On 2016/04/14 10:23 PM, Cecil Westerhof wrote: > 2016-04-14 22:10 GMT+02:00 Richard Hipp <drh at sqlite.org>: > >> On 4/14/16, Cecil Westerhof <cldwesterhof at gmail.com> wrote: >>> ?Yes that makes sense. But could not a type of PRAGMA be used? So if the >>> PRAGMA is not defined the old functionality and your historical data is >>> save. And if the PRAGMA is defined the new functionality. >>> >> Easier: Just declare the columns in question as NOT NULL. Or use a >> WITHOUT ROWID table which *does* enforce NOT NULL. >> > ?The problem is that you do not expect that values in the primary key can > be NULL, so you do not use NOT NULL. I now know that I need to do this and > I do it now, but the unaware will be bitten, just as I was. It was not a > big problem, because I just started playing with SQLite, but when you have > a very big database which gets a lot of data every day and a year later you > find out you are bitten by this ?
What makes you "expect" this? Nothing in the SQLite documentation I hope. Is it perhaps habit based on how some other DBs do it? How will "those who don't know about NOT NULL" be better serviced by a pragma which they also don't know about? The documentation is very clear on the matter, even a cursory glance would educate "the unaware". Making more things with more documentation to read, in order to help people with other documentation they didn't read - not a salient course of action I think. Cheers, Ryan