總和結果是Oracle乘以行數

Question

我有一個名爲b_balance返回以下記錄視圖：

SECURITIES_CODE BUY_SELL_FLAG C_BALANCE P_BALANCE 
--------------- ------------- --------- --------- 
10042    BUY     200   0 
10042    BUY     500   0 
10042    SELL     200   0 
10042    BUY      0  5000 
10042    SELL     0  2500 
10043    BUY     300   0 
10043    SELL     0  2500 

並命名爲t_balance返回以下記錄另一個視圖：

SECURITIES_CODE BUY_SELL_FLAG C_BALANCE P_BALANCE 
--------------- ------------- --------- --------- 
10042    BUY      0  5000 
10043    BUY     300   0 
10042    SELL     200   0 
10042    SELL     0  2500 
10043    SELL     0  2500 
10042    BUY     200   0 
10042    BUY     500   0 

現在出現的問題，當我執行我的SQL

SELECT TO_CHAR(to_date('20170801','yyyyMMdd'), 'MM/dd/yyyy') AS TRADE_DATE, 
    b.securities_code           AS SECURITIES_CODE, 
    b.buy_sell_flag           AS SIDE, 
    SUM(NVL(t.c_balance,0))         AS C_t_balance, 
    SUM(NVL(b.c_balance,0))         AS C_b_balance, 
    SUM(NVL(t.c_balance,0)) - SUM(NVL(b.c_balance,0))  AS C_DIFFERENCE, 
    SUM(NVL(t.p_balance,0))         AS P_t_balance, 
    SUM(NVL(b.p_balance,0))         AS P_b_balance, 
    SUM(NVL(t.p_balance,0)) - SUM(NVL(b.p_balance,0))  AS P_DIFFERENCE 
FROM b_balance b 
FULL OUTER JOIN t_balance t 
ON b.securities_code = t.securities_code 
AND b.buy_sell_flag = t.buy_sell_flag 
GROUP BY b.securities_code, 
    b.buy_sell_flag 
ORDER BY SECURITIES_CODE, 
    SIDE ; 

此返回以下記錄：

TRADE_DATE SECURITIES_CODE SIDE C_T_BALANCE C_B_BALANCE C_DIFFERENCE P_T_BALANCE P_B_BALANCE P_DIFFERENCE 
---------- --------------- ---- ----------- ----------- ------------ ----------- ------------ ------------ 
08/01/2017 10042    BUY   2100   2100    0   15000   15000    0 
08/01/2017 10042    SELL   400   400    0   5000   5000    0 
08/01/2017 10043    BUY   300   300    0    0    0    0 
08/01/2017 10043    SELL    0    0    0   2500   2500    0 

這意味着結果被乘以行數。 我檢查堆棧溢出，並沒有發現任何錯誤根據this答案。

那麼我的SQL有什麼問題？

Answer 1

你應該加入彙總結果

SELECT TO_CHAR(to_date('20170801','yyyyMMdd'), 'MM/dd/yyyy') AS TRADE_DATE, 
    t1.SECURITIES_CODE, 
    t1.SIDE, 
    t1.C_b_balance, 
    t1.P_b_balance, 
    t2.C_t_balance, 
    t2.P_t_balance, 
from (
    SELECT 
     b.securities_code   AS SECURITIES_CODE, 
     b.buy_sell_flag    AS SIDE, 
     SUM(NVL(b.c_balance,0))  AS C_b_balance, 
     SUM(NVL(b.p_balance,0))  AS P_b_balance, 
    FROM b_balance b 
    GROUP BY b.securities_code, b.buy_sell_flag) t1 
left join (
    SELECT 
     t.securities_code   AS SECURITIES_CODE, 
     t.buy_sell_flag    AS SIDE, 
     SUM(NVL(t.c_balance,0))  AS C_t_balance, 
     SSUM(NVL(t.p_balance,0))  AS P_t_balance, 
    FROM t_balance t 
    GROUP BY t.securities_code, t.buy_sell_flag 
) on t1.securities_code = t2.securities_code and t1.buy_sell_flag = t2.buy_sell_flag 

Answer 2

當與來自不同表的聚集工作，聚集在加入之前（未聚合的加盟值）：

select 
    date '2017-08-01'         as trade_date, 
    securities_code         as securities_code, 
    buy_sell_flag          as side, 
    nvl(t.sum_c_balance, 0)       as c_t_balance, 
    nvl(b.sum_c_balance, 0)       as c_b_balance, 
    nvl(t.sum_c_balance, 0) - nvl(b.sum_c_balance, 0) as c_difference, 
    nvl(t.sum_p_balance, 0)       as p_t_balance, 
    nvl(b.sum_p_balance, 0)       as p_b_balance, 
    nvl(t.sum_p_balance, 0) - nvl(b.sum_c_balance, 0) as p_difference 
from 
(
    select 
    securities_code, 
    buy_sell_flag, 
    sum(c_balance) as sum_c_balance, 
    sum(p_balance) as sum_p_balance 
    from b_balance 
    group by securities_code, buy_sell_flag 
) b 
full outer join 
(
    select 
    securities_code, 
    buy_sell_flag, 
    sum(c_balance) as sum_c_balance, 
    sum(p_balance) as sum_p_balance 
    from t_balance 
    group by securities_code, buy_sell_flag 
) t using (securities_code, buy_sell_flag) 
order by securities_code, buy_sell_flag;