在數(shù)據(jù)倉庫中經(jīng)常查詢的SQL總帶有下列特征:

◆幾個表進(jìn)行關(guān)聯(lián)

◆只有一個數(shù)據(jù)量巨大的表, 稱為事實表

◆其他的都是編碼表, 稱為維表

◆維表和事實表之間有主外鍵關(guān)系

假設(shè)有D1(key1),D2(key2),D3(key3),D4(key)四個小的維表和一個事實表F(key1,key2,key3,key4), 那么經(jīng)常進(jìn)行的查詢將是:

SELECT

D1.xxx, D2.xxx, D3.xxx, D4.xxx,

SUM(F.xxx), SUM(F.xxx)

FROM F, D1, D2, D3, D4

WHERE F.KEY1=D1.KEY1 AND F.KEY2=D2.KEY2

AND F.KEY3=D3.KEY3 AND F.KEY4=D4.KEY4

AND D1.xxx=? AND D2.xxx=?

AND D3.xxx=? AND D4.xxx=?

GROUP BY D1.xxx, D2.xxx, D3.xxx, D4.xxx

而為了提高查詢速度, 根據(jù)數(shù)據(jù)特征, key1,key2,key3,key4這四個字段會比較適合每個字段上建一個位圖索引(Bitmap Index), 但是上面的查詢語句并不能用到位圖索引, 除非是進(jìn)行了星型轉(zhuǎn)換, 這個轉(zhuǎn)換需要將SQL轉(zhuǎn)換成下面的格式:

SELECT

D1.xxx, D2.xxx, D3.xxx, D4.xxx,

SUM(F.xxx), SUM(F.xxx)

FROM F, D1, D2, D3, D4

WHERE F.KEY1=D1.KEY1 AND F.KEY2=D2.KEY2

AND F.KEY3=D3.KEY3 AND F.KEY4=D4.KEY4

AND D1.xxx=? AND D2.xxx=?

AND D3.xxx=? AND D4.xxx=?

AND F.KEY1 IN (SELECT D1.KEY1 FROM D1 WHERE D1.xxx=?)

AND F.KEY2 IN (SELECT D2.KEY2 FROM D2 WHERE D2.xxx=?)

AND F.KEY3 IN (SELECT D3.KEY3 FROM D3 WHERE D3.xxx=?)

AND F.KEY4 IN (SELECT D4.KEY4 FROM D4 WHERE D4.xxx=?)

GROUP BY D1.xxx, D2.xxx, D3.xxx, D4.xxx

有條件的可以試試, 使用星型轉(zhuǎn)換有三個條件:

◆事實表的每個代碼列上有位圖索引

◆參數(shù)STAR_TRANSFORMATION_ENABLED的值要設(shè)為TRUE

◆必須使用CBO, 所以最好對表進(jìn)行適當(dāng)?shù)姆治?/P>