tsql join with dateparts in subqueries - sql

Question

I have two queries with subqueries for dateparts which I like to join.
SELECT DateMonth, DateYear, Datestring,
MAX(CouponTotalCount) NoOfCouponsViewed
FROM (
SELECT *, DATEPART(MONTH, DateInsert) DateMonth, DATEPART(YEAR, DateInsert) DateYear,
CONVERT(CHAR(4), DateInsert, 100) + CONVERT(CHAR(4), DateInsert, 120) Datestring
FROM FlurryCouponViewed
) sub
where couponID=249
GROUP BY DateMonth, DateYear, Datestring
SELECT DateMonth, DateYear, Datestring,
MAX(CouponTotalCount) NoOfCouponsRedeemed
FROM (
SELECT *, DATEPART(MONTH, DateInsert) DateMonth, DATEPART(YEAR, DateInsert) DateYear,
CONVERT(CHAR(4), DateInsert, 100) + CONVERT(CHAR(4), DateInsert, 120) Datestring
FROM FlurryCouponRedeemed
) sub
where couponID=249
GROUP BY DateMonth, DateYear, Datestring
The output is of the two queries is:
DateMonth DateYear Datestring NoOfCouponsViewed
----------- ----------- ---------- -----------------
2 2012 Feb 2012 5
3 2012 Mar 2012 12
4 2012 Apr 2012 25
5 2012 May 2012 25
DateMonth DateYear Datestring NoOfCouponsRedeemed
----------- ----------- ---------- -------------------
2 2012 Feb 2012 3
3 2012 Mar 2012 4
4 2012 Apr 2012 5
5 2012 May 2012 11
What I like to achive is two have one joined query giving me:
DateMonth DateYear Datestring NoOfCouponsViewed NoOfCouponsRedeemed
----------- ----------- ---------- ----------------- -------------------
2 2012 Feb 2012 5 3
3 2012 Mar 2012 12 4
4 2012 Apr 2012 25 5
5 2012 May 2012 25 11
How can I do this ?

Answer 1

Make a inner join between the two queries and it should work:
SELECT sub.DateMonth, sub.DateYear, sub.Datestring,
MAX(sub.CouponTotalCount) NoOfCouponsViewed,
MAX(sub2.CouponTotalCount) NoOfCouponsViewed
FROM (
SELECT *, DATEPART(MONTH, DateInsert) DateMonth, DATEPART(YEAR, DateInsert) DateYear,
CONVERT(CHAR(4), DateInsert, 100) + CONVERT(CHAR(4), DateInsert, 120) Datestring
FROM FlurryCouponViewed
) sub
INNER JOIN
( SELECT *, DATEPART(MONTH, DateInsert) DateMonth, DATEPART(YEAR, DateInsert) DateYear,
CONVERT(CHAR(4), DateInsert, 100) + CONVERT(CHAR(4), DateInsert, 120) Datestring
FROM FlurryCouponRedeemed
) sub2 on sub.DateMonth = sub2.DateMonth and sub.DateYear = sub2.DateYear and sub.Datestring = sub2.Datestring
where sub.couponID=249 and sub2.couponID=249
GROUP BY sub.DateMonth, sub.DateYear, sub.Datestring

Answer 2

or a UNION:
SELECT u.DateMonth, u.DateYear, u.Datestring, MAX(u.NoOfCouponsViewed), MAX(u.NoOfCouponsRedeemed)
FROM (
SELECT DateMonth, DateYear, Datestring,
MAX(CouponTotalCount) NoOfCouponsViewed, 0 AS NoOfCouponsRedeemed
FROM (
SELECT *, DATEPART(MONTH, DateInsert) DateMonth, DATEPART(YEAR, DateInsert) DateYear,
CONVERT(CHAR(4), DateInsert, 100) + CONVERT(CHAR(4), DateInsert, 120) Datestring
FROM FlurryCouponViewed
) sub
where couponID=249
GROUP BY DateMonth, DateYear, Datestring
UNION
SELECT DateMonth, DateYear, Datestring, 0 AS NoOfCouponsViewed,
MAX(CouponTotalCount) NoOfCouponsRedeemed
FROM (
SELECT *, DATEPART(MONTH, DateInsert) DateMonth, DATEPART(YEAR, DateInsert) DateYear,
CONVERT(CHAR(4), DateInsert, 100) + CONVERT(CHAR(4), DateInsert, 120) Datestring
FROM FlurryCouponRedeemed
) sub
where couponID=249
GROUP BY DateMonth, DateYear, Datestring
) u
GROUP BY u.DateMonth, u.DateYear, u.Datestring

Answer 3

I would use UNION rather than JOIN
SELECT MonthInserted,
LEFT(DATENAME(MONTH, Datestring), 3) + ' ' + DATENAME(YEAR, MonthInserted) AS DateString
MAX(Viewed) NoOfCouponsViewed,
MAX(Redeemed) NoOfCouponsRedeemed
FROM ( SELECT CouponID,
DATEADD(MONTH, DATEDIFF(MONTH, 0, DateInsert)) [MonthInserted],
CouponTotalCount AS Viewed,
0 AS Redeemed
FROM FlurryCouponViewed
UNION ALL
SELECT CouponID,
DATEADD(MONTH, DATEDIFF(MONTH, 0, DateInsert)),
0,
CouponTotalCount
FROM FlurryCouponRedeemed
) sub
WHERE couponID = 249
GROUP BY MonthInserted
I think a UNION would perform better than the JOIN in the accepted answer as the MAX implies that there are multiple rows per month, and since month is the only common field so it will end up cross joining queries (i.e. if there are 1000 coupons viewed in june 2012 and 500 are redeemed the cross join means you would be selecting the max from 50,000 rows instead of 1500). I am not certain of the schema and logic so this may not be possible, but if there are dates in FlurryCouponRedeemed that are not in FlurryCouponViewed then these will not show.
I also like to keep dates as dates for as long as possible to help the optimiser do it's job, this is why I've replaced DATEPART(YEAR... & DATEPART(MONTH... AND CONVERT(VARCHAR(4), DateInsert... with DATEADD(MONTH, DATEDIFF(MONTH, 0, DateInsert))