On 3/4/16, tonyp at acm.org <tonyp at acm.org> wrote: > I ran some tests and almost all of my databases (about 100 of them with > different content mix and with the biggest one being around 500MB) inflated
Can you run sqlite3_analyzer on some of your databases and send me the output? > > by a lot while only a couple or so shrunk in size by not much. (It could be > > that I'm just unlucky with my database contents and this change simply > happens to affect me in a negative way while for most users the results > would be different.) > > For me, this effect alone is enough reason to not wish a change of the > default size as I would always have to manually change it back. Database > size is more important (to me) than access times. > > What is the expected gain of this change supposed to be (not only in terms > of database size but in general)? Is it just for hypothetical access speed In my tests, a 4096-byte page runs about 5% faster (measured using cachegrind) than a 1024-byte page. Real-world timings are harder to make, because on a modern machine, the same program will vary in performance by 5% from one run to the next, but I think I'm seeing closer to 10% faster. Nothing about this change forces you to use a larger page size. The "PRAGMA page_size=1024" command still works fine, as does the -DSQLITE_DEFAULT_PAGE_SIZE=1024 compile-time option. > > improvements? And, if so, for what size databases? (I would suspect only > for the very large ones for which usually one wouldn't normally choose > SQLite3.) > > (BTW, will this change also affect FOSSIL repos?) > > -----Original Message----- > From: Richard Hipp > Sent: Friday, March 04, 2016 5:48 PM > To: General Discussion of SQLite Database ; sqlite-dev > Subject: [sqlite] Changing the default page_size in 3.12.0 > > The tip of trunk (3.12.0 alpha) changes the default page size for new > database file from 1024 to 4096 bytes. > > https://www.sqlite.org/draft/releaselog/3_12_0.html > https://www.sqlite.org/draft/pgszchng2016.html > > This seems like a potentially disruptive change, so I want to give > you, the user community, plenty of time to consider the consequences > and potentially talk me out of it. > > The "Pre-release Snapshot" on the download page > (https://www.sqlite.org/download.html) contains this change, if you > want to actually evaluate the latest changes in your application. > > We hope to release 3.12.0 in early April, or maybe even sooner, so if > you want to provide feedback, you should do so without unnecessary > delay. > -- > D. Richard Hipp > drh at sqlite.org > _______________________________________________ > sqlite-users mailing list > sqlite-users at mailinglists.sqlite.org > http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users > > _______________________________________________ > sqlite-users mailing list > sqlite-users at mailinglists.sqlite.org > http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users > -- D. Richard Hipp drh at sqlite.org
