SQL Server 中Inner join 和where的效率差異
今天,手頭上正在作的一個項目,在生成報表時,客戶感覺太慢,于是,各處檢查,看可否提示效率。界面上的都改進了,提升不大。如是在SQL 語句上下功夫。(我這人比較懶,對簡單的語句和查詢都沒有經過仔細優化的,一般只對姚使用left join,outer join,group by 以及carsor的語句會仔細寫并用數據庫理論考慮和檢查---因為這種語句一般測試時如果發現錯誤,檢查和調試很麻煩)
先在網上Google搜索“Join 與 Where 效率”以及察看SQL Server 幫助文檔,希望能獲得“捷徑”些的優化思路。
搜索的結果是,各大論壇,包括MSDN上很多人提出了這個問題,但回答是眾說紛紜。總體上總結出來時說:對小數據量(<N萬)的來說效率幾乎無差異,更有說法說Inner join 和Where只是SQL標準不同,在查詢分析器中SQL Server查詢分析器是將Where直接轉換為Join后查詢的。
還是自己來做試驗吧。
如是有了如下比較結果(均在查詢分析器中查詢和計時):
語句(1)declare @operatorName nvarchar(50)set @operatorName = '%'
select distinct item.*; from item , customer_item , customer_operator ,operatorwhere item.itemcode = customer_item.itemCode and customer_item.customerCode =; customer_operator.customerCodeand customer_operator.operatorId =; customer_operator.operatorIdand operator.operatorName like @operatorNameand item.deleted = 0 and customer_item.deleted = 0 and customer_operator.deleted = 0查詢結果,74行,共時間0:00:04
語句(2)declare @operatorName nvarchar(50)
set @operatorName = '%'
select distinct item.*; from item inner join; customer_item on; item.itemcode = customer_item.itemCode inner join customer_operator on customer_item.customerCode = customer_operator.customerCodeinner join operator on customer_operator.operatorId = operator.operatorIdwhere; operator.operatorName like @operatorNameand item.deleted = 0 and customer_item.deleted = 0 and customer_operator.deleted = 0共74行,時間0:00:01
后檢查發現語句(1)中有一個重復自查詢條件 :customer_operator.operatorId =; customer_operator.operatorId將其葉加到語句2中,語句(3)declare @operatorName nvarchar(50)
set @operatorName = '%'
select distinct item.*; from item inner join; customer_item on; item.itemcode = customer_item.itemCode inner join customer_operator on customer_item.customerCode = customer_operator.customerCodeinner join operator on customer_operator.operatorId = operator.operatorIdwhere; operator.operatorName like @operatorNameand item.deleted = 0 and customer_item.deleted = 0 and customer_operator.deleted = 0and customer_operator.operatorId =; customer_operator.operatorId
所用時間和結果都為74行,時間0:00:01。
將語句(1)中的去掉該條件后成為語句(4)declare @operatorName nvarchar(50)set @operatorName = '%'
select distinct item.*; from item , customer_item , customer_operator ,operatorwhere item.itemcode = customer_item.itemCode and customer_item.customerCode =; customer_operator.customerCode--and customer_operator.operatorId =; customer_operator.operatorIdand operator.operatorName like @operatorNameand item.deleted = 0 and customer_item.deleted = 0 and customer_operator.deleted = 0
時間和結果為74行,時間0:00:01。
終于發現了些他們的差異。
結論: 盡量使用Join 而不是Where來列出關聯條件,特別是多個表聯合的時候。原因是: (1)在效率上,Where可能具有和Inner join一樣的效率。但基本可以肯定的(通過SQLServer幫助和其它資料,以及本測試)是Join的效率不比Where差。 (2)使用Join可以幫助檢查語句中的無效或者誤寫的關聯條件
