postgresql 模糊查询优化

1、安装pg_trgm插件

1. 到postgres软件安装目录：

　　　　/home/postgres/soft/postgresql-9.6.12/contrib/pg_trgm

　　2、翻译安装

　　　　make USE_PGXS=1  && make USE_PGXS=1 install

　　3、检查安装是否成功。

　　　　ll $PGHOME/lib/pg_trgm.so

　　　　显示如下，安装成功：

　　　　-rwxr-xr-x 1 postgres postgres 53112 3月  25 10:42 /home/postgres/postgresql-9.6.12/lib/pg_trgm.so

　　4、创建extension

testdb=# create extension pg_trgm;

CREATE EXTENSION

2、优化过程

　　1、准备表和测试数据。　　

testdb=# \dt+ t_user
                     List of relations
 Schema |  Name  | Type  |  Owner   |  Size  | Description 
--------+--------+-------+----------+--------+-------------
 public | t_user | table | postgres | 769 MB | 
(1 row)

testdb=# select count(1) from t_user;
  count  
---------
 6000000
(1 row)

testdb=# 

　　2、优化前：

testdb=# \d+ t_user;
                               Table "public.t_user"
 Column |         Type          | Modifiers | Storage  | Stats target | Description 
--------+-----------------------+-----------+----------+--------------+-------------
 id     | bigint                | not null  | plain    |              | 
 name   | character varying(50) |           | extended |              | 
 email  | character varying(50) |           | extended |              | 
 type   | integer               |           | plain    |              | 
 iphone | character varying(13) |           | extended |              | 
Indexes:
    "t_user_pkey" PRIMARY KEY, btree (id)

testdb=# explain ANALYZE select * from t_user where name like '%abc%';
                                                  QUERY PLAN                                                   
---------------------------------------------------------------------------------------------------------------
 Seq Scan on t_user  (cost=0.00..173361.26 rows=600 width=99) (actual time=0.179..2261.550 rows=43722 loops=1)
   Filter: ((name)::text ~~ '%abc%'::text)
   Rows Removed by Filter: 5956278
 Planning time: 0.665 ms
 Execution time: 2265.169 ms
(5 rows)

testdb=# explain ANALYZE select * from t_user where name like '%abc%';
                                                  QUERY PLAN                                                   
---------------------------------------------------------------------------------------------------------------
 Seq Scan on t_user  (cost=0.00..173361.26 rows=600 width=99) (actual time=0.108..1494.702 rows=43722 loops=1)
   Filter: ((name)::text ~~ '%abc%'::text)
   Rows Removed by Filter: 5956278
 Planning time: 0.162 ms
 Execution time: 1496.646 ms
(5 rows)

　　3、创建索引，并收集统计信息：

testdb=# CREATE INDEX idx_trgm_user_name_ ON t_user USING GIN(name gin_trgm_ops);

CREATE INDEX
testdb=# 
testdb=# ANALYZE t_user;
ANALYZE

　　4、优化后：

testdb=# explain ANALYZE select * from t_user where name like '%abc%';           
                                                             QUERY PLAN                                                              
-------------------------------------------------------------------------------------------------------------------------------------
 Bitmap Heap Scan on t_user  (cost=20.65..2283.92 rows=600 width=99) (actual time=36.398..132.911 rows=43722 loops=1)
   Recheck Cond: ((name)::text ~~ '%abc%'::text)
   Heap Blocks: exact=35367
   ->  Bitmap Index Scan on idx_trgm_user_name_  (cost=0.00..20.50 rows=600 width=0) (actual time=21.842..21.842 rows=43722 loops=1)
         Index Cond: ((name)::text ~~ '%abc%'::text)
 Planning time: 8.502 ms
 Execution time: 136.646 ms
(7 rows)

testdb=# explain ANALYZE select * from t_user where name like '%abc%';
                                                             QUERY PLAN                                                              
-------------------------------------------------------------------------------------------------------------------------------------
 Bitmap Heap Scan on t_user  (cost=20.65..2283.92 rows=600 width=99) (actual time=34.787..124.970 rows=43722 loops=1)
   Recheck Cond: ((name)::text ~~ '%abc%'::text)
   Heap Blocks: exact=35367
   ->  Bitmap Index Scan on idx_trgm_user_name_  (cost=0.00..20.50 rows=600 width=0) (actual time=20.599..20.599 rows=43722 loops=1)
         Index Cond: ((name)::text ~~ '%abc%'::text)
 Planning time: 0.214 ms
 Execution time: 128.587 ms
(7 rows)

结论：

没有创建索引时。查询需要2秒左右。优化后只需要0.1秒。相差20倍