SQL Server Group By Month

Question

I have a table which has this schema

ItemID    UserID    Year    IsPaid    PaymentDate  Amount
1         1         2009    0         2009-11-01  300
2         1         2009    0         2009-12-01  342
3         1         2010    0         2010-01-01  243
4         1         2010    0         2010-02-01  2543
5         1         2010    0         2010-03-01  475

I'm trying to get a query working which shows the totals for each month. So far I've tried DateDiff and nested selects, but neither gives me what I want. This is the closest I have I think:

DECLARE @start [datetime] = 2010/4/1;
SELECT ItemID, IsPaid,
(SELECT SUM(Amount) FROM Payments WHERE Year = 2010 And DateDiff(m, PaymentDate, @start) = 0 AND UserID = 100) AS "Apr",
(SELECT SUM(Amount) FROM Payments WHERE Year = 2010 AND DateDiff(m, PaymentDate, @start) =1 AND UserID = 100) AS "May",
(SELECT SUM(Amount) FROM Payments WHERE Year = 2010 AND DateDiff(m, PaymentDate, @start) =2 AND UserID = 100) AS "Jun", 
(SELECT SUM(Amount) FROM Payments WHERE Year = 2010 AND DateDiff(m, PaymentDate, @start) =3 AND UserID = 100) AS "Jul", 
(SELECT SUM(Amount) FROM Payments WHERE Year = 2010 AND DateDiff(m, PaymentDate, @start) =4  AND UserID = 100) AS "Aug", 
(SELECT SUM(Amount) FROM Payments WHERE Year = 2010 AND DateDiff(m, PaymentDate, @start) =5  AND UserID = 100) AS "Sep", 
(SELECT SUM(Amount) FROM Payments WHERE Year = 2010 AND DateDiff(m, PaymentDate, @start) =6  AND UserID = 100) AS "Oct", 
(SELECT SUM(Amount) FROM Payments WHERE Year = 2010 AND DateDiff(m, PaymentDate, @start) =7 AND UserID = 100) AS "Nov", 
(SELECT SUM(Amount) FROM Payments WHERE Year = 2010 AND DateDiff(m, PaymentDate, @start) =8 AND UserID = 100) AS "Dec", 
(SELECT SUM(Amount) FROM Payments WHERE Year = 2010 AND DateDiff(m, PaymentDate, @start) =9 AND UserID = 100) AS "Jan", 
(SELECT SUM(Amount) FROM Payments WHERE Year = 2010 AND DateDiff(m, PaymentDate, @start) =10 AND UserID = 100) AS "Feb", 
(SELECT SUM(Amount) FROM Payments WHERE Year = 2010 AND DateDiff(m, PaymentDate, @start) =11 AND UserID = 100) AS "Mar" 
FROM LIVE L INNER JOIN Payments I ON I.LiveID = L.RECORD_KEY 
WHERE UserID = 16178 

But I just get nulls when I should be getting values. Am I missing something?

Answer 1

DECLARE @start [datetime] = 2010/4/1;

Should be...

DECLARE @start [datetime] = '2010-04-01';

The one you have is dividing 2010 by 4, then by 1, then converting to a date. Which is the 57.5th day from 1900-01-01.

Try SELECT @start after your initialisation to check if this is correct.

Answer 2

SELECT CONVERT(NVARCHAR(10), PaymentDate, 120) [Month], SUM(Amount) [TotalAmount]
FROM Payments
GROUP BY CONVERT(NVARCHAR(10), PaymentDate, 120)
ORDER BY [Month]

You could also try:

SELECT DATEPART(Year, PaymentDate) Year, DATEPART(Month, PaymentDate) Month, SUM(Amount) [TotalAmount]
FROM Payments
GROUP BY DATEPART(Year, PaymentDate), DATEPART(Month, PaymentDate)
ORDER BY Year, Month

Answer 3

If you need to do this frequently, I would probably add a computed column PaymentMonth to the table:

ALTER TABLE dbo.Payments ADD PaymentMonth AS MONTH(PaymentDate) PERSISTED

It's persisted and stored in the table - so there's really no performance overhead querying it. It's a 4 byte INT value - so the space overhead is minimal, too.

Once you have that, you could simplify your query to be something along the lines of:

SELECT ItemID, IsPaid,
(SELECT SUM(Amount) FROM Payments WHERE Year = 2010 And PaymentMonth = 1 AND UserID = 100) AS 'Jan',
(SELECT SUM(Amount) FROM Payments WHERE Year = 2010 And PaymentMonth = 2 AND UserID = 100) AS 'Feb',
.... and so on .....
FROM LIVE L 
INNER JOIN Payments I ON I.LiveID = L.RECORD_KEY 
WHERE UserID = 16178 

Answer 4

Restrict the dimension of the NVARCHAR to 7, supplied to CONVERT to show only "YYYY-MM"

SELECT CONVERT(NVARCHAR(7),PaymentDate,120) [Month], SUM(Amount) [TotalAmount]
FROM Payments
GROUP BY CONVERT(NVARCHAR(7),PaymentDate,120)
ORDER BY [Month]

Answer 5

Now your query is explicitly looking at only payments for year = 2010, however, I think you meant to have your Jan/Feb/Mar actually represent 2009. If so, you'll need to adjust this a bit for that case. Don't keep requerying the sum values for every column, just the condition of the date difference in months. Put the rest in the WHERE clause.

SELECT 
      SUM( case when DateDiff(m, PaymentDate, @start) = 0 
           then Amount else 0 end ) AS "Apr",
      SUM( case when DateDiff(m, PaymentDate, @start) = 1 
           then Amount else 0 end ) AS "May",
      SUM( case when DateDiff(m, PaymentDate, @start) = 2 
           then Amount else 0 end ) AS "June",
      SUM( case when DateDiff(m, PaymentDate, @start) = 3 
           then Amount else 0 end ) AS "July",
      SUM( case when DateDiff(m, PaymentDate, @start) = 4 
           then Amount else 0 end ) AS "Aug",
      SUM( case when DateDiff(m, PaymentDate, @start) = 5 
           then Amount else 0 end ) AS "Sep",
      SUM( case when DateDiff(m, PaymentDate, @start) = 6 
           then Amount else 0 end ) AS "Oct",
      SUM( case when DateDiff(m, PaymentDate, @start) = 7 
           then Amount else 0 end ) AS "Nov",
      SUM( case when DateDiff(m, PaymentDate, @start) = 8 
           then Amount else 0 end ) AS "Dec",
      SUM( case when DateDiff(m, PaymentDate, @start) = 9 
           then Amount else 0 end ) AS "Jan",
      SUM( case when DateDiff(m, PaymentDate, @start) = 10 
           then Amount else 0 end ) AS "Feb",
      SUM( case when DateDiff(m, PaymentDate, @start) = 11 
           then Amount else 0 end ) AS "Mar"
   FROM 
      Payments I
         JOIN Live L
            on I.LiveID = L.Record_Key
   WHERE 
          Year = 2010 
      AND UserID = 100

Answer 6

Another approach, that doesn't involve adding columns to the result, is to simply zero-out the day component of the date, so 2016-07-13 and 2016-07-16 would both be 2016-07-01 - thus making them equal by month.

If you have a date (not a datetime) value, then you can zero it directly:

SELECT
    DATEADD( day, 1 - DATEPART( day, [Date] ), [Date] ),
    COUNT(*)
FROM
    [Table]
GROUP BY
    DATEADD( day, 1 - DATEPART( day, [Date] ), [Date] )

If you have datetime values, you'll need to use CONVERT to remove the time-of-day portion:

SELECT
    DATEADD( day, 1 - DATEPART( day, [Date] ),  CONVERT( date, [Date] ) ),
    COUNT(*)
FROM
    [Table]
GROUP BY
    DATEADD( day, 1 - DATEPART( day, [Date] ),  CONVERT( date, [Date] ) )