> through a version of dijkstra's routing algorithm Just out of interest, what data is this working on?
RBS On Tue, Mar 22, 2011 at 7:25 AM, Amit Chaudhuri <[email protected]> wrote: > [Not at all expert in sqlite but here's a practical example of speed up > using ":memory:" and perhaps a slightly different strategy for getting at > the persistent data.] > > I use sqlite3 with Qt4 / C++ for an application which reads in an undirected > graph and then chunks through a version of dijkstra's routing algorithm. A > colleague runs this on his machine and it takes all night on a large network > running on a database on disk. On my own machine which is more powerful it > probably runs a lot faster but still takes a couple of hours plus. Changing > to an in memory database, reading data in and processing in memory brings > the run time down to a couple of minutes. So yes - running in memory can be > much quicker. At the end of the run I attach an on disk database and copy > out the tables I need to save using "create table select...." . > > A > > On Mon, Mar 21, 2011 at 1:13 PM, Simon Friis <[email protected]> wrote: > >> I know how to create a database that exists only in memory by using >> the :memory: filename. This however, creates a new database every time >> and it can not be saved. >> >> Is is possible to make SQLite load a database file into memory and >> then save it back to the file again when the connection to the >> database is closed? >> >> Would it improve speed? >> >> - paldepind >> _______________________________________________ >> sqlite-users mailing list >> [email protected] >> http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users >> > _______________________________________________ > sqlite-users mailing list > [email protected] > http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users > _______________________________________________ sqlite-users mailing list [email protected] http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
