in reply to
Re^3: Typeless Relational Database
in thread Typeless Relational Database
So it's your contention that "DAYOFMONTH(...)" is not a type-conversion? I see one type going in (date) and a different type coming out (number). Or do you mean to say that only type-conversions that are built-in functions (but excluding the to_X(...) ones) are ok to use? Seems like an arbitrary line. Also, I doubt whether the makers of RDBMS have figured out every possible built-in type-converting function that might make sense to exist (good RDMSs allow developers to define their own functions for a reason).
As for the second, yes, I know full well that it's an optimization. An ugly one, in fact. The sort of thing that I don't even write out by hand, but have perl code generate for me. However, I don't think that makes it "wrong". Premature optimization is wrong, sure, but necessary optimization is... well... necessary. And it is hardly "wrong".
I think I'm sounding upset, and that's not how I mean it. I'm not upset at all, nor am I trying to turn this into a heated argument. I was just being pedantic about the absolutism of your statement that type-conversion is ALWAYS wrong. I heartily agree that type-conversion is *usually* wrong. But it's a lot less always wrong than goto, for example, and even goto has, like, one or two legitimate uses.
Not an editor command: Wq