On Tue, 22 Jun 2004, Mauricio CASTRO wrote:
> You are right!
>
> I am doing something wrong? I understand that there is a difference but I
> don't know why.
Floating point math is inexact. The date_part(...) is returning a value
that is very close to, but not exactly 4.031 due to accuracy limitations.
So, when you multiply it by 1000 you can get a value very close to, but
not exactly 4031. When it tries to compare it to the bigint 4031, it
converts the bigint to a double as well and gets a different double from
the first.
