子查詢導致SQL Server的無限執行

Question

我遇到了一個SQL語句與子查詢結構：

SELECT a12.ship_date  SHIP_DATE, 
     Sum(a11.qty_sold) QTY_SOLD 
FROM order_detail a11 
     JOIN order_fact a12 
     ON (a11.customer_id = a12.customer_id 
       AND a11.emp_id = a12.emp_id 
       AND a11.order_date = a12.order_date 
       AND a11.order_id = a12.order_id) 
WHERE a12.ship_date IN (SELECT DISTINCT ship_date 
         FROM order_fact 
         WHERE order_date BETWEEN (SELECT DISTINCT day_date - 1 
                FROM lu_day 
                WHERE 
           day_date = CONVERT(DATETIME, 
              '2007-01-01 00:00:00', 
              120)) AND 
                (SELECT DISTINCT day_date + 1 
                FROM 
                lu_day 
                WHERE 
           day_date = CONVERT(DATETIME, 
              '2007-01-01 00:00:00', 
              120))) 
GROUP BY a12.ship_date 

它不會停止，直到我強迫這一點。但是，如果我改變的地方進入

where a12.SHIP_DATE in ('2009-09-08 00:00:00','2009-07-08 00:00:00') 

它可以被正確執行。

我猜的原因是子查詢包含ORDER_FACT哪個列在外部查詢也依賴於它，但我不知道爲什麼。

我不熟悉SQL Server，任何人都可以解釋爲什麼會發生？謝謝。

========更新=======

謝謝你們的回答。刪除未使用的子查詢時可以更快地執行，但我仍然想知道大規模子查詢中發生了哪些瓶頸。

爲每個答案投票，因爲他們是正確的解決方案，但我認爲最好是一些執行計劃將幫助我學習更多。謝謝。

Answer 1

不知道這會解決它，而是嘗試在子查詢更具體：

SELECT DISTINCT ship_date 
FROM order_fact OFT 
WHERE OFT.order_date BETWEEN 
    (select distinct day_date-1 
    from lu_day 
    where day_date=CONVERT(datetime, '2007-01-01 00:00:00', 120)) 
    AND 
    (SELECT DISTINCT day_date + 1 
    FROM lu_day 
    WHERE day_date = CONVERT(DATETIME, '2007-01-01 00:00:00', 120)) 

，因爲這可以更好地分離的東西。

Answer 2

嘗試使用JOIN而不是IN。此外，我無法理解BETWEEN條件它相當於@NickyW提到的常量（除非該數據不存在於lu_day表中，那麼它將返回NULL）。

SELECT a12.ship_date  SHIP_DATE, 
     Sum(a11.qty_sold) QTY_SOLD 
FROM order_detail a11 
     JOIN order_fact a12 
     ON (a11.customer_id = a12.customer_id 
       AND a11.emp_id = a12.emp_id 
       AND a11.order_date = a12.order_date 
       AND a11.order_id = a12.order_id) 
     JOIN (SELECT DISTINCT ship_date 
         FROM order_fact 
         WHERE order_date BETWEEN CONVERT(DATETIME, 
              '2007-01-01 00:00:00', 
              120)) - 1 
              AND 
              CONVERT(DATETIME, 
              '2007-01-01 00:00:00', 
              120)) +1 
      ) as SHIPDATES ON a12.ship_date = SHIPDATES.ship_date 
GROUP BY a12.ship_date 

Answer 3

的查詢功能相同

SELECT a12.ship_date  SHIP_DATE, 
     Sum(a11.qty_sold) QTY_SOLD 
FROM order_detail a11 
     JOIN order_fact a12 
     ON (a11.customer_id = a12.customer_id 
       AND a11.emp_id = a12.emp_id 
       AND a11.order_date = a12.order_date 
       AND a11.order_id = a12.order_id) 
WHERE a12.ship_date BETWEEN CONVERT(DATETIME, '2007-01-01 00:00:00', 120) - 1 
         AND CONVERT(DATETIME, '2007-01-01 00:00:00', 120) + 1 
GROUP BY a12.ship_date 

爲a12是order_fact別名所以a12.ship_date始終處於order_fact.ship_date的唯一途徑，在BETWEEN子查詢不CONVERT(DATETIME, '2007-01-01 00:00:00', 120)是，如果有在lu_day中沒有day_date，等於2007-01-01，在那種情況下它將變成

WHERE a12.ship_date BETWEEN NULL - 1 
         AND NULL + 1 

可能，因爲不會有行的子查詢

編輯
李自成查詢也許我錯了，因爲WHERE條件是在ORDER_DATE

WHERE a12.ship_date 
     IN (SELECT ship_date 
      FROM order_fact 
      WHERE order_date BETWEEN 
          CONVERT(DATETIME, '2007-01-01 00:00:00', 120) - 1 
         AND CONVERT(DATETIME, '2007-01-01 00:00:00', 120) + 1) 

如果我讀如果查詢正確，則結果將是訂單的ship date和訂單的售出數量，該訂單具有訂單的ship_date，該訂單的ship_date是在訂單的前一天和下一天之間訂購的，在這種情況下，可以放棄發貨日期的條件，在訂單當天促銷

SELECT a12.ship_date  SHIP_DATE, 
     Sum(a11.qty_sold) QTY_SOLD 
FROM order_detail a11 
     JOIN order_fact a12 
     ON (a11.customer_id = a12.customer_id 
       AND a11.emp_id = a12.emp_id 
       AND a11.order_date = a12.order_date 
       AND a11.order_id = a12.order_id) 
WHERE a12.order_date BETWEEN CONVERT(DATETIME, '2007-01-01 00:00:00', 120) - 1 
          AND CONVERT(DATETIME, '2007-01-01 00:00:00', 120) + 1 
GROUP BY a12.ship_date 

查詢的性能緊密地與索引和表的尺寸並列，例如在最後一個查詢上order_fact.order_date指數會有所幫助，並且兩個表上customer_id, emp_id, order_date, order_id指數將幫助JOIN操作。