Get result by week - sql-server

Question

I need you for order a query by the date and then if it's possible to simplify it :-)
This query will return the number of entries per week from the current day.
Here is the query :
SELECT CONVERT(VarChar(50), DATEADD(day, DATEDIFF(day, 0, GETDATE()), -7), 103) as periode_join, COUNT(u.usr_ID) as cptu
FROM [USR_USER] u
INNER JOIN [USI_USER_SITE] s ON u.USR_ID = s.USR_ID
WHERE u.[USR_JOINED_DT] >= DATEADD(day, DATEDIFF(day, 0, GETDATE()), -7) and u.[USR_JOINED_DT] <= DATEADD(day, DATEDIFF(day, 0, GETDATE()), 0)
UNION
SELECT CONVERT(VarChar(50), DATEADD(day, DATEDIFF(day, 0, GETDATE()), -14), 103) as periode_join, COUNT(u.usr_ID) as cptu
FROM [USR_USER] u
INNER JOIN [USI_USER_SITE] s ON u.USR_ID = s.USR_ID
WHERE u.[USR_JOINED_DT] >= DATEADD(day, DATEDIFF(day, 0, GETDATE()), -14) and u.[USR_JOINED_DT] <= DATEADD(day, DATEDIFF(day, 0, GETDATE()), -7)
UNION
SELECT CONVERT(VarChar(50), DATEADD(day, DATEDIFF(day, 0, GETDATE()), -21), 103) as periode_join, COUNT(u.usr_ID) as cptu
FROM [USR_USER] u
INNER JOIN [USI_USER_SITE] s ON u.USR_ID = s.USR_ID
WHERE u.[USR_JOINED_DT] >= DATEADD(day, DATEDIFF(day, 0, GETDATE()), -21) and u.[USR_JOINED_DT] <= DATEADD(day, DATEDIFF(day, 0, GETDATE()), -14)
UNION
SELECT CONVERT(VarChar(50), DATEADD(day, DATEDIFF(day, 0, GETDATE()), -28), 103) as periode_join, COUNT(u.usr_ID) as cptu
FROM [USR_USER] u
INNER JOIN [USI_USER_SITE] s ON u.USR_ID = s.USR_ID
WHERE u.[USR_JOINED_DT] >= DATEADD(day, DATEDIFF(day, 0, GETDATE()), -28) and u.[USR_JOINED_DT] <= DATEADD(day, DATEDIFF(day, 0, GETDATE()), -21)
UNION
SELECT CONVERT(VarChar(50), DATEADD(day, DATEDIFF(day, 0, GETDATE()), -35), 103) as periode_join, COUNT(u.usr_ID) as cptu
FROM [USR_USER] u
INNER JOIN [USI_USER_SITE] s ON u.USR_ID = s.USR_ID
WHERE u.[USR_JOINED_DT] >= DATEADD(day, DATEDIFF(day, 0, GETDATE()), -35) and u.[USR_JOINED_DT] <= DATEADD(day, DATEDIFF(day, 0, GETDATE()), -28)
UNION
SELECT CONVERT(VarChar(50), DATEADD(day, DATEDIFF(day, 0, GETDATE()), -42), 103) as periode_join, COUNT(u.usr_ID) as cptu
FROM [USR_USER] u
INNER JOIN [USI_USER_SITE] s ON u.USR_ID = s.USR_ID
WHERE u.[USR_JOINED_DT] >= DATEADD(day, DATEDIFF(day, 0, GETDATE()), -42) and u.[USR_JOINED_DT] <= DATEADD(day, DATEDIFF(day, 0, GETDATE()), -35)
ORDER BY periode_join desc
Here is the result :
periode_join cptu
28/05/2013 8740
25/06/2013 9773
18/06/2013 8212
11/06/2013 6644
04/06/2013 9420
02/07/2013 7868
Thanks a lot :-D

Answer 1

if you have SQL Server 2005 and above, you can try something like this:
;with CTE_Data as
(
select
u.usr_id,
cast(dateadd(dd, -(datediff(dd, u.usr_joined_dt, getdate()) / 7 + 1) * 7, getdate()) as date) as periode_join
from usr_user as u
inner join usi_user_site as s on u.usr_id = s.usr_id
)
select periode_join, count(usr_id) as cptu
from CTE_Data
group by periode_join
order by periode_join asc
SQL FIDDLE EXAMPLE
To split data by weeks we have to take difference in days between getdate() and date from table - datediff(dd, u.usr_joined_dt, getdate()), then we have to get number of whole weeks - / 7, then we have to substract whole number of weeks + 1.
So, for example, if date = '2013/06/30' then number of days = 9, whole number of weeks = 1 and we have to substract 1 + 1 weeks so we get '2013/06/25' and so on