> On Tue, Jul 6, 2010 at 1:05 PM, Martin Knafve > <mar...@hmailserver.com>wrote: > >> Hi, >> >> I'm considering using SQLite as backend for my software. What's stopping >> me is the low insert-performance. If I disable fsync/FlushFileBuffers, >> performance is good but I can't do with the risk of database corruption. >> >> I don't know the details of the WAL-implementation in 3.7, but if >> fsync/FlushFileBuffers are only required when data is moved from the >> write-ahead log to the database file, I would expect performance to be a >> lot faster. Reading the documentation about WAL, I get the impression >> that >> I can accomplish this using the following pragmas: >> >> sqlite3_exec(db, "PRAGMA journal_mode=WAL", NULL, 0, &db_err); >> sqlite3_exec(db, "PRAGMA synchronous=NORMAL", NULL, 0, &db_err); >> > > If you do the above, no fsyncs happen except when doing a checkpoint, and > a > checkpoint should only happen after about each megabyte of inserting. > > We don't have any git support here, so I don't know where you got your > snapshot, but it wasn't an official SQLite. Prerelease snapshots for > SQLite > 3.7.0 are available here: > > http://www.sqlite.org/draft/download.html >
Thanks for the info, I'll have the official snapshot a go. Kind regards, Martin _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
