如何使用ORDER BY子句的SQL Server 2014

Question

這裏是我的查詢但它越來越錯誤

錯誤

消息156，15級，狀態1，第6行
時使用UNION ALL關鍵字「訂單」附近的語法不正確。

Msg 156，Level 15，State 1，Line 13
關鍵字'order'附近的語法不正確。

Msg 156，Level 15，State 1，Line 20
關鍵字'order'附近的語法不正確。

查詢

(SELECT TOP 20 cl_RooSiteId, 
       cl_CrawlOrgUrl 
FROM tblCrawlUrls 
WHERE Sysutcdatetime() > Dateadd(MINUTE, 50000, cl_LastCrawlDate) 
     AND cl_DuplicateUrl_ByCanonical = 0 
     AND cl_RooSiteId = 1 
ORDER BY cl_LastCrawlDate ASC) 
UNION ALL 
(SELECT TOP 200 cl_RooSiteId, 
       cl_CrawlOrgUrl 
FROM tblCrawlUrls 
WHERE Sysutcdatetime() > Dateadd(MINUTE, 50000, cl_LastCrawlDate) 
     AND cl_DuplicateUrl_ByCanonical = 0 
     AND cl_RooSiteId = 2 
ORDER BY cl_LastCrawlDate ASC) 
UNION ALL 
(SELECT TOP 50 cl_RooSiteId, 
       cl_CrawlOrgUrl 
FROM tblCrawlUrls 
WHERE Sysutcdatetime() > Dateadd(MINUTE, 50000, cl_LastCrawlDate) 
     AND cl_DuplicateUrl_ByCanonical = 0 
     AND cl_RooSiteId = 3 
ORDER BY cl_LastCrawlDate ASC) 

那麼我們應該如何寫的方式，它應該工作？

Answer 1

複製查詢結果爲temp table和使用Union all

SELECT TOP 20 cl_RooSiteId, 
       cl_CrawlOrgUrl,cl_LastCrawlDate 
INTO #Temp1 
FROM tblCrawlUrls 
WHERE Sysutcdatetime() > Dateadd(MINUTE, 50000, cl_LastCrawlDate) 
     AND cl_DuplicateUrl_ByCanonical = 0 
     AND cl_RooSiteId = 1 
ORDER BY cl_LastCrawlDate ASC 

SELECT TOP 200 cl_RooSiteId, 
       cl_CrawlOrgUrl,cl_LastCrawlDate 
INTO #temp2 
FROM tblCrawlUrls 
WHERE Sysutcdatetime() > Dateadd(MINUTE, 50000, cl_LastCrawlDate) 
     AND cl_DuplicateUrl_ByCanonical = 0 
     AND cl_RooSiteId = 2 
ORDER BY cl_LastCrawlDate ASC 

SELECT TOP 50 cl_RooSiteId, 
       cl_CrawlOrgUrl,cl_LastCrawlDate 
INTO #temp3 
FROM tblCrawlUrls 
WHERE Sysutcdatetime() > Dateadd(MINUTE, 50000, cl_LastCrawlDate) 
     AND cl_DuplicateUrl_ByCanonical = 0 
     AND cl_RooSiteId = 3 
ORDER BY cl_LastCrawlDate ASC 

SELECT cl_RooSiteId,cl_CrawlOrgUrl FROM #Temp1 
UNION ALL 
SELECT cl_RooSiteId,cl_CrawlOrgUrl FROM #Temp2 
UNION ALL 
SELECT cl_RooSiteId,cl_CrawlOrgUrl FROM #Temp3 
Order by cl_LastCrawlDate 

或者使用Stacked CTE

;WITH cte1 
    AS (SELECT TOP 20 cl_RooSiteId, 
         cl_CrawlOrgUrl,cl_LastCrawlDate 
     FROM tblCrawlUrls 
     WHERE Sysutcdatetime() > Dateadd(MINUTE, 50000, cl_LastCrawlDate) 
       AND cl_DuplicateUrl_ByCanonical = 0 
       AND cl_RooSiteId = 1 
     ORDER BY cl_LastCrawlDate ASC), 
    cte2 
    AS (SELECT TOP 200 cl_RooSiteId, 
         cl_CrawlOrgUrl,cl_LastCrawlDate 
     FROM tblCrawlUrls 
     WHERE Sysutcdatetime() > Dateadd(MINUTE, 50000, cl_LastCrawlDate) 
       AND cl_DuplicateUrl_ByCanonical = 0 
       AND cl_RooSiteId = 2 
     ORDER BY cl_LastCrawlDate ASC), 
    cte3 
    AS (SELECT TOP 50 cl_RooSiteId, 
         cl_CrawlOrgUrl,cl_LastCrawlDate 
     FROM tblCrawlUrls 
     WHERE Sysutcdatetime() > Dateadd(MINUTE, 50000, cl_LastCrawlDate) 
       AND cl_DuplicateUrl_ByCanonical = 0 
       AND cl_RooSiteId = 3 
     ORDER BY cl_LastCrawlDate ASC) 
SELECT cl_RooSiteId,cl_CrawlOrgUrl FROM cte1 
UNION ALL 
SELECT cl_RooSiteId,cl_CrawlOrgUrl FROM cte2 
UNION ALL 
SELECT cl_RooSiteId,cl_CrawlOrgUrl FROM cte3 
ORDER BY cl_LastCrawlDate ASC 

Answer 2

公共表表達式應​​該做的伎倆SEE THIS LINK

公共表表達式（CTE）可以被認爲是在唱歌的執行範圍內定義的臨時結果集SELECT，INSERT，UPDATE，DELETE或CREATE VIEW語句。 CTE類似於派生表，因爲它不是作爲對象存儲的，而是僅在查詢期間持續。與派生表不同，CTE可以自引用，並且可以在同一個查詢中多次引用。

這裏有一個片段

WITH TopLevel(colum1, column2, ...) AS 
(
    SELECT column1, column2, ... 
    FROM table1 --without ORDER BY 

    UNION ALL 

    SELECT column1, column2, ... 
    FROM table2 --without ORDER BY 
    . 
    . 
    . 

) 
SELECT * FROM 
FROM TopLevel 
ORDER BY column1, column2... --ORDER BY HERE... 

Answer 3

假設你在你寫的順序希望這些：

SELECT cl_RooSiteId, cl_CrawlOrgUrl 
FROM ((SELECT TOP 20 cl_RooSiteId, cl_CrawlOrgUrl, cl_LastCrawlDate, 0 as priority 
     FROM tblCrawlUrls 
     WHERE Sysutcdatetime() > Dateadd(MINUTE, 50000, cl_LastCrawlDate) AND 
       cl_DuplicateUrl_ByCanonical = 0 AND 
       cl_RooSiteId = 1 
    ) UNION ALL 
     (SELECT TOP 200 cl_RooSiteId, cl_CrawlOrgUrl, cl_LastCrawlDate, 1 as priority 
     FROM tblCrawlUrls 
     WHERE Sysutcdatetime() > Dateadd(MINUTE, 50000, cl_LastCrawlDate) AND 
       cl_DuplicateUrl_ByCanonical = 0 AND 
       cl_RooSiteId = 2 
    ) UNION ALL 
     (SELECT TOP 50 cl_RooSiteId, cl_CrawlOrgUrl, cl_LastCrawlDate, 2 
     FROM tblCrawlUrls 
     WHERE Sysutcdatetime() > Dateadd(MINUTE, 50000, cl_LastCrawlDate) AND 
       cl_DuplicateUrl_ByCanonical = 0 AND 
       cl_RooSiteId = 3 
    ) 
    ) c 
ORDER BY priority, cl_LastCrawlDate ASC 

注意在子查詢中包含的priority和cl_LastCrawlDate。我知道priority是多餘的，因爲您可以使用ORDER BY cl_RooSiteId, cl_LastCrawlDate。

編輯：

你也可以做到這一點沒有union all：

SELECT cl_RooSiteId, cl_CrawlOrgUrl 
FROM (SELECT c.*, 
      ROW_NUMBER() OVER (PARTITION BY cl_RooSiteId ORDER BY cl_LastCrawlDate) as seqnum 
     FROM tblCrawlUrls c 
     WHERE Sysutcdatetime() > Dateadd(MINUTE, 50000, cl_LastCrawlDate) AND 
      cl_DuplicateUrl_ByCanonical = 0 
    ) c 
WHERE (cl_RooSiteId = 1 and seqnum <= 20) OR 
     (cl_RooSiteId = 2 and seqnum <= 200) OR 
     (cl_RooSiteId = 3 and seqnum <= 50) 
ORDER BY cl_RooSiteId, cl_LastCrawlDate;