Question

I have a table full of events. I’ve been asked to create an aggregate table of sessions; one session may have several events. Sessions are identified by having the same arrival time. e.g. (This is a simplification, I’m not typing out actual timestamps):

EventID  ArrivalTime  StartTime EndTime StaffID
1        0945         0950      0955    John
2        0945         0955      1000    Barb

Might be turned into something like:

ArrivalTime StartTime EndTime StaffID
0945        0950      1000    ???

With use of MIN(StartTime) and MAX(EndTime) to keep it to a single row.

The problem I’m running into, as the question marks above indicate, is getting a single staff ID – which staff member it is doesn’t matter too much, but I need someone. If it were just a string, as I’ve shown above, it could be done with MIN(StaffID), but the thing that’s doing my head in is that I’m required to look up StaffID in the Staff table and pull out the GUID that’s associated with the short code that’s in my table. And GUID’s don’t like functions like MIN(). Also, just to make matters worse, it’s feasible that the StaffID column in the Events table is NULL, so I have to stick with left joins or similar.

Someone suggested a subquery, but apparently my brain refuses to accept this on a Friday and can’t see how to get that to work.

As a baseline, here’s something along the lines of my current query:

SELECT NEWID() AS SessionID,
e.ArrivalTime,
MIN(e.StartTime),
MAX(e.EndTime),
s.StaffGUID
FROM Events e LEFT JOIN Staff s ON e.StaffID = s.StaffID
GROUP BY e.ArrivalTime, s.StaffGUID

The problem is that if two different staff members are in the list, the session will show up twice. Any ideas?

Answer 1

There’s the correlated subselect option (2000+ for using TOP):

  SELECT NEWID() AS SessionID,
         e.ArrivalTime,
         MIN(e.StartTime),
         MAX(e.EndTime),
         (SELECT TOP 1 s.StaffGUID
            FROM STAFF s
           WHERE s.staffid = e.staffid) AS staffguid
    FROM EVENTS e
GROUP BY e.arrivaltime, e.staffguid, staffguid

…or the derived table/inline view (2005+ for using ROW_NUMBER):

  SELECT NEWID() AS SessionID,
         e.ArrivalTime,
         MIN(e.StartTime),
         MAX(e.EndTime),
         s.staffguid
     FROM EVENTS e
LEFT JOIN (SELECT t.staffid,
                  t.staffguid,
                  ROW_NUMBER() OVER (PARTITION BY t.staffid) AS rank
             FROM STAFF t) s ON s.staffid = e.staffid
                            AND s.rank = 1
 GROUP BY e.arrivaltime, s.staffguid

My preference is for the derived table – correlated subqueries tend not to perform as well.