极乐门资源网 Design By www.ioogu.com

场景 lower(name) like 'pf%'

create table users (id int primary key, name varchar(255));
Create or replace function random_string(length integer) returns text as
$$
declare
 chars text[] := '{0,1,2,3,4,5,6,7,8,9,A,B,C,D,E,F,G,H,I,J,K,L,M,N,O,P,Q,R,S,T,U,V,W,X,Y,Z,a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t,u,v,w,x,y,z}';
 result text := '';
 i integer := 0;
begin
 if length < 0 then
 raise exception 'Given length cannot be less than 0';
 end if;
 for i in 1..length loop
 result := result || chars[1+random()*(array_length(chars, 1)-1)];
 end loop;
 return result;
end;
$$ language plpgsql;
insert into users values(generate_series(1,50000), random_string(15));

普通bt:不走索引

pg_trgm模块提供函数和操作符测定字母数字文本基于三元模型匹配的相似性,还有支持快速搜索相似字符串的索引操作符类。三元模型是一组从一个字符串中获得的三个连续的字符。我们可以通过计数两个字符串共享的三元模型的数量来测量它们的相似性。这个简单的想法证明在测量许多自然语言词汇的相似性时是非常有效的。

CREATE INDEX users_idx0 ON users (name);

全字匹配查询(走索引)

explain select * from users where name='pfDNQVmhqDrF1EY';
        QUERY PLAN
-------------------------------------------------------------------------
 Index Scan using users_idx0 on users (cost=0.29..8.31 rows=1 width=20)
 Index Cond: ((name)::text = 'pfDNQVmhqDrF1EY'::text)
(2 rows)

加函数全字匹配(不走索引)

explain select * from users where lower(name)='pfDNQVmhqDrF1EY';
      QUERY PLAN
-----------------------------------------------------------
 Seq Scan on users (cost=0.00..1069.00 rows=250 width=20)
 Filter: (lower((name)::text) = 'pfDNQVmhqDrF1EY'::text)
(2 rows)

模糊匹配(不走索引)

explain select * from users where name like 'pf%';
      QUERY PLAN
--------------------------------------------------------
 Seq Scan on users (cost=0.00..944.00 rows=5 width=20)
 Filter: ((name)::text ~~ 'pf%'::text)
explain select * from users where name like 'pf_';
      QUERY PLAN
--------------------------------------------------------
 Seq Scan on users (cost=0.00..944.00 rows=5 width=20)
 Filter: ((name)::text ~~ 'pf_'::text)

字段带函数的bt索引:函数走索引

drop index users_idx0;
CREATE INDEX users_dex1 ON users (lower(name));

加函数全字匹配(走索引)

explain select * from users where lower(name)='pfDNQVmhqDrF1EY';
        QUERY PLAN
---------------------------------------------------------------------------
 Bitmap Heap Scan on users (cost=6.23..324.34 rows=250 width=20)
 Recheck Cond: (lower((name)::text) = 'pfDNQVmhqDrF1EY'::text)
 -> Bitmap Index Scan on users_dex1 (cost=0.00..6.17 rows=250 width=0)
   Index Cond: (lower((name)::text) = 'pfDNQVmhqDrF1EY'::text)
(4 rows)

模糊匹配(不走索引)

explain select * from users where lower(name) like 'pf%';
      QUERY PLAN
-----------------------------------------------------------
 Seq Scan on users (cost=0.00..1069.00 rows=250 width=20)
 Filter: (lower((name)::text) ~~ 'pf%'::text)
(2 rows)

声明操作符类的bt索引:like走索引

定义索引的同时可以为索引的每个字段声明一个操作符类。

CREATE INDEX name ON table (column opclass [sort options] [, …]);

这个操作符类指明该索引用于该字段时要使用的操作符。

CREATE INDEX users_dex2 ON users (lower(name) varchar_pattern_ops);

模糊匹配(走索引)

explain select * from users where lower(name) like 'pf%';
            QUERY PLAN
------------------------------------------------------------------------------------------------------
 Bitmap Heap Scan on users (cost=4.82..144.00 rows=5 width=20)
 Filter: (lower((name)::text) ~~ 'pf%'::text)
 -> Bitmap Index Scan on users_dex2 (cost=0.00..4.82 rows=53 width=0)
   Index Cond: ((lower((name)::text) ~>=~ 'pf'::text) AND (lower((name)::text) ~<~ 'pg'::text))
(4 rows)

场景2 name like '%pf%'

Create or replace function random_string(length integer) returns text as
$$
declare
 chars text[] := '{0,1,2,3,4,5,6,7,8,9,A,B,C,D,E,F,G,H,I,J,K,L,M,N,O,P,Q,R,S,T,U,V,W,X,Y,Z,a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t,u,v,w,x,y,z}';
 result text := '';
 i integer := 0;
begin
 if length < 0 then
 raise exception 'Given length cannot be less than 0';
 end if;
 for i in 1..length loop
 result := result || chars[1+random()*(array_length(chars, 1)-1)];
 end loop;
 return result;
end;
$$ language plpgsql;
create table users (id int primary key, name varchar(255));
insert into users values(generate_series(1,50000), random_string(15));

声明操作符bt:不走索引

CREATE INDEX idx_name ON users USING btree (lower(name) varchar_pattern_ops);
explain (analyze true,format yaml, verbose true, buffers true) select * from users where lower(name) like '%pf%';      QUERY PLAN
-----------------------------------------------------------
 - Plan:             +
  Node Type: "Seq Scan"        +
  Parallel Aware: false        +
  Relation Name: "users"        +
  Schema: "public"          +
  Alias: "users"          +
  Startup Cost: 0.00         +
  Total Cost: 1069.00         +
  Plan Rows: 5           +
  Plan Width: 20          +
  Actual Startup Time: 0.320       +
  Actual Total Time: 86.841       +
  Actual Rows: 710          +
  Actual Loops: 1          +
  Output:            +
  - "id"            +
  - "name"           +
  Filter: "(lower((users.name)::text) ~~ '%pf%'::text)"+
  Rows Removed by Filter: 49290      +
  Shared Hit Blocks: 319        +
  Shared Read Blocks: 0        +
  Shared Dirtied Blocks: 0        +
  Shared Written Blocks: 0        +
  Local Hit Blocks: 0         +
  Local Read Blocks: 0         +
  Local Dirtied Blocks: 0        +
  Local Written Blocks: 0        +
  Temp Read Blocks: 0         +
  Temp Written Blocks: 0        +
 Planning Time: 0.188         +
 Triggers:            +
 Execution Time: 86.975

声明pg_trgm操作符bt:可以走索引

CREATE EXTENSION pg_trgm;
CREATE INDEX idx_users_name_trgm_gist ON users USING gist (name gist_trgm_ops);
explain (analyze true, verbose true, buffers true) select * from users where name like '%pf%';
                QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------------
 Bitmap Heap Scan on public.users (cost=32.19..371.08 rows=505 width=20) (actual time=19.314..53.132 rows=193 loops=1)
 Output: id, name
 Recheck Cond: ((users.name)::text ~~ '%pf%'::text)
 Rows Removed by Index Recheck: 49807
 Heap Blocks: exact=319
 Buffers: shared hit=972
 -> Bitmap Index Scan on idx_users_name_trgm_gist (cost=0.00..32.06 rows=505 width=0) (actual time=19.175..19.175 rows=50000 loops=1)
   Index Cond: ((users.name)::text ~~ '%pf%'::text)
   Buffers: shared hit=653
 Planning time: 0.188 ms
 Execution time: 53.231 ms
(11 rows)

以上为个人经验,希望能给大家一个参考,也希望大家多多支持。如有错误或未考虑完全的地方,望不吝赐教。

标签:
PostgreSQL,模糊匹配,走索引

极乐门资源网 Design By www.ioogu.com
极乐门资源网 免责声明:本站文章均来自网站采集或用户投稿,网站不提供任何软件下载或自行开发的软件! 如有用户或公司发现本站内容信息存在侵权行为,请邮件告知! 858582#qq.com
极乐门资源网 Design By www.ioogu.com

评论“PostgreSQL模糊匹配走索引的操作”

暂无PostgreSQL模糊匹配走索引的操作的评论...

《魔兽世界》大逃杀!60人新游玩模式《强袭风暴》3月21日上线

暴雪近日发布了《魔兽世界》10.2.6 更新内容,新游玩模式《强袭风暴》即将于3月21 日在亚服上线,届时玩家将前往阿拉希高地展开一场 60 人大逃杀对战。

艾泽拉斯的冒险者已经征服了艾泽拉斯的大地及遥远的彼岸。他们在对抗世界上最致命的敌人时展现出过人的手腕,并且成功阻止终结宇宙等级的威胁。当他们在为即将于《魔兽世界》资料片《地心之战》中来袭的萨拉塔斯势力做战斗准备时,他们还需要在熟悉的阿拉希高地面对一个全新的敌人──那就是彼此。在《巨龙崛起》10.2.6 更新的《强袭风暴》中,玩家将会进入一个全新的海盗主题大逃杀式限时活动,其中包含极高的风险和史诗级的奖励。

《强袭风暴》不是普通的战场,作为一个独立于主游戏之外的活动,玩家可以用大逃杀的风格来体验《魔兽世界》,不分职业、不分装备(除了你在赛局中捡到的),光是技巧和战略的强弱之分就能决定出谁才是能坚持到最后的赢家。本次活动将会开放单人和双人模式,玩家在加入海盗主题的预赛大厅区域前,可以从强袭风暴角色画面新增好友。游玩游戏将可以累计名望轨迹,《巨龙崛起》和《魔兽世界:巫妖王之怒 经典版》的玩家都可以获得奖励。