SUM with GROUP BY don't display zero sums - sql

Question

I'm trying to get data from a single table. I grouped by CURR. I have 12 condition listed. But some are zero.
SELECT ISNULL(SUM(AMOUNT), 0) AS TOPL,
CURR
FROM XXX
WHERE DATEPART(year, CONVERT(dateTime, DATE)) = 2018
AND DATEPART(month, CONVERT(dateTime, DATE)) = 1
GROUP BY CURR
This is returning 3 value. But I want 12 value including zero sums. I tried this with CASE, but I could not.
Thanks...

Answer 1

What the others have been trying to tell you is that you need a "list" of the CURR values you want to see in your results. Generally, these would come from another table in a properly normalized database. Do you have one? It seems not but it is worth asking. A properly normalized database would generally have one.
So how do you create this list dynamically? Let us assume that your existing table has at least one row for every CURR value you desire in your resultset - even if that row has a date that falls outside of your period of interest. We can use that to form this list and then outer join that list to your existing query that does the summing.
with curr_list as (select distinct CURR from dbo.XXX)
select curr_list.CURR,
sum(isnull(tbl.AMOUNT)) as TOPL
from curr_list left join dbo.XXX as tbl
on curr_list.CURR = tbl.CURR
and tbl.[DATE] >= '20180101'
and tbl.[DATE] < '20180201'
group by curr_list.CURR
order by curr_list.CURR;
That should work assuming I made no typos. The CTE (named curr_list) creates the list of ID values that you want to see in your resultset. When you outer join that to your transaction data you will get at least one row for each CURR value. You then sum the amounts to aggregate those rows into a single row for each CURR value. Notice the change to the date criteria. Your original approach prevents the optimizer from using any useful indexes on that column.
If your existing table does not have a row for every value of CURR you want in your results, then you can simply change the cte and hardcode the values you desire.

Answer 2

I may get what you are trying to do.
If you have 12 CURR (I'm guessing currency?) and want to get the total amount of the transaction in January 2018 grouped by currency.
If this is the case, here is what you should try to do :
SELECT ISNULL(SUM(AMOUNT), 0) AS TOPL,
CURR_TABLE.CURR
FROM CURR_TABLE
LEFT JOIN XXX ON XXX.CURR = CURR_TABLE.CURR
WHERE DATEPART(year, CONVERT(dateTime, XXX.DATE)) = 2018
AND DATEPART(month, CONVERT(dateTime, XXX.DATE)) = 1
GROUP BY CURR_TABLE.CURR
That way you'll get all currency listed (even if no reccord of that currency is available for that month.
EDIT:
I don't like that kind of syntax when you can avoid it, but you can :
SELECT ISNULL(SUM(AMOUNT), 0) AS TOPL,
CURR_TABLE.CURR
FROM (SELECT distinct CURR FROM XXX) CURR_TABLE
LEFT JOIN XXX ON XXX.CURR = CURR_TABLE.CURR
WHERE DATEPART(year, CONVERT(dateTime, XXX.DATE)) = 2018
AND DATEPART(month, CONVERT(dateTime, XXX.DATE)) = 1
GROUP BY CURR_TABLE.CURR

Answer 3

Please execute below query and compare with your expected output.
SELECT CURR, SUM(ISNULL(AMOUNT,0)) AS TOPL,
CURR
FROM XXX
WHERE DATEPART(year, CONVERT(dateTime, DATE)) = 2018
AND DATEPART(month, CONVERT(dateTime, DATE)) = 1
GROUP BY CURR