On Fri, Jan 28, 2011 at 11:38 AM, Pierre Chatelier <k...@club-internet.fr>wrote:
> > But I wondered if I could improve the performance with the following scheme > : > my disk DB is B > I create a memory DB, with the same structure, named A > I attach B to A > then in a a loop > I insert the rows in A > When A is big enough, I flush A into B with "INSERT INTO B SELECT * from > A" > I make A empty > and so on until the input data is exhausted > > But the overall performance is comparable to not using A at all. Is it a > stupid idea (regarding the way sqlite is already optimized) or can I do > something clever with that ? > How you fill your primary key of newly inserted data comparing to the file db ids? I mean if memory ids are always bigger than Max(Id) of the file table, this should be a very quick insert comparable with the speed of harddisk. But if these ids have intermediate values comparing with the existing ones, the speed can drop significantly since sqlite should read/write at virtually random positions of the file db. Max _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
