On Wed, 30 Aug 2006, Mario Frasca wrote: > On 2006-0829 13:15:02, [EMAIL PROTECTED] wrote: > > >> To my surprise (perhaps "horror") I find that SQLite has > >> for a very long time allowed NULL values in PRIMARY KEY > >> columns. [...] > > I understand your concern about legacy programs, but most of us expect > PRIMARY KEY to imply NOT NULL... don't we? what about looking for > alternative good solutions? we could put the correction code in the > source, conditionally compiled (not the default) and with the next major > release reverse the condition (new 'corrected' source becomes default > and old 'legacy' behaviour still available if desired) ... ? > > maybe 'the best of both worlds', hope you agree. > > regards, > Mario
<MHO> I have to go along with Mario, here. This is a potential show stopper, and I would grab a "fixed" version of SQLite ASAP were it made available. Alternatively, one might simply use an ifdef to restore the old (legacy) behaviour, something like: #ifdef WEIRD_AND_UNEXPECTED_BEHAVIOURS_DESIRED ... #endif I'd suggest that this is one for the regression suite as well. </MHO> Rob Sciuk ----------------------------------------------------------------------------- To unsubscribe, send email to [EMAIL PROTECTED] -----------------------------------------------------------------------------