[转帖]PG语法解剖--基本sql语句用法入门

PG语法解剖--基本sql语句用法入门

https://www.toutiao.com/i6710897833953722894/

COPY 命令挺好的 需要学习一下. 

原创 波波说运维 2019-07-12 00:02:00

概述

今天主要对PG数据库的一些基本SQL语句用法做个介绍，做个简单了解，也做备忘！

下面主要用例子来说明。

---

1、建表语句

create table test (id int8 primary key,info text,crt_time timestamp);

注意保留字

---

2、select into & create table as

postgres=# select * into table new_tbl from pg_class;postgres=# create table tbl_1 as select * from pg_class;

---

3、插入\更新\删除\查询

insert into tbl (xx,xx) values (xx,xx);update tbl set xx=xx where xxx;delete from tbl where xxx=xxx;select xx from xx where xx...;

如果是delete|update limit，则：

update tbl set xx=xx where ctid = any ( array (select ctid from tbl where xx limit ? for update));delete from tbl where ctid = any ( array (select ctid from tbl where xx limit ? for update));

---

4、批量DML

insert into xx values (),(),...(); copy xx from stdin; copy xx from 'file'; pg_bulkloadupdate t set info=t1.info,crt_time=t1.crt_time from t1,t2 where (t.id=t1.id) and t1.id=t2.id;update tbl_1 set relname=tmp.rel from (values (1,'test1'),(2,'test2')) tmp (id, rel) where tmp.id=tbl_1.id;delete from t using t1 where t.id=t1.id;delete from tbl_1 using (values (1),(2)) tmp (rel) where tmp.rel=tbl_1.reltype;

注意update , delete 批量操作，JOIN不是一一对应时，更新目标可能会随机匹配。

---

5、DB端copy+客户端copy

• https://github.com/digoal/blog/blob/master/201805/20180516_03.md

• https://github.com/digoal/blog/blob/master/201805/20180510_01.md

5.1、copy为什么快？

协议：

5.2、DB 端copy

copy tbl to 'file';copy (SQL) to 'file';copy tbl from 'file';

5.3、客户端copy

copy tbl from stdin;copy (SQL) to stdout;copy tbl to stdout;psql (\copy to | from); -- copy协议

---

6、排序+ offset limit

select * from tbl_1 order by relname nulls first;select * from tbl_1 order by relname nulls last;select * from tbl_1 order by relname;select * from tbl_1 order by relname limit 10 offset 10;select * from tbl_1 order by relname::text collate "C";

---

7、聚合+解耦合

select string_agg(relname,',' order by xx) from tbl_1; select g,avg(c1) from tbl group by g;

---

8、distinct

select distinct relname,relnamespace from pg_class;SELECT id, COUNT_DISTINCT(val) FROM test_table GROUP BY 1;select count(distinct (relname,relnamespace)) from pg_class;select distinct on (c3) c2,c3 from tbl;

---

9、INNER|OUTER JOIN

•inner

select * from t1 join t2 on (t1.x=t2.x) where xxxx;

• left

1)scan filter

select t1.*,t2.* from t1 left join t2 on (t1.x=t2.x) where t1.x=x;

2)join filter

select t1.*,t2.* from t1 left join t2 on (t1.x=t2.x and t1.x=x);

• right

把上面的left join改成right join即可，这里就不多说了。