Question

I’ve got an sqlite table which contains start/stop timestamps. I would like to create a query which returns a total elapsed time from there.

Right now I have a SELECT (e.g. SELECT t,type FROM event WHERE t>0 AND (name='start' or name='stop) and eventId=xxx ORDER BY t) which returns a table which looks something like this:

+---+-----+
|t  |type |
+---+-----+
|  1|start|
| 20|stop |
|100|start|
|150|stop |
+---+-----+

To produce the total elapsed time in the above example would be accomplished by (20-1)+(150-100) = 69

One idea I had was this: I could run two separate queries, one for the “start” fields and one for the “stop” fields, on the assumption that they would always line up like this:

+---+---+
|(1)|(2)|
+---+---+
|  1| 20|
|100|150|
+---+---+

(1) SELECT t FROM EVENT where name='start' ORDER BY t
(2) SELECT t FROM EVENT where name='stop' ORDER BY t

Then it would be simple (I think!) to just sum the differences. The only problem is, I don’t know if I can join two separate queries like this: I’m familiar with joins that combine every row with every other row and then eliminate those that don’t match some criteria. In this case, the criteria is that the row index is the same, but this isn’t a database field, it’s the order of the resulting rows in the output of two separate selects – there isn’t any database field I can use to determine this.

Or perhaps there is some other way to do this?

Answer 1

I do not use SQLite but this may work. Let me know.

SELECT SUM(CASE WHEN type = 'stop' THEN t ELSE -t END) FROM event

This assumes the only values in type are start/stop.