如何有效利用索引编写高效过程

USE  AdventureWorks;
GO
-- 主键只用来保证数据，而聚集索引影响数据的逻辑排序
-- 下面的语句执行计划中并没有看到排序操作，证明数据是按SalesOrderID排过序的，而且是一种双向链表
SELECT   TOP ( 10 )  *   FROM  Sales.SalesOrderHeader  ORDER   BY  SalesOrderID
SELECT   TOP ( 10 )  *   FROM  Sales.SalesOrderHeader  ORDER   BY  SalesOrderID  DESC

SELECT   TOP ( 10 ) SalesOrderID  FROM  Sales.SalesOrderHeader  ORDER   BY  SalesOrderID
SELECT   TOP ( 10 ) SalesOrderID  FROM  Sales.SalesOrderHeader

-- ======================================================================
-- 查看表内数据是如何存储的
-- ======================================================================
SELECT   object_name (p. object_id )  as  tablename,i.name  as  indexname,
    rows,a.type_desc  as  page_type_desc,
    total_pages  as  pages,first_page
FROM  sys.partitions p
     JOIN  sys.system_internals_allocation_units a
         ON  p.partition_id  =  a.container_id
     JOIN  sys.indexes i 
         ON  p.index_id  =  i.index_id  AND  p. object_id   =  i. object_id
WHERE  p. object_id = object_id (N ' Sales.SalesOrderHeader ' )

-- 查看所含的pages字段最多的则为聚集索引叶级，其余的为别的索引页
-- 0xA01400000100得知第一页为5280，此为聚集索引叶的第一页
DBCC  TRACEON( 3604 )
GO
DBCC  PAGE(AdventureWorks, 1 , 5280 , 3 ) -- 查看第4～7字节的值是按照SalesOrderID从小到大排列的
GO
-- 查看IX_SalesOrderHeader_CustomerID索引的第一页的第一个值是否为1
-- 0x105200000100得到第一页为21008
DBCC  PAGE(AdventureWorks, 1 , 21008 , 1 )
GO


-- ======================================================================
-- 行长度的限制8060
-- ======================================================================
USE  tempdb;
GO
IF   OBJECT_ID (N ' bigrows ' ,N ' U ' )  IS   NOT   NULL
     DROP   TABLE  dbo.bigrows
GO
CREATE   TABLE  dbo.bigrows
(
    a  char ( 3000 ),
    b  char ( 3000 ),
    c  char ( 2000 ),
    d  char ( 60 )
)
GO
INSERT   INTO  dbo.bigrows
     SELECT   REPLICATE ( ' a ' , 3000 ), REPLICATE ( ' b ' , 3000 ),
             REPLICATE ( ' c ' , 2000 ), REPLICATE ( ' d ' , 6 )
-- 2005使用行溢出数据突破了変长列的限制，但这样会增加I/O操作。
IF   OBJECT_ID (N ' bigrows ' ,N ' U ' )  IS   NOT   NULL
     DROP   TABLE  dbo.bigrows
GO
CREATE   TABLE  dbo.bigrows
(
    a  varchar ( 3000 ),
    b  varchar ( 3000 ),
    c  varchar ( 3000 ),
    d  varchar ( 3000 )
)
GO
INSERT   INTO  dbo.bigrows
     SELECT   REPLICATE ( ' a ' , 3000 ), REPLICATE ( ' b ' , 3000 ),
             REPLICATE ( ' c ' , 3000 ), REPLICATE ( ' d ' , 3000 )
SELECT   object_name ( object_id )  as  name,
    partition_id,partition_number  as  pnum,rows,
    allocation_unit_id  as  au_id,type_desc  as  page_type,total_pages  as  pages
FROM  sys.partitions p    
     JOIN  sys.allocation_units a
         ON  p.partition_id  =  a.container_id
WHERE   object_id   =   object_id (N ' dbo.bigrows ' );
-- ======================================================================
-- 非聚集索引通过聚集索引进行查找
-- ======================================================================
-- SQL2005新增用于查看表中各索引的统计的动态函数
SELECT   *   FROM  sys.dm_db_index_physical_stats( DB_ID (N ' AdventureWorks ' ),  OBJECT_ID (N ' Sales.SalesOrderHeader ' ),  NULL ,  NULL  ,  ' DETAILED ' );
-- 索引的深度
SELECT   INDEXPROPERTY ( object_ID ( ' Sales.SalesOrderHeader ' ),N ' PK_SalesOrderHeader_SalesOrderID ' , ' IndexDepth ' )
SELECT   INDEXPROPERTY ( object_ID ( ' Sales.SalesOrderHeader ' ),N ' AK_SalesOrderHeader_SalesOrderNumber ' , ' IndexDepth ' )
-- 聚集索引查找，因为索引的深度为3，所以此时逻辑读为3
SELECT   *   FROM  Sales.SalesOrderHeader  WHERE  SalesOrderID = 43659
-- 非聚集索引通过聚集索引进行查找(逻辑读有5次，因为聚集为3，非聚集为2)
SELECT   *   FROM  Sales.SalesOrderHeader  WHERE  SalesOrderNumber = N ' SO45283 '

-- ======================================================================
-- 复合索引测试
-- ======================================================================
BEGIN   TRANSACTION
GO
DROP   INDEX  IX_SalesOrderHeader_CustomerID  ON  Sales.SalesOrderHeader
GO
CREATE   NONCLUSTERED   INDEX  IX_SalesOrderHeader_CustomerID  ON  Sales.SalesOrderHeader
    (
    CustomerID,
    CreditCardID
    )  WITH ( STATISTICS_NORECOMPUTE  =   OFF , IGNORE_DUP_KEY  =   OFF , ALLOW_ROW_LOCKS  =   ON , ALLOW_PAGE_LOCKS  =   ON )  ON   [ PRIMARY ]
GO
COMMIT

UPDATE   STATISTICS  Sales.SalesOrderHeader IX_SalesOrderHeader_CustomerID
-- 查看此索引的统计信息
DBCC  SHOW_STATISTICS( ' Sales.SalesOrderHeader ' ,IX_SalesOrderHeader_CustomerID)

SELECT   COUNT ( * )  -- EQ_ROWS等于上限值的个数
FROM  Sales.SalesOrderHeader
WHERE  CustomerID = 1

SELECT   COUNT ( * )  -- RANGE_ROWS此范围内不包括边界的记录数
FROM  Sales.SalesOrderHeader
WHERE  CustomerID > 1   AND  CustomerID < 19

SELECT    COUNT ( DISTINCT  CustomerID) -- DISTINCT_RANGE_ROWS 此范围内不包括边界的不同记录数
FROM  Sales.SalesOrderHeader
WHERE  CustomerID > 1   AND  CustomerID < 19

SELECT    1.0 * COUNT ( * ) / COUNT ( DISTINCT  CustomerID) -- AVG_RANGE_ROWS 数据分布程度
FROM  Sales.SalesOrderHeader
WHERE  CustomerID > 1   AND  CustomerID < 19
-- 查看Sales.SalesOrderHeader表中索引信息
SELECT   *   FROM  
sys.dm_db_index_physical_stats( DB_ID (N ' AdventureWorks ' ),  OBJECT_ID (N ' Sales.SalesOrderHeader ' ),  NULL ,  NULL  ,  ' DETAILED ' );
-- IX_SalesOrderHeader_CustomerID的索引编号为5深度为2，叶级有71个页面，根叶1页
SELECT   QUOTENAME (name)
FROM  sys.indexes
WHERE    object_id   =   722101613   AND  index_id  =   5 ;

DBCC  DROPCLEANBUFFERS;
GO
DBCC  FREEPROCCACHE;
GO
-- 非聚集索引查找+键值查找(逻辑读为8，非聚集索引查找2页加每个值一次聚集索引查找)
SELECT   *   FROM  Sales.SalesOrderHeader  WHERE  CustomerID = 19965
-- 非聚集索引扫描+键值查找(因此使用复合索引除第一个字段外，不会进行索引查找)
SELECT   *   FROM  Sales.SalesOrderHeader  WHERE  CreditCardID = 15466
-- 覆盖索引查找(逻辑读为2，说明只在非聚集索引中查找到了所需的数据)
SELECT  SalesOrderID,CreditCardID  FROM  Sales.SalesOrderHeader
WHERE  CustomerID = 19965

-- 作为查询条件的字段会自动创建统计
SELECT   *   FROM  Production.Location  WHERE  ModifiedDate > ' 20070101 '
-- 所有自动创建的统计
SELECT  m.name + ' . ' + object_name (s. object_id )  as   [ table ] ,s.name  as   [ statistics ] ,c.name  as   [ column ]
FROM  sys.stats s
     join  sys.stats_columns sc
         on  s.stats_id  =  sc.stats_id  and  s. object_id   =  sc. object_id
     join  sys.columns c 
         on  sc.column_id  =  c.column_id  and  c. object_id   =  sc. object_id
     join  sys.objects o 
         on  s. object_id   =  o. object_id
     join  sys.schemas m
         on  m.schema_id  =  o.schema_id
where  s.auto_created = 1   and  o.type = ' U '

SELECT   object_name (id),name,statblob -- 统计信息二进制大型对象(BLOB)，仅内部使用，返回NULL.
FROM  sys.sysindexes

-- SELECT * FROM sys.stats_columns
-- ====================================================================================
-- 选择度过低而无法有效利用索引查找而选择表扫描，查询2004-7-01之后的订单
-- ====================================================================================
USE  AdventureWorks;
GO
BEGIN   TRANSACTION
GO
DROP   INDEX  IX_OrderDate  ON  Sales.SalesOrderHeader
GO
CREATE   NONCLUSTERED   INDEX  IX_OrderDate  ON  Sales.SalesOrderHeader
    (
    OrderDate
    )  WITH ( STATISTICS_NORECOMPUTE  =   OFF , IGNORE_DUP_KEY  =   OFF , ALLOW_ROW_LOCKS  =   ON , ALLOW_PAGE_LOCKS  =   ON )  ON   [ PRIMARY ]
GO
COMMIT
DBCC  DROPCLEANBUFFERS;
GO
DBCC  FREEPROCCACHE;
GO
-- 分别使用2004-7-01低选择度与2004-7-26高选择度比较两种查询的性能差异
SELECT  SalesOrderID,OrderDate,Status,CustomerID,TaxAmt,TotalDue
FROM  Sales.SalesOrderHeader
WHERE  OrderDate > ' 2004-7-01 '

-- 首先在非聚焦索引叶级中扫描获得最大及最小的订单号，由于此时表扫描只是在叶级所以I/O操作会明显减少
DECLARE   @min   int ,
         @max   int
SELECT   @min = MIN (SalesOrderID), @max = MAX (SalesOrderID) 
FROM  Sales.SalesOrderHeader
WHERE  OrderDate > ' 2004-7-01 '
-- 使用聚焦索引
SELECT  SalesOrderID,OrderDate,Status,CustomerID,TaxAmt,TotalDue
FROM  Sales.SalesOrderHeader
WHERE  SalesOrderID  BETWEEN   @min   AND   @max
GO
-- ================================================
-- 什么是查询参数规范SARGS
-- ================================================
-- 1、不要对字段做运算
SELECT   *   FROM  Sales.SalesOrderHeader  WHERE  CustomerID + 1 = 2
SELECT   *   FROM  Sales.SalesOrderHeader  WHERE  CustomerID = 1
-- 2、不要对字段使用函数(数学函数、日期函数、字符串函数等)
SELECT   *   FROM  Sales.SalesOrderHeader  WHERE   ABS (SalesOrderID - 44659 ) < 1
SELECT   *   FROM  Sales.SalesOrderHeader  WHERE  SalesOrderID < 44660   AND  SalesOrderID > 44658
-- 3、不要使用负向查询(NOT、!=、<>、!>、!<、NOT EXISTS、NOT IN)
SELECT   *   FROM  Sales.SalesOrderDetail  WHERE  OrderQty != 1
SELECT   *   FROM  Sales.SalesOrderDetail  WHERE  OrderQty > 1   OR  OrderQty < 1
-- 4、小心使用OR(虽然有聚集索引但是因Status无索引可用只能进行表扫描)
SELECT   *   FROM  Sales.SalesOrderHeader  WHERE  SalesOrderID = 43659   OR   [ Status ] = 1

-- ================================================================================
-- 三种联结(嵌套循环、合并联结、哈稀联结)，所占用的资源依次增加，SQL优先选用嵌套循环
-- ================================================================================
-- 1、嵌套循环(外表、内表)，记录少的表作为外表，在内表中根据每个联结字段的值进行
-- 循环，只有在外表数据量较少时使用
SELECT   *
FROM  Sales.SalesOrderDetail D  JOIN  Sales.SalesOrderHeader O
     ON  D.SalesOrderID  =  O.SalesOrderID
WHERE  O.SalesOrderID  BETWEEN   43659   AND   43665
-- 2、合并联结的算法前提是字段已排序
SELECT   *
FROM  Sales.SalesOrderDetail D  JOIN  Sales.SalesOrderHeader O
     ON  D.SalesOrderID  =  O.SalesOrderID
WHERE  O.SalesOrderID  BETWEEN   43659   AND   54000
-- 3、哈稀联结(在要联结的字段上无可用索引时使用此连接算法)
IF   OBJECT_ID (N ' SumPrice ' ,N ' U ' )  IS   NOT   NULL
     DROP   TABLE  SumPrice
GO
SELECT  O.SalesOrderID, YEAR (O.OrderDate) + MONTH (O.OrderDate)  AS  Date, CAST ( SUM (D.OrderQty * D.UnitPrice * ( 1 - D.UnitPriceDiscount))  AS   decimal ( 18 , 2 ))  AS  Price
INTO  SumPrice
FROM  
    Sales.SalesOrderDetail D 
         JOIN  
    Sales.SalesOrderHeader O
         ON  D.SalesOrderID = O.SalesOrderID
GROUP   BY  O.SalesOrderID,O.OrderDate
GO
SELECT  O.SalesOrderID,O.
FROM  SumPrice D  JOIN  Sales.SalesOrderHeader O
     ON  D.SalesOrderID  =  O.SalesOrderID
WHERE  O.SalesOrderID  BETWEEN   43659   AND   54000

-- ==========================================================================================
-- 代替LEFT JOIN的方法
-- 为了防止在查询中出现LEFT操作，在设计数据库时应尽量避免关联表中不一致的情况出现
-- 比如产品的类别和具体产品的对应关系中，可在类别加入一个无类别的记录以对应不知是何类别的产品
-- ==========================================================================================
BEGIN   TRANSACTION
GO
DROP   INDEX  IX_SalesOrderHeader_CustomerID  ON  Sales.SalesOrderHeader
GO
CREATE   NONCLUSTERED   INDEX  IX_SalesOrderHeader_CustomerID  ON  Sales.SalesOrderHeader
    (
    CustomerID
    )  WITH ( STATISTICS_NORECOMPUTE  =   OFF , IGNORE_DUP_KEY  =   OFF , ALLOW_ROW_LOCKS  =   ON , ALLOW_PAGE_LOCKS  =   ON )  ON   [ PRIMARY ]
GO
COMMIT
DBCC  FLUSHPROCINDB( 8 );
GO
DBCC  DROPCLEANBUFFERS;
GO
-- 1、查询所有客户的总订单数量
SELECT  C.CustomerID, COALESCE ( SUM (OrderQty), 0 )  AS  Quantity
FROM  Sales.Customer C  LEFT   JOIN             
    (    Sales.SalesOrderHeader O 
         JOIN  Sales.SalesOrderDetail D  ON  O.SalesOrderID = D.SalesOrderID
    )  ON  C.CustomerID = O.CustomerID
GROUP   BY  C.CustomerID
GO
-- 2、替代的查询方式
DECLARE   @CustomerInfo   table
(
    CustomerID     int              not   null ,
    Quantity     smallint      not   null
)
INSERT   INTO   @CustomerInfo
SELECT  CustomerID, 0
FROM  Sales.Customer;

WITH  SumQuantity  AS
(
     SELECT  O.CustomerID, COALESCE ( SUM (OrderQty), 0 )  AS  Quantity
     FROM  Sales.SalesOrderHeader O 
     JOIN  Sales.SalesOrderDetail D  ON  O.SalesOrderID = D.SalesOrderID
     GROUP   BY  O.CustomerID
)
UPDATE   @CustomerInfo
SET  C.Quantity = S.Quantity
FROM   @CustomerInfo  C  JOIN  SumQuantity S
     ON  S.CustomerID = C.CustomerID;

SELECT   *   FROM   @CustomerInfo ;
GO

-- ==========================================================================================
-- 控制表之间的联结
-- ==========================================================================================
USE  Northwind;
GO
SELECT   DISTINCT  C.CompanyName  AS  customer, S.CompanyName  AS  supplier
FROM  dbo.Customers  AS  C
   LEFT   OUTER   JOIN  dbo.Orders  AS  O
     ON  O.CustomerID  =  C.CustomerID
   LEFT   OUTER   JOIN  dbo. [ Order Details ]   AS  OD
     ON  OD.OrderID  =  O.OrderID
   LEFT   OUTER   JOIN  dbo.Products  AS  P
     ON  P.ProductID  =  OD.ProductID
   LEFT   OUTER   JOIN  dbo.Suppliers  AS  S
     ON  S.SupplierID  =  P.SupplierID
-- OPTION(FORCE ORDER);
-- 上面的查询过多的使用外联结，可使用下面的方式替换
SELECT   DISTINCT  C.CompanyName  AS  customer, S.CompanyName  AS  supplier
FROM  dbo.Orders  AS  O
   JOIN  dbo. [ Order Details ]   AS  OD
     ON  OD.OrderID  =  O.OrderID
   JOIN  dbo.Products  AS  P
     ON  P.ProductID  =  OD.ProductID
   JOIN  dbo.Suppliers  AS  S
     ON  S.SupplierID  =  P.SupplierID
   RIGHT   OUTER   JOIN  dbo.Customers  AS  C
     ON  O.CustomerID  =  C.CustomerID;

SELECT   DISTINCT  C.CompanyName  AS  customer, S.CompanyName  AS  supplier
FROM  dbo.Customers  AS  C
   LEFT   OUTER   JOIN
    (dbo.Orders  AS  O
        JOIN  dbo. [ Order Details ]   AS  OD
          ON  OD.OrderID  =  O.OrderID
        JOIN  dbo.Products  AS  P
          ON  P.ProductID  =  OD.ProductID
        JOIN  dbo.Suppliers  AS  S
          ON  S.SupplierID  =  P.SupplierID)
       ON  O.CustomerID  =  C.CustomerID;


-- ============================================================
-- 存储过程相关
-- ============================================================
USE  Northwind;
GO
-- ***********************************************************
-- 1、计划缓存重用造成的无效查询计划
-- 造成上述问题的原因是因为SQLServer为了节省编译的开销会优先考虑使用缓存的计划
-- 2000下只能使用过程级重新编译，2005下提供了语句级重新编译的特性。从而避免了编译
-- 整个过程的开销
IF   OBJECT_ID (N ' dbo.usp_GetOrders ' ,N ' P ' )  IS   NOT   NULL
   DROP   PROC  dbo.usp_GetOrders;
GO
CREATE   PROC  dbo.usp_GetOrders
   @odate   AS   DATETIME
AS

SELECT  c.CustomerID, c.CompanyName,
    o.EmployeeID, o.OrderID, o.OrderDate
FROM  dbo.Orders o
     join  dbo.Customers c
         on  o.CustomerID  =  c.CustomerID
WHERE  OrderDate  >=   @odate ;
-- OPTION(RECOMPILE);
GO
-- 选择度高的测试
EXEC  dbo.usp_GetOrders  ' 19980506 ' ;
-- 选择度低的测试
EXEC  dbo.usp_GetOrders  ' 19960101 ' ;
-- 查看缓存计划
SELECT   *
FROM  sys.syscacheobjects
WHERE  sql  NOT   LIKE   ' %cache% '
   AND  sql  LIKE   ' %usp_GetOrders% ' ;
-- 清除缓存计划
DBCC  FREEPROCCACHE;
GO
-- ***********************************************************
-- 2、参数嗅探测
-- 以当前日期插入一条记录，高选择度查询
INSERT   INTO  dbo.Orders(OrderDate, CustomerID, EmployeeID)  VALUES ( GETDATE (), N ' ALFKI ' ,  1 );
GO
ALTER   PROC  dbo.usp_GetOrders
   @d   AS   INT   =   0
AS
DECLARE   @odate   AS   DATETIME ;
SET   @odate   =   DATEADD ( day ,  - @d ,  CONVERT ( VARCHAR ( 8 ),  GETDATE (),  112 ));

SELECT  OrderID, CustomerID, EmployeeID, OrderDate
FROM  dbo.Orders
WHERE  OrderDate  >=   @odate ;
GO
-- 查看表dbo.Orders空间情况
SELECT   *   FROM  
sys.dm_db_index_physical_stats( DB_ID (N ' Northwind ' ),  OBJECT_ID (N ' dbo.Orders ' ),  NULL ,  NULL  ,  ' DETAILED ' );
-- 此时虽然选择度高，但是因为在过程编译时优化器无法根据参数进行最优化
EXEC  dbo.usp_GetOrders;

-- 使用聚集索引代替非聚集索引进行查询的参数嗅探测试
USE  AdventureWorks;
GO
IF   OBJECT_ID (N ' dbo.GetCutomerOrderByDate ' ,N ' P ' )  IS   NOT   NULL
   DROP   PROC  dbo.GetCutomerOrderByDate;
GO
CREATE   PROCEDURE  dbo.GetCutomerOrderByDate
     @start   datetime
AS
BEGIN
     SET  NOCOUNT  ON ;

     DECLARE   @min   int ,
         @max   int
     SELECT   @min = MIN (SalesOrderID), @max = MAX (SalesOrderID) 
     FROM  Sales.SalesOrderHeader
     WHERE  OrderDate > @start
     SELECT  SalesOrderID,OrderDate,Status,CustomerID,TaxAmt,TotalDue
     FROM  Sales.SalesOrderHeader
     WHERE  SalesOrderID  BETWEEN   @min   AND   @max
END
GO
DBCC  FREEPROCCACHE;
GO
-- 低选择度测试
EXEC  dbo.GetCutomerOrderByDate  ' 2004-5-01 '
-- 高选择度测试
EXEC  dbo.GetCutomerOrderByDate  ' 2004-7-26 '
-- 缓存计划
SELECT   *
FROM  sys.syscacheobjects
WHERE  sql  NOT   LIKE   ' %cache% '
   AND  sql  LIKE   ' %GetCutomerOrderByDate% ' ;

-- ***********************************************************
-- 3、Execute/sp_executesql
USE  Northwind;
GO
DBCC  FREEPROCCACHE;
GO
--
DECLARE   @i          AS   INT ;
DECLARE   @sql      AS   VARCHAR ( 52 );
SET   @i   =   10248 ;
WHILE   @i < 10252
BEGIN
     SET   @sql   =   ' SELECT * FROM dbo.Orders WHERE OrderID =  '
       +   CAST ( @i   AS   VARCHAR ( 10 ))  +  N ' ; ' ;
     EXEC ( @sql );
     SET   @i = @i + 1
END
-- 查看缓存计划，可以看到除了参数化计划外还为每个不同的值保存一个缓存计划
SELECT  cacheobjtype, objtype, usecounts, sql
FROM  sys.syscacheobjects
WHERE  sql  NOT   LIKE   ' %cache% '
   AND  sql  NOT   LIKE   ' %sys.% ' ;
-- 存储过程中调用同样存在上面的问题，为每个参数分配一个缓存计划
IF   OBJECT_ID (N ' dbo.GetOrderByID ' ,N ' P ' )  IS   NOT   NULL
   DROP   PROC  dbo.GetOrderByID;
GO
CREATE   PROCEDURE  dbo.GetOrderByID
@i   AS   INT
AS
BEGIN
     SET  NOCOUNT  ON ;
     DECLARE   @sql   AS   VARCHAR ( 52 );
     SET   @sql   =   ' SELECT * FROM dbo.Orders WHERE OrderID =  '
       +   CAST ( @i   AS   VARCHAR ( 10 ))  +  N ' ; ' ;
     EXEC ( @sql );
END
GO
DECLARE   @i          AS   INT ;
SET   @i   =   10248 ;
WHILE   @i < 10252
BEGIN
     EXEC  dbo.GetOrderByID  @i
     SET   @i = @i + 1
END
GO
ALTER   PROCEDURE  dbo.GetOrderByID
@i   AS   INT
AS
BEGIN
     SET  NOCOUNT  ON ;
     DECLARE   @sql   AS   NVARCHAR ( 52 );
     SET   @sql   =  N ' SELECT * FROM dbo.Orders WHERE OrderID = @ID '
     EXEC  sp_executesql
         @stmt = @sql ,
         @params = N ' @ID as int ' ,
         @ID = @i
END
GO
DBCC  FREEPROCCACHE;
GO
DECLARE   @i          AS   INT ;
SET   @i   =   10248 ;
WHILE   @i < 10252
BEGIN
     EXEC  dbo.GetOrderByID  @i
     SET   @i = @i + 1
END
-- 动态语句的选择度测试
IF   OBJECT_ID (N ' dbo.usp_GetOrders ' ,N ' P ' )  IS   NOT   NULL
   DROP   PROC  dbo.usp_GetOrders;
GO

CREATE   PROC  dbo.usp_GetOrders
   @odate   AS   DATETIME
AS
SET  NOCOUNT  ON ;
     DECLARE   @sql   AS   NVARCHAR ( 100 );
     SET   @sql = N ' SELECT OrderID, CustomerID, EmployeeID, OrderDate
            FROM dbo.Orders
            WHERE OrderDate >= @odate OPTION(RECOMPILE); '
     EXEC  sp_executesql
         @stmt = @sql ,
         @params = N ' @odate as datetime ' ,
         @odate = @odate
GO
SELECT  OrderID, CustomerID, EmployeeID, OrderDate
             FROM  dbo.Orders
             WHERE  OrderDate  >=   ' 19980506 '   option (recompile)
-- 选择度高的测试
EXEC  dbo.usp_GetOrders  ' 19980506 ' ;
-- 选择度低的测试
EXEC  dbo.usp_GetOrders  ' 19960101 ' ;
-- 查看缓存计划
SELECT   *
FROM  sys.syscacheobjects
WHERE  sql  NOT   LIKE   ' %cache% '
   AND  sql  LIKE   ' %usp_GetOrders% ' ;
-- 清除缓存计划
DBCC  FREEPROCCACHE;
GO

-- 动态语句EXEC
IF   OBJECT_ID (N ' dbo.usp_GetOrders ' ,N ' P ' )  IS   NOT   NULL
   DROP   PROC  dbo.usp_GetOrders;
GO
CREATE   PROC  dbo.usp_GetOrders
   @odate   AS   varchar ( 10 )
AS
SET  NOCOUNT  ON ;
     DECLARE   @sql   AS   VARCHAR ( 120 );
     SET   @sql = ' SELECT OrderID, CustomerID, EmployeeID, OrderDate
            FROM dbo.Orders
            WHERE OrderDate >= ''' + @odate + ''' ; '
     print   @sql
     EXEC ( @sql )
-- OPTION(RECOMPILE);
GO
SELECT  OrderID, CustomerID, EmployeeID, OrderDate
             FROM  dbo.Orders
             WHERE  OrderDate  >= ' 19980506 '
select   convert ( datetime , ' 19980506 ' )
-- 选择度高的测试
EXEC  dbo.usp_GetOrders  ' 19980506 ' ;
-- 选择度低的测试
EXEC  dbo.usp_GetOrders  ' 19960101 ' ;
-- 查看缓存计划
SELECT   *
FROM  sys.syscacheobjects
WHERE  sql  NOT   LIKE   ' %cache% '
   AND  sql  LIKE   ' %usp_GetOrders% ' ;
-- 清除缓存计划
DBCC  FREEPROCCACHE;
GO
-- ============================================================
-- 日志相关
-- ============================================================
DBCC  LOGINFO;
GO
DBCC  SQLPERF(LOGSPACE);
GO
-- 2005
WHILE   1   =   1
BEGIN
   UPDATE   TOP ( 5000 ) dbo.A
     SET  CustomerID  =  N ' ABCDE '
   WHERE  CustomerID  =  N ' OLDWO ' ;

   IF   @@rowcount   <   5000   BREAK ;
END
GO
-- 2000
SET   ROWCOUNT   5000 ;
WHILE   1   =   1
BEGIN
   DELETE   FROM  dbo.LargeOrders
   WHERE  OrderDate  <   ' 19970101 ' ;

   IF   @@rowcount   <   5000   BREAK ;
END
SET   ROWCOUNT   0 ;