Question

I would prefer to use a JOIN for this, but none of the solutions are either viable due to performance impacts or they don’t return the right result set.

The RunTime column is a UNIX timestamp captured every night at midnight. Not every entry will occur every night, meaning an entry for Entry1 may have occurred two days ago, but not today.

Schema:

 `ID` int(11) unsigned NOT NULL AUTO_INCREMENT,
  `Name` varchar(32) NOT NULL,
  `Category` int(11) NOT NULL,
  `RunTime` int(11) NOT NULL,
  UNIQUE KEY `ID` (`ID`),

Example Desired Results:

+-------------+----------------+----------+
| Name        | LastRunTime        | Count    |
+-------------+----------------+----------+
| Random Name |     1339131600 |       73 |
| RandomName2 |     1337131600 |       13 |
... etc

Essentially, my working query looks like the below. It will query the table for yesterdays data.

select Name,
       RunTime AS LastRunTime,
       count(*) AS Count
from TABLE
where RunTime = 
    (
    select DISTINCT( RunTime ) from TABLE WHERE r.Name=Name order by RunTime LIMIT 1,1
    ) 
and Category in 
    (
    select AnotherTable .ID from AnotherTable where AnotherTable.Status = 1
    )

group by Name

The problem is, this query is slow. I want to move away from the first subquery, on the RunTime column, but the LIMIT and association is getting the best of me. The above takes a very, very, long time.

Does anyone have an example of a way to:

Get the second most recent RunTime, the count of rows for the second most recent RunTime, quickly and efficiently where RunTime is not consistent across all rows?

Any suggestions are appreciated!

Answer 1

I stuck with a subquery but changed the timeframe to WHERE RunTime = (SELECT MAX(RunTime) - 86400....

All other solutions or attempts where either too demanding or convoluted.