We help IT Professionals succeed at work.

SQL Query Syntax Cross Join

Medium Priority
383 Views
Last Modified: 2012-06-27
With previous help from the experts I have this working query. It simply counts the number of orders per year for a customer. Now I have been ask to add a second set of columns that will be the identical set of data only for a second customer. They are looking to see a side by side comparison of two customers for the same period of time. I believe a cross join is going to be in order because customer A may or may not order the same items as customer B and vice versa. So the new query will also have to have the custnmbr field as part of the group. How do you restructure this query to list

Item Numb, Desc, Cust Count #1 1stYear, Cust Count #2 1st Year,Cust Count #2 2nd Year,Cust Count #2 2nd Year



      select itemnmbr,itemdesc
      , sum(case when YEAR(sh.docdate)=2011 then 1 else 0 end ) as NumberPerYear2011
      , sum(case when YEAR(sh.docdate)=2012 then 1 else 0 end ) as NumberPerYear2012
      from SOP30300 sl
      inner join SOP30200 sh on sl.SOPNUMBE=sh.SOPNUMBE and sl.SOPTYPE=sh.SOPTYPE
      where custnmbr='195' and sh.SOPTYPE=3 and YEAR(sh.docdate) in (2011,2012)
      group by ITEMNMBR,ITEMDESC
                                            1:
Comment
Watch Question

SharathData Engineer
CERTIFIED EXPERT

Commented:
You can try like this.

;with Customer1 as (
    select itemnmbr,itemdesc
      , sum(case when YEAR(sh.docdate)=2011 then 1 else 0 end ) as NumberPerYear2011
      , sum(case when YEAR(sh.docdate)=2012 then 1 else 0 end ) as NumberPerYear2012
      from SOP30300 sl
      inner join SOP30200 sh on sl.SOPNUMBE=sh.SOPNUMBE and sl.SOPTYPE=sh.SOPTYPE
      where custnmbr='195' and sh.SOPTYPE=3 and YEAR(sh.docdate) in (2011,2012)
      group by ITEMNMBR,ITEMDESC),

Customer2 as (
    select itemnmbr,itemdesc
      , sum(case when YEAR(sh.docdate)=2011 then 1 else 0 end ) as NumberPerYear2011
      , sum(case when YEAR(sh.docdate)=2012 then 1 else 0 end ) as NumberPerYear2012
      from SOP30300 sl
      inner join SOP30200 sh on sl.SOPNUMBE=sh.SOPNUMBE and sl.SOPTYPE=sh.SOPTYPE
      where custnmbr='196' and sh.SOPTYPE=3 and YEAR(sh.docdate) in (2011,2012)
      group by ITEMNMBR,ITEMDESC)
select *
  from Customer1 c1
  full join Customer1 c2
    on c1.itemnmbr = c2.itemnmbr and c1.itemdesc = c2.itemdesc
rwheeler23President

Author

Commented:
What does the

;with   do? Where can I find documentation on this structure?

This works. Thanks.
Scott PletcherSenior DBA
CERTIFIED EXPERT
Most Valuable Expert 2018
Distinguished Expert 2019

Commented:
One note just for general info:

YEAR(sh.docdate) in (2011,2012)

should be replaced with:

sh.docdate >= '20110101' AND sh.docdate < '20130101'

The second style allows an index with docdate to be used, whereas the first style does not.

When you use a function on any column, SQL can no longer use an index to search for the value.
rwheeler23President

Author

Commented:
Thanks for the tip!!!
rwheeler23President

Author

Commented:
I just realized those years may not always be consectutive. That is why I used the 'IN()'
Scott PletcherSenior DBA
CERTIFIED EXPERT
Most Valuable Expert 2018
Distinguished Expert 2019

Commented:
For years 2010 and 2010:

((sh.docdate >= '20100101' AND sh.docdate < '20110101') OR
 (sh.docdate >= '20120101' AND sh.docdate < '20130101'))
rwheeler23President

Author

Commented:
That is true. Good point. The time portion does not cause a problem with this? I remember writing some code and I set the begin date to the same date as the end date and it would not return any results. It was only when there was at least a one day difference would it return anything.
Senior DBA
CERTIFIED EXPERT
Most Valuable Expert 2018
Distinguished Expert 2019
Commented:
Good point about the time portion -- that can be a real pain when dealing with datetime.

It does not cause an issue with the specific style I used --
because note that I am using < the NEXT day, rather than <= the ending day.

That is, instead of this:

sh.docdate BETWEEN '20110101' AND '20111231 23:59:59.997' --every day and time in 2011

I use this:

sh.docdate >= '20110101' AND sh.docdate < '20120101' --every day and time in 2011

That way it continues to work properly even if the datatype of docdate is changed to datetime2 (or smalldatetime).
rwheeler23President

Author

Commented:
Top notch answer. Thank you.

Explore More ContentExplore courses, solutions, and other research materials related to this topic.