On Mon, Jul 4, 2016 at 1:18 PM, R Smith <rsm...@rsweb.co.za> wrote: > On 2016/07/04 1:13 PM, Dominique Devienne wrote: > >> On Mon, Jul 4, 2016 at 12:08 PM, Hick Gunter <h...@scigames.at> wrote: >> >>> A simple UPDATE <table> set <field>=<value> where <key_condition> will >>> translate into about 40 instructions; adding your trigger adds an >>> additional estimated 400 instructions of trigger program. >>> >> But the question is whether these extra VDBE instructions can cause >> slowdowns, even when not executed at runtime? (i.e. branch using them is >> not taken). > > The slowness is not due to the Trigger code being skipped or executed... > The slowness is resulting from the trigger code having to be compiled into > the statement. Whether or not it gets executed might affect another layer > of slowdown, but not in the OP's case.
Ah ah, I see it now, thanks! --DD PS: I never worry about SQLite's parser performance (because I'm binding in part), but in this case it's akin to compiling a dozen different statements, with the associated name lookups and execution plan building, so it could add up if done in a tight loop indeed. Great insights. Thanks again. _______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
