我如何嵌套这些SQL语句而不会出现group by子句错误？

Question

I am trying to query a database that has these params:

Transaction Date, User Email Address

What I have done is use this query:

SELECT [User Email Address],  COUNT(*) AS 'count'
    FROM
      [DATABASE].[TABLE]
    GROUP BY [User Email Address]

which displays a table with params:

Email Address, count

In this case the count column shows the number of occurrences of the user email in the original table.

What I am trying to do next is look at the Transaction Date column for the last year up to today, and compare the count column for this subset with the count column of the orginal (which goes back some 3 years). Specifically, I want my end resultant table to be:

User User Email Address, countDiff

where countDiff is the difference in counts from the one year subset and the original subset.

I have tried:

SELECT [User Email Address], [Transaction Date], [count - COUNT(*)] AS 'countdDifference'

FROM (

    SELECT [User Email Address],  COUNT(*) AS 'count'
    FROM
      [DATABASE].[TABLE]
    GROUP BY [User Email Address]


) a

WHERE a.[Transaction Date] >= '2011-08-07 00:00:00.000'


ORDER BY [count] DESC

But I get the error that [Transaction Date] is not in the Group By clause or aggregate. If I put it in the Group By next to [User Email Address] , it messes up the data.

This is actually a common problem I've had. Any ways to circumvent this?

Answer 1

You need to use two different subqueries: One that counts the full entries and another one that counts the entries of the last year.

Maybe this will help you:

SELECT a.*, a.[count] - Coalesce(b.[count], 0) as 'countDif' 
FROM 
    (
        SELECT [User Email Address],  COUNT(*) AS 'count'
        FROM [DATABASE].[TABLE]
        GROUP BY [User Email Address]
    ) AS a
    LEFT JOIN (
        SELECT [User Email Address],  COUNT(*) AS 'count'
        FROM [DATABASE].[TABLE]
        WHERE [Transaction Date] >= '2011-08-07 00:00:00.000'
        GROUP BY [User Email Address]
    ) AS b ON a.[User Email Address] = b.[User Email Address]

Answer 2

You can do both counts in one SELECT:

SELECT [User Email Address],  
    SUM(CASE WHEN [Transaction Date] >= '2011-08-07' THEN 1 ELSE 0 END) AS 'FilteredCount',
    COUNT(*) AS 'TotalCount',
    COUNT(*) 
       - SUM(CASE WHEN [Transaction Date] >= '2011-08-07' THEN 1 ELSE 0 END)
         AS 'CountDifference'
    FROM
      [DATABASE].[TABLE]
    GROUP BY [User Email Address]

Answer 3

You can use this:

SELECT [User Email Address], 
       [Transaction Date], 
       count1.count - isnull(count2.count,0) as  [countdDifference]
FROM    (SELECT [User Email Address],  COUNT(*) AS 'count'
        FROM    [DATABASE].[TABLE] t1
        GROUP BY t1.[User Email Address]) as count1
LEFT JOIN (SELECT [User Email Address],  COUNT(*) AS 'count'
        FROM    [DATABASE].[TABLE] t2
        GROUP BY [User Email Address] 
        WHERE t2.[Transaction Date] >= '2011-08-07 00:00:00.000') as count2
ON      count2.[User Email Address] = count1.[User Email Address]
ORDER BY 3 DESC

You should also start thinking about 1.) not using count as it's used a bunch (it's almost a reserved word but not quite); 2.) don't use spaces in your field names; 3.) finding a easy-to-read way to organize your SQL :)

Answer 4

,How about something like this?

SELECT
    [User Email Address],
    count(*) AS Total,
    sum(CASE 
            WHEN [Transaction Date] >= '2011/08/07 00:00:00.000' THEN 1
            ELSE 0 
         END) AS WithinDateRange,
    count(*) 
        - sum(CASE 
                  WHEN [Transaction Date] >= '2011/08/07 00:00:00.000' THEN 1
                  ELSE 0 
              END) AS Difference.
FROM [DATABASE].[TABLE]
GROUP BY [User Email Address]