Hi, I'm working on integrating sqlite into Mozilla. The problem is that some Linux users have their profile (and hence, database) over a network. Ignoring the safety issues for a moment, I'm trying to get good performance. Sometimes it takes the network "a while" to read a page, so I'm trying to read as few pages as possible.
I notice that it removes pages from the cache after each transaction. Why is this done instead of letting them hang around for the next transaction? When starting a new transaction, it could of check the file for modifications and only clear the cache if the file has been touched by another connection. This would really save us a lot of time since in most cases there is only one connection and most of the database should fit in the cache. If this is doable, how much effort do you think would be required us to change the cache clearing mechanism? Thanks, Brett