我们经常会对数据字典中的系统表进行遍历，从而写一些SHELL脚本或者动态造数据等等。 这里我用PLsql演示了三种方法来遍历一张表。

表结构如下，

t_girl=# \d tmp_1;
           Unlogged table "public.tmp_1"
  Column  |            Type             | Modifiers
----------+-----------------------------+-----------
 id       | integer                     |
 log_time | timestamp without time zone |

在这里我创建里一个自定义类型来保存我的函数返回值。

create type ytt_record as (id int,log_time timestamp without time zone);

现在来看第一个函数。 也是用最笨的方法来遍历。

create or replace function sp_test_record1(
IN f_id int
) returns setof ytt_record as
$ytt$
declare i int;
declare cnt int;
declare o_out ytt_record;
begin
   i := 0;
   cnt := 0;
   select count(*)   into  cnt from tmp_1 where id > f_id;
   while i < cnt
   loop
       select id,log_time into strict o_out from tmp_1 where id > f_id order by log_time desc limit 1 offset i;
     i := i + 1;
     return next o_out;
   end loop;
end;
$ytt$ language plpgsql;

我们来执行下结果，花费了3毫秒左右。

t_girl=# select  * from sp_test_record1(60);
 id |          log_time         
----+----------------------------
 85 | 2014-01-11 17:52:11.696354
 73 | 2014-01-09 17:52:11.696354
 77 | 2014-01-04 17:52:11.696354
 80 | 2014-01-03 17:52:11.696354
 76 | 2014-01-02 17:52:11.696354
 65 | 2013-12-31 17:52:11.696354
 80 | 2013-12-30 17:52:11.098336
 85 | 2013-12-27 17:52:11.098336
 97 | 2013-12-26 17:52:11.696354
 94 | 2013-12-24 17:52:09.321394
(10 rows)
Time: 3.338 ms

现在来看第二个函数，这个就比较优化了， 用了系统自带的循环遍历结构。

create or replace function sp_test_record2(
IN f_id int
) returns setof ytt_record as
$ytt$
declare o_out ytt_record;
begin
  for o_out in   select id,log_time  from tmp_1 where id > f_id order by log_time desc
  loop
     return next o_out;
   end loop;
end;
$ytt$ language plpgsql;

这次运行结果看看，时间不到1毫秒。

t_girl=# select * from sp_test_record2(60);
 id |          log_time         
----+----------------------------
 85 | 2014-01-11 17:52:11.696354
 73 | 2014-01-09 17:52:11.696354
 77 | 2014-01-04 17:52:11.696354
 80 | 2014-01-03 17:52:11.696354
 76 | 2014-01-02 17:52:11.696354
 65 | 2013-12-31 17:52:11.696354
 80 | 2013-12-30 17:52:11.098336
 85 | 2013-12-27 17:52:11.098336
 97 | 2013-12-26 17:52:11.696354
 94 | 2013-12-24 17:52:09.321394
(10 rows)
Time: 0.660 ms

最后一个函数， 利用RETURN QUERY 直接返回结果集。

create or replace function sp_test_record3(
IN f_id int
) returns setof ytt_record as
$ytt$
begin
  return query select id,log_time  from tmp_1 where id > f_id order by log_time desc ;
end;
$ytt$ language plpgsql;

这个结果其实等同于直接从表SELECT，响应时间和第二个差不多。

t_girl=# select sp_test_record3(60);
          sp_test_record3         
-----------------------------------
 (85,"2014-01-11 17:52:11.696354")
 (73,"2014-01-09 17:52:11.696354")
 (77,"2014-01-04 17:52:11.696354")
 (80,"2014-01-03 17:52:11.696354")
 (76,"2014-01-02 17:52:11.696354")
 (65,"2013-12-31 17:52:11.696354")
 (80,"2013-12-30 17:52:11.098336")
 (85,"2013-12-27 17:52:11.098336")
 (97,"2013-12-26 17:52:11.696354")
 (94,"2013-12-24 17:52:09.321394")
(10 rows)
Time: 0.877 ms
t_girl=#

本文出自 “上帝，咱们不见不散！” 博客，请务必保留此出处http://www.jb51.cc/article/p-febgcqik-bs.html