Sheeri K. Cabral wrote:
The standard doesn't say much about implementation details, and that
seems like an implementation detail to me.
I think rolling back the statement is the right thing to do as well.
I forget, did we keep INSERT IGNORE? If so, I think the big question
is what happens with that? (I'd say it should just not insert that
one row that's null, but not rollback)....
It's clear to me that INSERT IGNORE should ignore all errors, including
implementation errors, misunderstandings, blunders, and intentional
maliciousness. In other words, it's appropriate (and necessary) to fill
a record here and there with random data pretty much out of the blue. I
mean, why restrict it to intentional bad data, or, if you believe there
is no such thing as bad data, then misunderstood data? Generating our
own bad data is icing on the cake.
_______________________________________________
Mailing list: https://launchpad.net/~drizzle-discuss
Post to : [email protected]
Unsubscribe : https://launchpad.net/~drizzle-discuss
More help : https://help.launchpad.net/ListHelp
