存储过程需要多长时间才能执行?

时间:2013-02-18 09:58:26

标签: mysql stored-procedures query-optimization

DELIMITER $$

CREATE PROCEDURE Load_Fact_List()

BEGIN

  DECLARE Project_Number_Temp INT;
  DECLARE Panel_Id_Temp INT;
  DECLARE Employee_Id_Temp INT;
  DECLARE Zip_Temp VARCHAR(255);
  DECLARE Created_Date_Temp DATE;
  DECLARE Country_Temp VARCHAR(255);


  DECLARE no_more_rows BOOLEAN;
  DECLARE loop_cntr INT DEFAULT 0;
  DECLARE num_rows INT DEFAULT 0;


  DECLARE   load_cur    CURSOR FOR
SELECT  Project_Id, Panel_Id, Employee_Id, Zip, Created_Date
    FROM  Fact_List;



  DECLARE CONTINUE HANDLER FOR NOT FOUND
    SET no_more_rows = TRUE;


  OPEN load_cur;
  select FOUND_ROWS() into num_rows;

  the_loop: LOOP

    FETCH  load_cur
    INTO   Project_Number_Temp, Panel_Id_Temp, Employee_Id_Temp, Zip_Temp, Created_Date_Temp;


    IF no_more_rows THEN
        CLOSE load_cur;
        LEAVE the_loop;
    END IF;

SET Country_Temp= (select Country from Zip where Zip= Zip_Temp);

INSERT INTO Test_Fact
(   
        Project_Key, 
        Campaign_Key, 
        Respondents_Key, 
        Event_Key, 
        Employee_Key, 
        Geography_Key, 
        Date_Key    
)

SELECT (SELECT Project_Key from Project_Dim where Project_Id= Project_Number_Temp AND Quota_Country= Country_Temp),0,(SELECT MAX(Respondents_Key) from Respondents_Dim WHERE Panel_Id= Panel_Id_Temp),1,(select MAX(Employee_Key) from Employee_Dim WHERE Employee_Id= Employee_Id_Temp),(Select Geography_Key from Geography_Dim where Zip= Zip_Temp), (Select Date_Key from Date_Dim where Full_Date= Created_Date_Temp);

    SET loop_cntr = loop_cntr + 1;
  END LOOP the_loop;


  select num_rows, loop_cntr;


END $$

上面的代码正常工作,但速度很慢。每1小时加载1000条记录。我没有记录加载到事实表中。任何人都可以建议我进行任何优化吗?

要求是通过循环遍历其他表并从维度表中收集所需的键值来加载事实表。

1 个答案:

答案 0 :(得分:1)

通常的程序实际上是这样的。

您已构建维度,并且您只是将要插入的数据收集到临时表中的事实表中。然后将此数据插入另一个临时表中,如下所示:

INSERT INTO tmp_fact_table
(
fact_key,
dim1_key,
dim2_key,
...
fact1,
fact2
...
)
SELECT 
ISNULL (f.fact_key, 0),
ISNULL (d1.sid, 0) as whatever,
ISNULL (d2.sid, 0) as whatever2,
...
ISNULL (tt.fact1, 0),
ISNULL (tt.fact2, 0)
FROM
yourTempTable tt 
LEFT JOIN Dim1 d1 ON tt.identifying_column = d1.identifying_column
...
LEFT JOIN fact_table f ON 
f.dim1_key = d1.sid
AND f.dim2_key = d2.sid

,其中

  • fact_key是事实表中的标识列
  • dim1_key是事实表中维度
  • 的外键
  • fact1等等是你在事实表中想要的事实,清楚
  • 当没有找到条目时,ISNULL()函数返回0。 0是未知数据的每个维度中虚拟行的ID

然后,您将拥有一个表,其中您的维度ID链接到您要导入到事实表中的数据,当事实表中的条目尚不存在时,其中0为事实键,并且ID为事实表条目另有说明。

然后更新事实表,其中tmp_fact_table.fact_key!= 0

然后插入事实表中tmp_fact_table.fact_key = 0

那就是它。

我这样做了数百万行,大约需要半个小时。花生300,000行。