Question

Would it be possible to calculate time online from a table of something like

logtype   time  
----------------
login     2:30  
logout    2:45  
login     3:20  
logout    4:50  
login     5:00  
login     5:10  
logout    6:00  

It would have extra logins because sometimes the server crashes and doesn’t add logout.
Can you do something like this in MySQL only or would you have to use possibly ColdFusion?
And if so, how would you do it in MySQL/ColdFusion?

All I’m wanting is a (rough) total of time online, hours, days, minutes etc for a particular user.

Answer 1

As others have suggested, modifying the table structure is a better approach in the long run. Trying to pair up log records (with no matching key) is difficult, and sometimes impossible. A slightly modified structure would be easier to query, and would yield more accurate results. However, if you absolutely cannot modify the table, it might be possible to calculate a rough total – IF your record ID is numeric/sequential.

The assumption is a “logout” record is always inserted after a “login”. So the logout’s record ID would always be greater than that of the corresponding “login”. A subquery could convert the login information into pairs: ie current and next login record. The corresponding “logout” record id would fall within that range. Once you have matched up login/logouts, use date/time functions to calculate the time elapsed in minutes. Then SUM() the values and format the results however you wish. Note: Obviously proper indexing of the table is a must.

That said, this method only provides an estimate. It has several flaws:

• Some databases do not optimize
  subqueries very well. So even with
  indexes, the query performance can
  be quite poor if there is a lot of
  data involved.
• Relies solely upon the record id
  for identification.
• Does not account for stray logout
  records or erroneous/duplicates. So
  it may be less accurate than other
  methods
• Requires more convoluted sql than
  alternative structures

Anyway, I am not sure of the MySQL syntax off the top of my head. But ignoring the MS Sql Server specific functions, this should be close. The performance in MS SQL was decent with proper indexes, and about 100K records.

SELECT  SUM(DateDiff(n, li.LogTime, lo.LogTime)) AS TotalTimeInMinutes
FROM    (
            SELECT  li.RecordID, li.UserID, li.LogType, li.LogTime,
                   (    SELECT MIN(t.RecordID) FROM  YourTable t
                        WHERE t.UserID   = li.UserID
                        AND   t.LogType  = li.logType
                        AND   t.RecordID > li.RecordID
                    ) AS NextRecordID
            FROM   YourTable li
            WHERE  li.UserID = 123
            AND    li.LogType = 'login'
        ) 
        li INNER JOIN YourTable lo ON lo.UserID = li.UserID
             AND lo.LogType = 'logout'
             AND lo.RecordID BETWEEN li.RecordID AND li.NextRecordID