【原创】POSTGRESQL 分区表初次体验

POSTGRESQL的分区和MYSQL不同，MYSQL是有专门的分区表， 而POSTGRESQL的分区则利用它本身的面向对象的特性来做。 下面我们来简单的体验下。


我们先创建一张父表。 记住，所有的分区表都得继承他。

t_girl=# create table num_master (id int not null primary key);
CREATE TABLE

接下来我们创建一个简单的函数来动态创建分区表。

t_girl=# create or replace function create_partition_table () returns void as $$
t_girl$# declare i int;
t_girl$# declare cnt int;
t_girl$# declare stmt text;
t_girl$# begin
t_girl$# -- Created by ytt at 2013/12/15. Dynamic creating partition tables.
t_girl$# i:= 0;
t_girl$# cnt:=4;
t_girl$# <<lable1>> while i < cnt loop
t_girl$#   stmt := 'create table num_slave'||i+1||'(check(id >='||i*100||' and id <'||(i+1)*100||')) inherits(num_master)';
t_girl$# execute stmt;
t_girl$#   i:=i + 1;
t_girl$# end loop lable1;
t_girl$# return;
t_girl$# end;
t_girl$# $$ language plpgsql;
CREATE FUNCTION
t_girl=#

OK。 现在可以执行了。

t_girl=# select create_partition_table();
 create_partition_table
------------------------

(1 row)

列出所有的表

t_girl=# \d
           List of relations
 Schema | Name | Type | Owner
--------+------------+-------+----------
 ytt | num_master | table | postgres
 ytt | num_slave1 | table | postgres
 ytt | num_slave2 | table | postgres
 ytt | num_slave3 | table | postgres
 ytt | num_slave4 | table | postgres
 ytt | t1 | table | t_girl
(6 rows)

我们针对父表建立一个触发器函数体，对应其分区表的数据分布。

t_girl=# create or replace function num_insert_trigger()
t_girl-# returns trigger as $$
t_girl$# begin
t_girl$# -- Created by ytt at 2013/12/15. Do how to distribute data.
t_girl$# if (new.id >=0 and new.id <100) then
t_girl$# insert into num_slave1 values (new.*);
t_girl$# elsif (new.id >=100 and new.id <200) then
t_girl$# insert into num_slave2 values(new.*);
t_girl$# elsif (new.id >=200 and new.id <300) then
t_girl$# insert into num_slave3 values (new.*);
t_girl$# elsif (new.id >=300 and new.id <400) then
t_girl$# insert into num_slave4 values (new.*);
t_girl$# else
t_girl$# raise exception 'Column id out of range.';
t_girl$# end if;
t_girl$# return null;
t_girl$# end;
t_girl$# $$
t_girl-# language plpgsql;
CREATE FUNCTION

我们看看已经建好的触发器：

t_girl=# \d+ num_master
                       Table "ytt.num_master"
 Column | Type | Modifiers | Storage | Stats target | Description
--------+---------+-----------+---------+--------------+-------------
 id | integer | not null | plain | |
Indexes:
    "num_master_pkey" PRIMARY KEY, btree (id)
Triggers:
    insert_num_slave_trigger BEFORE INSERT ON num_master FOR EACH ROW EXECUTE PROCEDURE ytt.num_insert_trigger()
Child tables: num_slave1,
              num_slave2,
              num_slave3,
              num_slave4
Has OIDs: no

我们现在生成简单的测试数据。

t_girl=# select func_create_sample_data();
 func_create_sample_data
-------------------------

(1 row)

上面的函数生成了大概400行的数据。




为了查看优化器是如何处理查询的，我们来看看简单的查询

t_girl=# explain select * from num_master where id > 30 and id < 120;
                           QUERY PLAN
-----------------------------------------------------------------
 Append (cost=0.00..5.00 rows=91 width=4)
   -> Seq Scan on num_master (cost=0.00..0.00 rows=1 width=4)
         Filter: ((id > 30) AND (id < 120))
   -> Seq Scan on num_slave1 (cost=0.00..2.50 rows=70 width=4)
         Filter: ((id > 30) AND (id < 120))
   -> Seq Scan on num_slave2 (cost=0.00..2.50 rows=20 width=4)
         Filter: ((id > 30) AND (id < 120))
(7 rows)


t_girl=#

我也是今天刚刚接触到POSTGRESQL的分区表，有问题，还希望提出。