On 19 Sep 2011, at 2:42pm, Jaco Breitenbach wrote: > On 19 September 2011 13:23, Simon Slavin <slav...@bigfraud.org> wrote: > >> If you're actually concerned about filesize, then WAL mode is probably not >> the right mode to use. You may be better off with 'PRAGMA journal_mode = >> DELETE'. If your platform stores data on an SSD, then you may be better off >> with 'PRAGMA journal_mode = PERSIST'. For further details see > > My application is currently using the PERSIST journal mode. Actual file > size is not important. I've mentioned the file sizes only to add more > detail and a clearer indication of where I get the numbers from. The > PERSIST mode is used because it seems to have the least overhead of the > available journal modes.
Okay. If I understand the page correctly, then if you have the mode set to 'PERSIST' then you are not using WAL mode. This means SQLite uses the journal file (what you called the log file) very differently and you should ignore the link to <http://www.sqlite.org/draft/wal.html#ckpt> I posted earlier. Also, it means that ... On 19 Sep 2011, at 10:13am, Jaco Breitenbach wrote: > When it comes to checkpointing (every 10 minutes), I assume the log file is > read once in its entirety, and the data file written once for every record > inserted. If you're not using WAL mode, then I don't think the journal file will ever be read in its entirety unless SQLite is trying to recover from a crash or power failure. In this mode the data file is kept pretty-much up-to-date. The journal file contains rollback and recovery information. Simon. _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
