我們有時寫的sql 會多次查詢和使用相同的結果集，而事實上每次查詢都會消耗資源和降低sql的整體查詢效率，特別是對大量數據表，耗時特別長。
使用with as 將想要查詢的數據集保存到一張虛表中，數據查詢均從這張虛擬表（視圖）中查詢獲得，減少對數據庫的直接訪問。
1、該sql性能太低，其中有兩張表（數據量均比較大）數據拼接操作，沒有先過濾掉不需要的數據，全表掃描兩張大表然後拼接，然後過濾
SELECT A.CUST_ID,
A.CUST_CODE,
A.CUST_NAME,
A.CREATED_DATE,
A.CUST_TYPE,
C.CERT_TYPE_NAME,
B.CERT_NBR,
D.STATE,
D.STATE_DATE,
E.MDN_STATE MSISDN_STATE
FROM CUST A,
CERT B,
CERT_TYPE C,
ECONET_CUST_STATE D,
(SELECT G.CUST_ID, WM_CONCAT(G.MDN_STATE) MDN_STATE
FROM (SELECT *
FROM (SELECT F.CUST_ID,
F.MDN_STATE,
ROW_NUMBER() OVER(PARTITION BY F.CUST_ID ORDER BY F.MDN_STATE) TOP
FROM (SELECT DISTINCT S.CUST_ID,
S.PREFIX || S.ACC_NBR || '|' ||
P.PROD_STATE MDN_STATE
FROM SUBS S, PROD P
WHERE S.SUBS_ID = P.PROD_ID) F)
WHERE TOP <= 5) G
WHERE G.CUST_ID IS NOT NULL
GROUP BY G.CUST_ID) E
WHERE 1 = 1
AND A.CUST_ID = D.CUST_ID
AND A.CERT_ID = B.CERT_ID(+)
AND B.CERT_TYPE_ID = C.CERT_TYPE_ID(+)
AND A.CUST_ID = E.CUST_ID(+)
AND D.STATE IN ('F', 'A', 'D')
2、第一次優化，先過濾掉數據然後拼接，然而部分數據由於拼接元素爲null,通過條件過濾的話會漏掉，因此要單獨查詢 然後union（性能有所提高）
SELECT G.CUST_ID,
G.CUST_CODE,
G.CUST_NAME,G.CREATED_DATE,G.CUST_TYPE,G.CERT_TYPE_NAME,
G.CERT_NBR,G.STATE,G.STATE_DATE, WM_CONCAT(G.MDN_STATE) MDN_STATE
FROM (SELECT *
FROM (SELECT F.CUST_ID,
F.CUST_CODE,
F.CUST_NAME,F.CREATED_DATE,F.CUST_TYPE,F.CERT_TYPE_NAME,
F.CERT_NBR,F.STATE,F.STATE_DATE,
F.MDN_STATE,
ROW_NUMBER() OVER(PARTITION BY F.CUST_ID ORDER BY F.MDN_STATE) TOP
FROM ( SELECT A.CUST_ID,
A.CUST_CODE,
A.CUST_NAME,
A.CREATED_DATE,
A.CUST_TYPE,
C.CERT_TYPE_NAME,
B.CERT_NBR,
D.STATE,
D.STATE_DATE,S.PREFIX || S.ACC_NBR || '|' ||P.PROD_STATE MDN_STATE
FROM SUBS S, PROD P, CUST A,CERT B,CERT_TYPE C, ECONET_CUST_STATE D
WHERE S.SUBS_ID = P.PROD_ID
AND B.CERT_TYPE_ID = C.CERT_TYPE_ID(+)
AND A.CERT_ID = B.CERT_ID(+)
AND A.CUST_ID = S.CUST_ID(+)
AND A.CUST_ID = D.CUST_ID
AND D.STATE IN ('F', 'A', 'D')) F)
WHERE TOP <= 5) G
WHERE G.CUST_ID IS NOT NULL
GROUP BY G.CUST_ID,
G.CUST_CODE,
G.CUST_NAME,G.CREATED_DATE,G.CUST_TYPE,G.CERT_TYPE_NAME,
G.CERT_NBR,G.STATE,G.STATE_DATE
UNION ALL
SELECT A.CUST_ID,
A.CUST_CODE,
A.CUST_NAME,
A.CREATED_DATE,
A.CUST_TYPE,
C.CERT_TYPE_NAME,
B.CERT_NBR,
D.STATE,
D.STATE_DATE,WM_CONCAT(S.CUST_ID) MDN_STATE
FROM CUST A,CERT B,CERT_TYPE C, ECONET_CUST_STATE D,SUBS S
WHERE 1=1
AND B.CERT_TYPE_ID = C.CERT_TYPE_ID(+)
AND A.CERT_ID = B.CERT_ID(+)
AND A.CUST_ID = D.CUST_ID
AND D.CUST_ID = S.CUST_ID(+)
AND D.STATE IN ('F', 'A', 'D')
AND S.CUST_ID IS NULL
GROUP BY A.CUST_ID,
A.CUST_CODE,
A.CUST_NAME,
A.CREATED_DATE,
A.CUST_TYPE,
C.CERT_TYPE_NAME,
B.CERT_NBR,
D.STATE,
D.STATE_DATE
3、第二次優化（第一次優化 使用了union all,其中SUBS S,CUST A,CERT B,CERT_TYPE C, ECONET_CUST_STATE D 這些個關聯查詢在union all 兩邊 都做了，如果只查詢一次的話，性能會更好點，因此抽出來，用with as 先做一次查詢 保存SCCCE 這張虛表中 這樣只會對subs 這張數據量很大的表做一次全表掃描）
WITH SCCCE AS (SELECT A.CUST_ID,S.SUBS_ID,
A.CUST_CODE,
A.CUST_NAME,
A.CREATED_DATE,
A.CUST_TYPE,
C.CERT_TYPE_NAME,
B.CERT_NBR,
D.STATE,
D.STATE_DATE,
S.PREFIX || S.ACC_NBR MISDN
FROM SUBS S,CUST A,CERT B,CERT_TYPE C, ECONET_CUST_STATE D
WHERE B.CERT_TYPE_ID = C.CERT_TYPE_ID(+)
AND A.CERT_ID = B.CERT_ID(+)
AND A.CUST_ID = S.CUST_ID(+)
AND A.CUST_ID = D.CUST_ID
AND D.STATE IN ('F', 'A', 'D'))
SELECT G.CUST_ID,
G.CUST_CODE,
G.CUST_NAME,G.CREATED_DATE,G.CUST_TYPE,G.CERT_TYPE_NAME,
G.CERT_NBR,G.STATE,G.STATE_DATE, WM_CONCAT(G.MDN_STATE) MDN_STATE
FROM (SELECT *
FROM (SELECT F.CUST_ID,
F.CUST_CODE,
F.CUST_NAME,F.CREATED_DATE,F.CUST_TYPE,F.CERT_TYPE_NAME,
F.CERT_NBR,F.STATE,F.STATE_DATE,
F.MDN_STATE,
ROW_NUMBER() OVER(PARTITION BY F.CUST_ID ORDER BY F.MDN_STATE) TOP
FROM (SELECT SC.CUST_ID,
SC.CUST_CODE,
SC.CUST_NAME,
SC.CREATED_DATE,
SC.CUST_TYPE,
SC.CERT_TYPE_NAME,
SC.CERT_NBR,
SC.STATE,
SC.STATE_DATE,
SC.MISDN|| '|' ||P.PROD_STATE MDN_STATE
FROM PROD P,SCCCE SC
WHERE SC.SUBS_ID = P.PROD_ID)F)
WHERE TOP <= 5) G
WHERE G.CUST_ID IS NOT NULL
GROUP BY G.CUST_ID,
G.CUST_CODE,
G.CUST_NAME,G.CREATED_DATE,G.CUST_TYPE,G.CERT_TYPE_NAME,
G.CERT_NBR,G.STATE,G.STATE_DATE
UNION ALL
SELECT SC.CUST_ID,
SC.CUST_CODE,
SC.CUST_NAME,
SC.CREATED_DATE,
SC.CUST_TYPE,
SC.CERT_TYPE_NAME,
SC.CERT_NBR,
SC.STATE,
SC.STATE_DATE,WM_CONCAT( SC.MISDN) MDN_STATE
FROM SCCCE SC
WHERE 1=1
AND SC.SUBS_ID IS NULL
GROUP BY SC.CUST_ID,
SC.CUST_CODE,
SC.CUST_NAME,
SC.CREATED_DATE,
SC.CUST_TYPE,
SC.CERT_TYPE_NAME,
SC.CERT_NBR,
SC.STATE,
SC.STATE_DATE