Are you sure the structure on disk is the same as in memory? My initial guess is that you have some indexes that need to be updated when you execute your inserts/updates. However, if the structure is exactly the same on disk and in memory I'm left clueless.
Can you provide me with a copy of the data and the code that does the inserts? If so, I would test to see if I get the same behaviour on Linux x86 and x86_64. Samuel On 2007-12-21 10:13, [EMAIL PROTECTED] wrote: > hi all, > > i got a question concerning in-memory-behaviour of sqlite. > > this is what i'm trying to do: > > i load an existing file db into a memory one (create a :memory:-connection, > attaching the file db which has ~ 60 mb and then copying its tables and > indices and detaching the file db - this works quite great so far, lasts > around 5 secs). now i begin a transaction on the in-memory-db and start > executing my queries (some 10k inserts, updates and deletes). and this > process is horribly (sic!) slow. > > in comparison i did all the inserts into an empty memory-db (without > cloning the whole file db first) which is faster by around factor 1000... > > anyone has a clue what happens there? i switched off windows virtual memory > paging - the same effect. disk io is near zero, it's really the db commands > only that causes this huge difference... working on the file db is much > faster than on memory given the same process. do i have to pragma sth > special (tried almost everything here, too...) > > thnx for your help, > pong *stuck ----------------------------------------------------------------------------- To unsubscribe, send email to [EMAIL PROTECTED] -----------------------------------------------------------------------------
