SQL技巧总结

1. Extracting query executor plan from Procedure Cache

代码

WITH  XMLNAMESPACES ( ' http://schemas.microsoft.com/sqlserver/2004/07/showplan '   AS  sql)
SELECT   top   100
  C.value( ' @StatementId ' , ' INT ' )  AS   [ No ] ,
  C.value( ' (./@StatementText) ' , ' NVARCHAR(MAX) ' )  AS   [ Statement Text ] ,
  qplan.query_plan  AS   [ Query Plan ]
FROM  ( SELECT   DISTINCT  plan_handle  FROM  sys.dm_exec_query_stats)  AS  qstats
   CROSS  APPLY sys.dm_exec_query_plan(qstats.plan_handle)  AS  qplan
   CROSS  APPLY query_plan.nodes( ' /sql:ShowPlanXML/sql:BatchSequence/sql:Batch/sql:Statements/
descendant::*[attribute::StatementText] ' )
     AS  T(C)
ORDER   BY  plan_handle,  [ No ] ;

2. Get Index States:

代码

-- -查看数据库中缺失的索引
select   distinct   DB_NAME (database_id)  as  DataBaseName,
       statement  as  table_name,
       equality_columns,
       inequality_columns,
       included_columns
from  sys.dm_db_missing_index_details
where   DB_NAME (database_id) = ' ASGDM '
order   by  statement,equality_columns,included_columns


-- -查看数据库中表索引使用情况
select   DB_Name (database_id)  as  dbname,c.name  as  tablename,b.name  as  indexname,
       a.user_seeks,a.user_scans,a.user_lookups,
       a.last_user_seek,a.last_user_scan,a.last_user_lookup
from  sys.dm_db_index_usage_stats a
inner   join  sys.indexes b  on  a.index_id = b.index_id  and  a. object_id = b. object_id
inner   join  sys.objects c  on  a. object_id = c. object_id
where   DB_Name (database_id) = ' ASGDM '   and  b.index_id >= 1
order   by  c.name

3. SMSS启用丢弃结果：

4. 修改数据库的兼容级别
exec sp_dbcmptlevel dbname, dbversion;
5. 1.*impression/click --> int 转化为 float

6. select into 快于 insert select；left join is null  = not exists

7.  控制执行顺序：拆分、括号、Hint

8. 递归CTE：对CTE名称的内部引用代表“前一个结果集”

9.  实际执行计划也不一定可靠
    执行计划不会捕捉下面语句的执行    set @comm = 'select ......'    insert into     exec(@comm)

　 执行计划的显示的percentage有时和实际不相符，最好还是利用Set statistics time on来获取执行时间

10. 最先应用最有效的过滤/连接

11. TABLESAMPLE select * from orders TABLESAMPLE (1000 rows)

12. HINT
   with index     

   option( use plan ...

13. Except和Not Exist对Null值处理逻辑不同；

14. Order By应用Case：

Order   by  col1,
case   when  col1  =   1   then  orderID  end ,
case   when  col1  =   2   then  orderdate  end   desc ;

15. Insert/Update/Delete Output

代码

-- Insert
INSERT   INTO  dbo.CustomersDim(CustomerID, CompanyName)
    OUTPUT inserted.CustomerID, inserted.KeyCol
     INTO   @NewCusts
     --  OUTPUT inserted.CustomerID, inserted.KeyCol
   SELECT  CustomerID, CompanyName
   FROM  Northwind.dbo.Customers
   WHERE  Country  =  N ' UK ' ;

SELECT  CustomerID, KeyCol  FROM   @NewCusts ;

-- Update
UPDATE   TOP ( @n ) dbo.Messages  WITH (READPAST)  SET  status  =   ' open '
    OUTPUT inserted.msgid, inserted.msgdate, inserted.msg  INTO   @Msgs
    OUTPUT inserted.msgid, inserted.msgdate, inserted.msg
   WHERE  status  =   ' new ' ;

-- Delete
DELETE   TOP ( 5000 )  FROM  dbo.LargeOrders
      OUTPUT deleted. *   INTO  dbo.OrdersArchive
     WHERE  OrderDate  <   ' 20010101 ' ;

16. Delete Top --一次删除太多记录比较慢，使用DELETE TOP可以一部分一部分删除

代码

WHILE   1 = 1
BEGIN
   BEGIN   TRAN
     DELETE   TOP ( 5000 )  FROM  dbo.LargeOrders
      OUTPUT deleted. *   INTO  dbo.OrdersArchive
     WHERE  OrderDate  <   ' 20010101 ' ;
     IF   @@rowcount   <   5000
     BEGIN
       COMMIT   TRAN
       BREAK ;
     END
   COMMIT   TRAN
END

17. CHECKSUM函数。如果我们需要在一个较大的字符串列上建立索引，我们可以使用CHECKSUM函数。CHECKSUM可以根据字符串产生一个4字节的整数，我们可以把索引建立在CHECKSUM的计算列上。当然，CHECKSUM并不能保证完全唯一，所以在查询时需要用到字符串列；

18. 最好把最具选择性的列放到索引最前面。