Help regarding a select Query

Question

hi All,
I have one requirement.I wonder whether i can do it in a single SQL Statement or not.
I have tab1 with columns c1,c2,c3,c4,....,Logon_time
I have tab2 with columns C1,c2,c3,c4.....,Logoff_time
All the columns are same for two tables except that logon_time and logoff_time as mentioned above.
tab1 will have all the information of user who logon to database.
tab2 will have all the information of user who logoff from DB.
Now i i have to generate a report which shows all the user info and logon Duration.For that i can use logoff_time - logon_time.but the challenge here is logoff_time dont contain rows for the users who logon to database byt yet to logoff.
So for them logoff_time-logon_time showing -ve Duration.
So I want Duration as logoff_time-logon_time for all the rows.But for those users who yet to logoff i want duration as "YET to Logoff"
Is this possible with single Select Query.
Thanks
Pramod

Answer 1

Hi, Pramod,
You need to join tab_1 to tab_2. Since not every row in tab_1 will have a matching row in tab_2, make it an outer join.
How do you know which row in tab_1 matches which row in tab_2?
I assume there's some column (let's say c1) that identifies a user. A user can have several sessions going at the same time, however.
The following query matches the n-th logon_time in tab_1 with the n-th logoff_time in tab_2 for the same user:
WITH     numbered_tab_1     AS
     SELECT  c1, c2, c3, c4, ..., logon_time
     ,     ROW_NUMBER () OVER ( PARTITION BY  c1     -- or whatever identifies user
                               ORDER BY          logon_time
                       ) AS r_num
     FROM     tab_1
,     numbered_tab_2     AS
     SELECT  c1, c2, c3, c4, ..., logon_time
     ,     ROW_NUMBER () OVER ( PARTITION BY  c1     -- or whatever identifies user
                               ORDER BY          logoff_time
                       ) AS r_num
     FROM     tab_2
SELECT     t1.c1, t1.c2, t1.c3, t1.c4, ..., t1.logon_time     -- or whatever you want
,     t2.c1, t2.c2, t2.c3, t2.c4, ..., t2.logoff_time
,     NVL ( TO_CHAR (24 * (t2.logff_time - t1.logon_time))    -- duration in hours
         , 'Yet to Logoff'
         )         AS duration
FROM             numbered_tab_1     t1
LEFT OUTER JOIN     numbered_tab_2     t2     ON   t1.c1     = t2.c1
                                 AND  t1.r_num  = t2.r_num
;I can test this if you'll post some sample data and the results you want from that data.
The query above matches logons and logoffs only by user and time
For example, say I log on at 10:00.
At 10:45, I open another window and log on again, but log off at 10:50 after 5 minutes.
At 11:30 I log off the first session.
If you were interested in sessions that lasted over 1 hour, the query above would not be accurate. It would show one session starting at 10:00 and ending at 10:50, and another session starting at 10:45 and ending at 11:30. To know that the first session to end was the second sessiion in order of beginning, you would have to have another column, say a unique session id.
If you're only interested in the number of sessions, or the total time connected per user, then the query above will be good enough.
Edited by: Frank Kulash on Jul 6, 2009 12:07 PM