使用Postgresql 实现快速插入测试数据

原作者: [db:作者] 来自: [db:来源] 收藏邀请

1.创建常规的企业信息表

create table t_centerprises(
 objectid bigint not null, /*唯一编号(6位行政区号+6位sn)*/
 divid uuid not null, /*行政区唯一代码*/
 name text not null, /*企业名称*/
 address text not null, /*企业地址*/
 post text, /*企业邮编*/
 contacts text, /*联系人*/
 tel text, /*联系电话*/
 fax text, /*传真*/
 describe text, /*企业备注*/ 
 date timestamp default now() not null, /*创建日期*/
 constraint pk_centerprisess_objectid primary key (objectid),
 constraint fk_centerprises_divid foreign key(divid) references ts_divisions(objectid) on delete cascade
);
create index idx_centerprises_divid on t_centerprises(divid);

2.需要使用的函数

/*转换16进制到字符*/
drop function if exists hex_to_string(text);
create or replace function hex_to_string( text) 
 returns text as 
$$
 declare
 result text;
 begin
 execute 'select U&''\' || $1 || '''' INTO result;
 return result;
 end;
$$ language plpgsql;
 
/*随机生成汉字
 汉字范围U+4E00..U+9FA5
*/
drop function if exists gen_random_zh(int,int);
create or replace function gen_random_zh(imin int,imax int) 
 returns text as 
$$
 declare
 vlen integer;
 result text;
 begin
 result := '';
 vlen = floor(random()*(imax-imin)+imin);
 for i in 1..vlen loop
  result := result || hex_to_string(to_hex(floor(random()*(42191-19968)+19968)::integer));
 end loop;
 return result;
 end;
$$ language plpgsql;

3.常规测试数据插入(5000000条)

insert into t_centerprises(objectid,divid,name,address,post,contacts,tel,fax,describe)
  select (vdivid|| lpad(id::text,6,'0'))::bigint as objectid,'110101',
  gen_random_zh(5,25) as name,gen_random_zh(10,50) as address,
  floor(random()*(699999-600000)+600000) as post,gen_random_zh(2,8) as contacts,
  floor(random()*(69999999-60000000)+60000000) as tel,floor(random()*(69999999-60000000)+60000000) as fax,
  gen_random_zh(32,128) as describe
 from generate_series(1,5000000) as id;

在普通pc机上插入,大概完成时间约8小时,过程不可监控,并且cpu/内存占用率高,磁盘基本满负荷动作,读写率基本上都是100%.

4.改进后的方法, 插入(10000000条)

do $$
 declare vStart bigint;
 declare vEnd bigint;
 declare MAXVALE bigint;
 declare INTERVAL bigint;
 declare vprovince integer;
 declare vprefecture integer;
 declare vcounty integer;
 declare vdivid text;
 declare vdividex uuid;
begin
 vprovince := 10;vprefecture := 1;vcounty := 1;
 
 MAXVALE := 1000000;
 INTERVAL := 1000; vStart := 1 ;vEnd := INTERVAL;
 vdivid := (lpad(vprovince::text,2,'0') || lpad(vprefecture::text,2,'0') || lpad(vcounty::text,2,'0'))::text;
 vdividex := (select objectid from ts_divisions where province=vprovince and prefecture=vprefecture and county=vcounty);
 loop 
 insert into t_centerprises(objectid,divid,name,address,post,contacts,tel,fax,describe)
  select (vdivid|| lpad(id::text,6,'0'))::bigint as objectid,vdividex as divid,
  gen_random_zh(5,25) as name,gen_random_zh(10,50) as address,
  floor(random()*(699999-600000)+600000) as post,gen_random_zh(2,8) as contacts,
  floor(random()*(69999999-60000000)+60000000) as tel,floor(random()*(69999999-60000000)+60000000) as fax,
  gen_random_zh(32,128) as describe
 from generate_series(vStart,vEnd) as id;
 
 raise notice '%', vEnd;
 vStart := vEnd + 1; vEnd := vEnd + INTERVAL;
 if( vEnd > MAXVALE ) then
  return;
 elsif(vEnd = MAXVALE) then
  vEnd := vEnd - 1;
 end if;
 end loop;
end$$;

因为运算原因, cpu/内存占用率仍然很高, 硬盘负荷较小,读写率也比较低,大概完成时间约1.5小时.

补充：postgreSQL数据库向表中快速插入1000000条数据

不用创建函数，直接向表中快速插入1000000条数据

create table tbl_test (id int, info text, c_time timestamp);
insert into tbl_test select generate_series(1,100000),md5(random()::text),clock_timestamp();
select count(id) from tbl_test; --查看个数据条数

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极客世界。如有错误或未考虑完全的地方，望不吝赐教。