthdistance;
- --
- -- The radius of the Earth we are using.
- --
- SELECT earth()::numeric(20,5);
- earth
- ---------------
- 6378168.00000
- (1 row)
- --
- -- Convert straight line distances to great circle distances.把直線距離轉成大圓距離
- --
- SELECT (pi()*earth())::numeric(20,5);
- numeric
- ----------------
- 20037605.73216
- (1 row)
- SELECT sec_to_gc(0)::numeric(20,5);
- sec_to_gc
- -----------
- 0.00000
- (1 row)
- --
- -- Convert great circle distances to straight line distances.
- --
- SELECT gc_to_sec(0)::numeric(20,5);
- gc_to_sec
- -----------
- 0.00000
- (1 row)
- SELECT gc_to_sec(sec_to_gc(2*earth()))::numeric(20,5);
- gc_to_sec
- ----------------
- 12756336.00000
- (1 row)
- --
- -- Set coordinates using latitude and longitude.
- -- Extract each coordinate separately so we can round them.
- --
- SELECT cube_ll_coord(ll_to_earth(0,0),1)::numeric(20,5),
- cube_ll_coord(ll_to_earth(0,0),2)::numeric(20,5),
- cube_ll_coord(ll_to_earth(0,0),3)::numeric(20,5);
- cube_ll_coord | cube_ll_coord | cube_ll_coord
- ---------------+---------------+---------------
- 6378168.00000 | 0.00000 | 0.00000
- (1 row)
- SELECT cube_ll_coord(ll_to_earth(360,360),1)::numeric(20,5),
- cube_ll_coord(ll_to_earth(360,360),2)::numeric(20,5),
- cube_ll_coord(ll_to_earth(360,360),3)::numeric(20,5);
- cube_ll_coord | cube_ll_coord | cube_ll_coord
- ---------------+---------------+---------------
- 6378168.00000 | 0.00000 | 0.00000
- (1 row)
- --
- -- Test getting the latitude of a location.
- --
- SELECT latitude(ll_to_earth(0,0))::numeric(20,10);
- latitude
- --------------
- 0.0000000000
- (1 row)
- SELECT latitude(ll_to_earth(45,0))::numeric(20,10);
- latitude
- ---------------
- 45.0000000000
- (1 row)
- --
- -- Test getting the longitude of a location.
- --
- SELECT longitude(ll_to_earth(0,0))::numeric(20,10);
- longitude
- --------------
- 0.0000000000
- (1 row)
- SELECT longitude(ll_to_earth(45,0))::numeric(20,10);
- longitude
- --------------
- 0.0000000000
- (1 row)
- --
- -- For the distance tests the following is some real life data.
- --
- -- Chicago has a latitude of 41.8 and a longitude of 87.6.
- -- Albuquerque has a latitude of 35.1 and a longitude of 106.7.
- -- (Note that latitude and longitude are specified differently
- -- in the cube based functions than for the point based functions.)
- --
- --
- -- Test getting the distance between two points using earth_distance.
- --
- SELECT earth_distance(ll_to_earth(0,0),ll_to_earth(0,0))::numeric(20,5);
- earth_distance
- ----------------
- 0.00000
- (1 row)
- SELECT earth_distance(ll_to_earth(0,0),ll_to_earth(0,180))::numeric(20,5);
- earth_distance
- ----------------
- 20037605.73216
- (1 row)
- --
- -- Test getting the distance between two points using geo_distance.
- --
- SELECT geo_distance('(0,0)'::point,'(0,0)'::point)::numeric(20,5);
- geo_distance
- --------------
- 0.00000
- (1 row)
- SELECT geo_distance('(0,0)'::point,'(180,0)'::point)::numeric(20,5);
- geo_distance
- --------------
- 12436.77274
- (1 row)
- --
- -- Test getting the distance between two points using the <@> operator.
- --
- SELECT ('(0,0)'::point <@> '(0,0)'::point)::numeric(20,5);
- numeric
- ---------
- 0.00000
- (1 row)
- SELECT ('(0,0)'::point <@> '(180,0)'::point)::numeric(20,5);
- numeric
- -------------
- 12436.77274
- (1 row)
- --
- -- Test for points that should be in bounding boxes.
- --
- SELECT earth_box(ll_to_earth(0,0),
- earth_distance(ll_to_earth(0,0),ll_to_earth(0,1))*1.00001) @>
- ll_to_earth(0,1);
- ?column?
- ----------
- t
- (1 row)
- SELECT earth_box(ll_to_earth(0,0),
- earth_distance(ll_to_earth(0,0),ll_to_earth(0,0.1))*1.00001) @>
- ll_to_earth(0,0.1);
- ?column?
- ----------
- t
- (1 row)
- --
- -- Test for points that shouldn't be in bounding boxes. Note that we need
- -- to make points way outside, since some points close may be in the box
- -- but further away than the distance we are testing.
- --
- SELECT earth_box(ll_to_earth(0,0),
- earth_distance(ll_to_earth(0,0),ll_to_earth(0,1))*.57735) @>
- ll_to_earth(0,1);
- ?column?
- ----------
- f
- (1 row)
- SELECT earth_box(ll_to_earth(0,0),
- earth_distance(ll_to_earth(0,0),ll_to_earth(0,0.1))*.57735) @>
- ll_to_earth(0,0.1);
- ?column?
- ----------
- f
- (1 row)