存储过程运行比查询慢得多

Question

我有一个存储过程，它为优势比创建一个2x2表。基本比值比表如下所示：

编辑 - 此查询最终完成，并且在两分钟和32次单独调用函数后确实返回了正确的答案。我不明白为什么这是递归运行，任何想法，SO？

A - only records that satisfy both thing 1 and thing 2 go here
B - only records that satisfy thing 1 (people with thing 2 CANNOT go here)
C - only records that satisfy thing 2 (people with thing 1 CANNOT go here)
D - people with thing 1 OR thing 2 cannot go here

表格中的所有单元格都是代表一群人的整数。

我试图学习一些新语法，并决定使用intersect和except。我想创建thing 1和thing 2个变量，因此我将下面的查询放入存储过程。

CREATE PROC Findoddsratio (@diag1 NVARCHAR(5), 
                           @diag2 NVARCHAR(5)) 
AS 
    IF Object_id('tempdb..#temp') IS NOT NULL 
      DROP TABLE #temp 

    CREATE TABLE #temp 
      ( 
         squarenumber CHAR(1), 
         counts       FLOAT 
      ) 

    INSERT INTO #temp 
                (squarenumber, 
                 counts) 
    SELECT * 
    FROM   ( 
           --both + 
           SELECT 'a'                                    AS squareNumber, 
                  Cast(Count(DISTINCT x.counts)AS FLOAT) AS counts 
           FROM   (SELECT DISTINCT ic.patid AS counts 
                   FROM   icdclm AS ic 
                   WHERE  ic.icd LIKE @diag1 
                   INTERSECT 
                   SELECT DISTINCT ic.patid AS counts 
                   FROM   icdclm AS ic 
                   WHERE  ic.icd LIKE @diag2)x 
           UNION 
           --only 1+ 
           SELECT 'b', 
                  Count(DISTINCT x.counts) 
           FROM   (SELECT DISTINCT ic.patid AS counts 
                   FROM   icdclm AS ic 
                   WHERE  ic.icd LIKE @diag1 
                   EXCEPT 
                   SELECT DISTINCT ic.patid AS counts 
                   FROM   icdclm AS ic 
                   WHERE  ic.icd LIKE @diag2)AS x 
           UNION 
           --only 2+ 
           SELECT 'c', 
                  Count(DISTINCT x.counts) 
           FROM   (SELECT DISTINCT ic.patid AS counts 
                   FROM   icdclm AS ic 
                   WHERE  ic.icd LIKE @diag2 
                   EXCEPT 
                   SELECT DISTINCT ic.patid AS counts 
                   FROM   icdclm AS ic 
                   WHERE  ic.icd LIKE @diag1)AS x 
            UNION 
            --both - 
            SELECT 'd', 
                   Count(DISTINCT x.counts) 
            FROM   (SELECT DISTINCT ic.patid AS counts 
                    FROM   icdclm AS ic 
                    EXCEPT 
                    SELECT DISTINCT ic.patid AS counts 
                    FROM   icdclm AS ic 
                    WHERE  ic.icd LIKE @diag2 
                    EXCEPT 
                    SELECT DISTINCT ic.patid AS counts 
                    FROM   icdclm AS ic 
                    WHERE  ic.icd LIKE @diag1) AS x)y 

    --i used a pivot table to make the math work out easier 
    SELECT Round(Cast(( a * d ) / ( b * c ) AS FLOAT), 2) AS OddsRatio 
    FROM   (SELECT [a], 
                   [b], 
                   [c], 
                   [d] 
            FROM   (SELECT [squarenumber], 
                           [counts] 
                    FROM   #temp) p 
                   PIVOT ( Sum(counts) 
                         FOR [squarenumber] IN ([a], 
                                                [b], 
                                                [c], 
                                                [d]) ) AS pvt)t 

ICDCLM是一个包含patid=int, icd=varchar(5)

结构的表格

ICDCLM中有〜百万行。当我运行此查询而不使其成为存储过程时，它会在几秒钟内运行。如果我尝试exec FindsOddsRation 'thing1%','thing2%'。它运行并运行，但从不返回任何内容（> 2分钟）。存储过程花了这么长时间的差异是什么？ SQL Server 2008 R2 小提琴here

Answer 1

如果您运行与存储过程完全相同的SQL且时间不同，则您的存储过程可能依赖于过时的元数据。尝试更新统计信息或重新编译存储过程。