使用多个CASE语句时出错[SQL]

Question

在Microsoft SQL Server中使用SQL时，我遇到了一个非常具体的问题，即我要比较两个月之间的数据。 此数据可以具有两种状态之一-为简单起见，它可以是“ 1”或“ 0”，或者基本上是“是”或“否”。

我想考虑以下情况：

• 如果某行在一个月内的状态为“ 1”，而在下个月变为“ 0”， 反之亦然
• 如果状态没有从一个月更改为另一个月（即“ 1”更改为“ 1”或“ 0”更改为“ 0”）
• 如果状态值在上个月为空，并且在下个月更改为“ 1”或“ 0”

我假设我可以为此使用多个CASE语句，到目前为止，在遇到问题之前，我得到的是以下内容：

select
     --determine if status changes from 1 to 0 between two months or stays the same
     case
     when T.[status] = '1' and T.date between '01-01-2017' and '02-28-2017'
     then 'No Change'
     else
     'Status Changed to 0'
     end,

     --determine if status changes from 0 to 1 between two months or stays the same
     case
     when T.[status] = '0' and T.date between '01-01-2017' and '02-28-2017'
     then 'No Change'
     else 
     'Status Changed to 1'
     end,

     --determine if there was new row that didn't appear in first month with a status of 1 or 0 in second month
     case
     when T.status is null in T.date between '01-01-2017' and '01-31-2017'  
     and T.status = '1' in T.date between '02-01-2017' and '02-28-2017' then 'New Status to 1'
     else
     'New Status of 1'
     end, *
from table T
where ....

该代码适用于第一个CASE语句，以在新列中显示从1到0的状态更改，但是它使第二列显示另一状态，因此我不确定它是否按预期工作。 而且，由于我正在一起检查两个月，因此我似乎无法逻辑地将其他方案编码到我的SQL中。

我试图找到一种方法来压缩所有CASE语句，使其仅产生一个额外的列，而我遇到的另一个问题是，第三条CASE语句在第一个“ IN”处引发了语法错误。 希望这对某人有意义，并在此先感谢您提供的任何输入或技巧。

Answer 1

首先考虑最后一个问题，即“ IN”的不正确语法：

 case
 when T.status is null in t.date between '01-01-2017' and '01-31-2017'  
 and T.status = '1' in T.date between '02-01-2017' and '02-28-2017' then 'New Status to 1'
 else
 'test'
 end

尝试：

 case
 when (T.status is null AND t.date between '01-01-2017' and '01-31-2017')  
 OR (T.status = '1' AND T.date between '02-01-2017' and '02-28-2017') then 'New Status to 1'
 else
 'test'
 end

为了更轻松地检查状态更改，请以一个月的补偿将表连接到自身。

Answer 2

我在WITH子句中添加了带有输入数据的示例表，它可以满足您的要求。 我正在使用SQL Server支持的LAG（）分析函数。 请记住，在分析窗口分区的第一行中，LAG（）返回NULL-因为在这种情况下没有前一行。

在SQL Server以外的其他DBMS中，您将需要ANSI串联运算符|| 而不是+来连接两个字符串。

WITH
input(input_id,monthfirst,status) AS (
          SELECT 1,DATE '2017-01-01',0
UNION ALL SELECT 1,DATE '2017-02-01',1
UNION ALL SELECT 1,DATE '2017-03-01',0
UNION ALL SELECT 2,DATE '2017-01-01',1
UNION ALL SELECT 2,DATE '2017-02-01',1
UNION ALL SELECT 2,DATE '2017-03-01',0
UNION ALL SELECT 3,DATE '2017-01-01',CAST(NULL AS INT)
UNION ALL SELECT 3,DATE '2017-02-01',1
UNION ALL SELECT 3,DATE '2017-03-01',0
)
,
input_with_prev_status AS (
SELECT
  *
, LAG(status) OVER (PARTITION BY input_id ORDER BY monthfirst) AS prev_status
FROM input
)
SELECT
  input_id
, monthfirst
, status
, CASE
    WHEN prev_status IS NULL AND status IS NOT NULL
      THEN 'New status to ' + CAST(status AS CHAR(1))
    WHEN prev_status <> status
      THEN 'Status changed to ' + CAST(status AS CHAR(1))
    WHEN prev_status = status
      THEN 'no status change'
    WHEN prev_status IS NULL AND status IS NULL
      THEN 'status remains missing'
    ELSE 'unforeseen status change'
  END AS status_change
FROM input_with_prev_status
;
input_id|monthfirst|status|status_change
       1|2017-01-01|     0|New status to 0
       1|2017-02-01|     1|Status changed to 1
       1|2017-03-01|     0|Status changed to 0
       2|2017-01-01|     1|New status to 1
       2|2017-02-01|     1|no status change
       2|2017-03-01|     0|Status changed to 0
       3|2017-01-01|-     |status remains missing
       3|2017-02-01|     1|New status to 1
       3|2017-03-01|     0|Status changed to 0

Answer 3

请参见下面的自我连接代码（由Dejan建议）。 您需要使用表T中的键列创建ON语句

select 
  case 
  when (T1.[status] = '1' and T2. [status] = '1') or (T1.[status] = '0' and T2. [status] = '0') then 'No Change'
  when T1.[status] = '1' and T2. [status] = '0' then 'Status Changed to 0'
  when T1.[status] = '0' and T2. [status] = '1' then 'Status Changed to 1'
  when T.status is null and  T2. [status] = '1' then 'New Status to 1'
 else
 'test'
  end
table T1 
inner join T2
   on ....join on the key column
where T1.date between '01-01-2017' and '31-01-2017' 
and T2.date between '01-02-2017' and '28-02-2017'