> but I am wondering if there is any > (somewhat) easy way to get direct fd access and maintain asyncronicity without > threads.
How do you see it's possible? Even if we assume that all files were opened in async mode and only aio functions were used what would that mean? You are calling sqlite3_step to change some data, SQLite calls aio functions to write all changed database pages to disk, then what? sqlite3_step returns and you continue to do something even though transaction is not committed yet? When and how should SQLite call fsync to ensure that journal is on the disk, to ensure that all changed database pages are on the disk? If you don't care about that just execute 'pragma synchronous = off' and your OS kernel will do async writes to disk for you without any user-space threads and without any changes to SQLite code base. Pavel On Wed, Feb 9, 2011 at 12:50 AM, Mike Blumenkrantz <m...@zentific.com> wrote: > Hi, > > I am writing an open source database library, and I would like to add an > sqlite > backend to it. I have checked out the documentation and found information > about > using sqlite in async mode with threads, but I am wondering if there is any > (somewhat) easy way to get direct fd access and maintain asyncronicity without > threads. > > I have read through the source a bit, and I am guessing that this is > unlikely to be possible due to the heavy use of mmap, but I thought I would > mail > and ask anyway :) > > -- > Mike Blumenkrantz > Zentific: NULL pointer dereferences now 50% off! > _______________________________________________ > sqlite-users mailing list > sqlite-users@sqlite.org > http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users > _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
