> > On Thu, Dec 11, 2014 at 10:58 AM, Paul <de...@ukr.net> wrote: > > > > > > > > I have yet to try and test if dropping stat tables worth the effort. > > > > > > > Most of the work is involved in loading sqlite_stat4. On the other hand, > > most of the benefit comes from sqlite_stat1. So consider compiling without > > SQLITE_ENABLE_STAT4. You will still probably get good query plans, but the > > startup time should be reduced. > > > > Thanks you for analysis of the data, Richard. > I am going to try your advice and post the results. >
I've done different tests, and my results are: - Compiling without stat4(3) is equivalent to simply dropping sqlite_stat4(3) table, and as stated before, on average it takes 1.2 ms (with) vs 0.4 ms (without) to open database. That is very good, because I don't have to compile two different version of library. - It takes roughly same time to read sqlite_stat3 as it takes to read sqlite_stat4, sqlite_stat3 wins by a few microseconds. So if I'd ever choose to ignore additional open overhead, I'd definitely go for sqlite_stat4. - Compiling with USE_PREAD macro saves a dozen more microseconds. I'll stick to the advice and drop sqlite_stat4 table, leaving only sqlite_stat1. Thanks everyone for help! Best regards, Paul _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users