"Khee Chin" <kheechin@gmail.com> writes:
> I'm not certain if this is a bug or an intended behavior.
> I noticed a discrepency in calculating the remainders across pgsql (PG),
> wolframalpha (W) and using Knuth's description of floored division
We follow whatever the C compiler's idea of modulo is. This is somewhat
standardized, in fact, and I'm not excited about changing it.
Especially not by introducing roundoff-prone double arithmetic into what
had been an exact operation.
regards, tom lane