有没有办法在一种语法中选择具有不同子查询的两个计数？答案

【问题标题】：Is there a way to select two counts with different subqueries in one syntax?有没有办法在一种语法中选择具有不同子查询的两个计数？
【发布时间】：2019-04-11 14:10:09
【问题描述】：

基本上我想得到两个计数。第一个计数是记录数。第二个计数是有数据的记录数。这两个计数来自名为调查的表，该表没有所有州名。所以我将它与另一个表 UsState 连接起来以获取所有状态名称。我整理了语法，但出现错误。

SELECT
Name as State,
count (*)from NamrsFrozen2017.Investigation2017 where FiscalYear = 2017 AND StateName in (SELECT distinct StateName FROM [NamrsFrozen2017].[Investigation2017] WHERE (not ReportDate = '' ))[Report Date Records],
count (*) from NamrsFrozen2017.Investigation2017 WHERE FiscalYear = 2017 AND ReportDate != '' AND StateName in (SELECT distinct StateName FROM [NamrsFrozen2017].[Investigation2017] WHERE (not ReportDate = '' ))[Report Date Records with Data]
From
(SELECT DISTINCT Name FROM NamrsFullDw.UsState) s
LEFT JOIN NamrsFrozen2017.Investigation2017 c ON s.Name = c.StateName AND  c.FiscalYear = 2017
  GROUP BY Name
  Order by Name

【问题讨论】：

不要使用COUNT(*)，而是将列的名称放在函数之一的括号中。聚合函数不包含 NULL 值。 SELECT COUNT(*),COUNT(I) FROM (VALUES(1),(2),(NULL))V(I); 分别返回 3 和 2。
我不认为我理解您的回复，您能否澄清一下。我在名称和 s @Larnu 上收到错误
好吧，你不能这样做SELECT Name, COUNT(*) FROM... AS {alias}, COUNT(*) FROM...，这没有意义。我们看不到您看到的内容，因此请发布一些示例数据和预期结果（text，不是图片）。
样本数据、所需结果以及对您想要做什么的解释会有所帮助。

标签： sql sql-server

【解决方案1】：

一些相关的子查询会对您有所帮助。它们看起来像内联选择，但使用对内联查询之外的表的引用。像这样：

SELECT 
   Name as State,
   (SELECT COUNT(*) 
    FROM NamrsFrozen2017.Investigation2017 
    --I would think you want to view count of records for the current state, and not ALL states
    -- and not ALL states. So filter like this:
    WHERE FiscalYear = 2017 AND StateName = s.Name -- <-Correlated
    --Don't need a full list, like you had before
    -- (SELECT distinct StateName FROM [NamrsFrozen2017].[Investigation2017] WHERE (not ReportDate = '' ))
   ) AS [Report Date Records],
   (SELECT COUNT(*) 
    FROM NamrsFrozen2017.Investigation2017 
    WHERE FiscalYear = 2017 AND StateName = s.Name -- <- correlated
        AND ReportDate != '' --or maybe ReportDate IS NOT NULL            
    --Don't need to query the full table after correlating this to your FROM clause
    --  in (SELECT distinct StateName FROM [NamrsFrozen2017].[Investigation2017] WHERE (not ReportDate = '' ))
   ) AS [Report Date Records with Data]
FROM
   (SELECT DISTINCT Name FROM NamrsFullDw.UsState) s
   LEFT JOIN NamrsFrozen2017.Investigation2017 c 
        ON s.Name = c.StateName AND c.FiscalYear = 2017
GROUP BY s.Name
ORDER BY s.Name

这可能会得到更多优化，但我不熟悉您的数据，或者不清楚您想用原始查询完成什么。

【讨论】：

【解决方案2】：

评论太长了。

您遇到的语法错误是因为您的内联子查询格式不正确。你不能从COUNT(*)函数开始；它们每个都需要完全形成SELECT 语句。然后每一个都需要用括号括起来。

您的查询应该是这样的。

SELECT
  Name AS State
 ,(
    SELECT
      COUNT(*)
    FROM
      NamrsFrozen2017.Investigation2017
    WHERE
      FiscalYear = 2017
      AND StateName IN
          (
            SELECT DISTINCT StateName
            FROM NamrsFrozen2017.Investigation2017
            WHERE ReportDate <> ''
          )
  ) AS [Report Date Records]
 ,(
    SELECT
      COUNT(*)
    FROM
      NamrsFrozen2017.Investigation2017
    WHERE
      FiscalYear = 2017
      AND ReportDate != ''
      AND StateName IN
          (
            SELECT DISTINCT StateName
            FROM NamrsFrozen2017.Investigation2017
            WHERE (NOT ReportDate = '')
          )
  ) AS [Report Date Records with Data]
FROM
  (SELECT DISTINCT Name FROM NamrsFullDw.UsState) AS s
LEFT JOIN
  NamrsFrozen2017.Investigation2017 AS c
    ON
    s.Name = c.StateName
      AND c.FiscalYear = 2017
GROUP BY
  Name
ORDER BY
  Name;

这里还有一个次要问题，即此查询是否会按照您的预期执行。对于 Larnu 在 cmets 中的观点，聚合函数在不同情况下以不同方式处理 NULL。但是，如果没有要测试的样本数据，我无法判断这个功能版本的查询是否会返回您想要的结果。

【讨论】：