On 8 Feb 2011, at 4:22pm, Sven L wrote: > Thank you very much for your detailed explanation!
You're welcome. I've actually never had to think out this feature of SQLite before, so it was interesting for me too. I hope Richard or the rest of the team will correct me if I got anything wrong. > I will comment out my calls to ANALYZE, and see how my software performs. > > The reason why I added it in the first place is that users are allowed to > create their own queries, and since not all of them are SQL experts, I wanted > the engine to be as tolerant as possible. Perhaps it's a better idea to > simply crave decent SQL! Allowing users to make up their own queries on the fly does make things a little more difficult as you worked out: you can't pre-make good indexes. The big server/client databases cache temporary indexes (and share them between users), so they handle unexpected queries far better: if any query comes in that doesn't suit any indexes it simply makes up a new temporary index and keeps it in case it's needed later. SQLite can't do this because it's designed for a tiny footprint and can't chew up lots of memory or disk space without a good reason. But your initial questions did sound a little like premature optimisation and I think you'll get decent results without worrying too much about it. I'm sure the overwhelming number of SQLite users have never used ANALYZE even once. Simon. _______________________________________________ sqlite-users mailing list [email protected] http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
