"Doug" <[EMAIL PROTECTED]> wrote in message news:[EMAIL PROTECTED] > Thank you Igor. The GROUP BY was the secret (I was trying to do a > GROUP BY on the outer select, but it wasn't quite cutting it). > > GROUP BY is very powerful, but I notice it has a performance cost. > Is there a way to use an index with it? I have EventTime indexed and > that index is being used. I suppose creating a combined index of > EventTime, ProcessID and FileName might help because the underlying > record wouldn't need to be looked up. Any thoughts on that idea?
No, I don't see how such an index would help. Since EventTime is the first component, this index can't be used to enumerate rows in the order defined by (ProcessID, FileName) pair. Imagine two records with the same ProcessID and FileName but with EventTimes that are far apart. Such records won't be adjacent in your index. You could create an index on (ProcessID, FileName). But SQLite can only use one index in a single query. You should test and measure which of the two indexes results in better performance. Igor Tandetnik _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
