Question

I have a table and I need to count all unique users on it. The schema of the table is:

idAccess | idClient | date             | uid
1        | 12       | 2012-04-03 10:59 | 1234-1234-1234
2        | 13       | 2012-01-03 11.23 | 2345-2345-2345
3        | 12       | 2012-04-03 10:59 | 1234-1234-1234
4        | 12       | 2012-04-03 11:59 | 1234-1234-1234
5        | 12       | 2012-02-23 02:39 | 5788-5788-5687
6        | 12       | 2011-12-03 12:31 | asdf-1234-asdf
7        | 12       | 2011-10-13 13:36 | eeef-1234-eeee
8        | 15       | 2010-11-23 17:33 | qwer-EeQE-fhjh

This is the query:

Select 
    count(*) AS ct, count(DISTINCT(uid)) as users, YEAR(date) as ano 
FROM 
    [AQUA_INTRANET].[dbo].[Accesos] 
WHERE 
    idClient IN (12,13,15) 
    AND date> '01-01-2006 00:00' AND date< '04-02-2012 23:59'   
GROUP BY 
    YEAR(date) 
ORDER BY 
    year

and the result:

ct | users  | year
5  |   6    |  2012
2  |   3    |  2011
1  |   1    |  2010 

If I execute this query, I’m getting an wrong results, because the sum of the results is different from this other query:

Select 
     count(DISTINCT(uid)) as users
FROM 
     [AQUA_INTRANET].[dbo].[Accesos] 
WHERE  
     idCliente IN (12,13,15) 
     AND fecha > '01-01-2006 00:00' 
     AND fecha < '04-02-2012 23:59'

and result is:

users
 6

that is different if you sum the users column on the other query.

This happens sometimes, not always, but it’s very strange, because I’m having different values if I group by year or if I get the sum of all unique users.

I don’t know where is the problem, and I tried with different queries, but the result it wrong if I group by year.

Answer 1

One uid can have records that are in more than one year.

Try these three queries

SELECT count(DISTINCT(uid)) as users
  FROM [AQUA_INTRANET].[dbo].[Accesos] 
 WHERE idCliente IN (12,13,15)
   AND fecha >= '01-01-2010 00:00' AND fecha < '01-01-2011'

SELECT count(DISTINCT(uid)) as users
  FROM [AQUA_INTRANET].[dbo].[Accesos] 
 WHERE idCliente IN (12,13,15)
   AND fecha >= '01-01-2011 00:00' AND fecha < '01-01-2012'

SELECT count(DISTINCT(uid)) as users
  FROM [AQUA_INTRANET].[dbo].[Accesos] 
 WHERE idCliente IN (12,13,15)
   AND fecha >= '01-01-2012 00:00' AND fecha < '04-02-2012 23:59'

Also, note that I use >= AND < to ensure no records ‘fall through the gaps’ or get double counted.