TSQL：外部连接表的计数产生不正确的结果

Posted 2023-02-24

【中文标题】TSQL：外部连接表的计数产生不正确的结果

【英文标题】：TSQL: Count on outer joined tables is producing incorrect results

【发布时间】：2015-04-02 03:32:30

【问题描述】：

生成关于 SQL Server 2008 的报告。我已使用 LEFT OUTER JOIN 将一个表与其他五个表连接起来。当我对那些其他表进行计数时，我得到了不正确的数据。我知道为什么，但我不确定如何解决它。

查询正在跟踪一所学校的招生候选人。当他们完成整个流程时，他们会在流程的每个主要阶段被标记。我需要的是计算在特定时期（年和月）中有多少人通过了特定阶段。大多数情况下它都有效。但是，如果候选人通过了该期间的任何阶段，则该候选人也会在之前的阶段中被“计数”，即使它们发生在查询期间之前。一个很好的例子是 AD_35，一个特定的学术项目应该有一个人，但输出显示为 2。当我只查询 AD_35 表时，我得到了正确的信息。所以我知道问题出在外连接上，但我不确定如何解决这个问题（在生成我的命名输出的子查询中尝试了各种标准）。对某人来说应该很容易......在此先感谢，代码如下。 :Year 和 :Month 用于用户输入，将填充数值（例如 2015 1）

连续波

SELECT DISTINCT 
   ad_candidacy.prog_cde,   
   ad_candidacy.stageyr,   
   ad_candidacy.stagemo,   
   Count (case when (ad_02.stageyr in (:Year, :Year -1, :Year-2) and ad_02.stagemo <= :month) then 1 else null end) as Inquiry,   
   Count (case when (ad_05.stageyr in (:Year, :Year -1, :Year-2) and ad_05.stagemo <= :month) then 1 else null end) as Applied,   
   Count (case when (ad_35.stageyr in (:Year, :Year -1, :Year-2) and ad_35.stagemo <= :month and ad_35.id_num = ad_candidacy.id_num and ad_35.stageyr = ad_candidacy.stageyr and ad_35.stagemo=ad_candidacy.stagemo) then 1 else null end) as Accepted,   
   Count (case when (ad_50.stageyr in (:Year, :Year -1, :Year-2) and ad_50.stagemo <= :month) then 1 else null end) as Matriculated,   
   Count (case when (ad_enroll.stageyr in (:Year, :Year -1, :Year-2) and ad_enroll.stagemo <= :month) then 1 else null end) as Enrolled,   
   ad_candidacy.stagemo_long  
FROM 
   ad_candidacy 
LEFT OUTER JOIN 
   ad_02 ON ad_candidacy.id_num = ad_02.id_num 
LEFT OUTER JOIN 
   ad_05 ON ad_candidacy.id_num = ad_05.id_num 
LEFT OUTER JOIN 
   ad_35 ON ad_candidacy.id_num = ad_35.id_num 
LEFT OUTER JOIN 
   ad_enroll ON ad_candidacy.id_num = ad_enroll.id_num 
LEFT OUTER JOIN 
   ad_50 ON ad_candidacy.id_num = ad_50.id_num  
WHERE 
   (ad_candidacy.stageyr in (:Year, :Year -1, :Year-2) ) 
   AND ( ad_candidacy.stagemo <= :Month )   
GROUP BY 
   ad_candidacy.prog_cde,   
   ad_candidacy.stageyr,   
   ad_candidacy.stagemo,   
   ad_candidacy.stagemo_long  
ORDER BY 
   ad_candidacy.stageyr ASC  

【问题讨论】：

您是说如果候选人被算作“已接受”，那么即使他们在 2 年多前申请，也算作“已申请”？我不明白怎么做。或者，如果您说的是去年申请并在今年被录取的候选人，那么他们被计算两次是有道理的。

我说的是第二个。为什么有意义？ （是的，我需要在这里上学，哈！）候选人于 2014 年 12 月被录取，并于 2015 年 1 月入学。如果我查询 AD_35（已接受）表，我会看到 12 月的日期。如果我查询并限制在 2015 年 1 月，则候选人不会出现。但是，当我运行包含联接的上述查询时，有一个候选人在 2015 年 1 月接受和注册。我认为 Count (case... 语句中的子查询可以解决这个问题，但没有。

【参考方案1】：

连接多个表需要考虑连接条件。第二个表可能有同一行的多行。为确保最终得到重复项，您可以在加入第一个表之前在子查询中搜索第二个表。

SELECT a.Name, 
       b.Total
FROM table1 as a
LEFT OUTER JOIN ( SELECT table1Id, Total = COUNT(b.some_measure) from table2 group by table1Id)as b ON a.table1Id = b.table1Id

【讨论】：

同意考虑加入条件。单独查询第二个表可以给我正确的信息。但是，您的代码示例为我提供了一些额外的素材来重新设计 OUTER JOIN 逻辑，我现在正在处理它的变体。

【参考方案2】：

Ako 的回答为我指明了正确的方向。我正在使用子查询，但他的示例导致了正确的输出。下面是代码的工作版本。谢谢！

SELECT DISTINCT 
   ad_candidacy.prog_cde,   
   ad_candidacy.stageyr,   
   ad_candidacy.stagemo, 
   ad_candidacy.StageMo_Long,
   COUNT (case when (Inquiry IS NOT NULL) then 1 else null end) as Inquiry,
   COUNT (case when (Applied IS NOT NULL) then 1 else null end) as Applied,
   count (case when (Accepted is not null) then 1 else null end) as Accepted,
   COUNT (case when (Matriculated IS NOT NULL) then 1 else null end) as Matriculated,
   count (case when (Enrolled is not null) then 1 else null end) as Enrolled
FROM 
   ad_candidacy
LEFT OUTER JOIN
 (select id_num, Inquiry = COUNT (id_num) from ad_02 where stageyr in (:year, :year-1, :year-2) and StageMo <= :month group by id_num) as ad_02 on ad_candidacy.id_num = ad_02.id_num     
LEFT OUTER JOIN
 (select id_num, Accepted = COUNT (id_num) from ad_35 where stageyr in (:year, :year-1, :year-2) and StageMo <= :month group by id_num) as ad_35 on ad_candidacy.id_num = ad_35.id_num    
LEFT OUTER JOIN
 (select id_num, Applied = COUNT (id_num) from ad_05 where stageyr in (:year, :year-1, :year-2) and StageMo <= :month group by id_num) as ad_05 on ad_candidacy.id_num = ad_05.id_num    
LEFT OUTER JOIN
 (select id_num, Matriculated = COUNT (id_num) from ad_50 where stageyr in (:year, :year-1, :year-2) and StageMo <= :month group by id_num) as ad_50 on ad_candidacy.id_num = ad_50.id_num   
LEFT OUTER JOIN
 (select id_num, Enrolled = COUNT (id_num) from ad_enroll where stageyr in (:year, :year-1, :year-2) and StageMo <= :month group by id_num) as ad_enroll on ad_candidacy.id_num = ad_enroll.id_num  
WHERE 
   (ad_candidacy.stageyr in (:year, :year-1, :year-2) ) 
   AND ( ad_candidacy.stagemo <= :month )   
GROUP BY 
   ad_candidacy.prog_cde,   
   ad_candidacy.stageyr,   
   ad_candidacy.stagemo,   
   ad_candidacy.stagemo_long