"scott.marlowe" <scott.marlowe@ihs.com> writes:
> So it would appear to be that the automatic assumptions about what is
> float and what is numeric changed from 7.2 to 7.3, i.e. it's assumed that
> numeric is the input type.
That's correct.
Looking at the code, round(numeric) always rounds xxx.5 values away from
zero (0.5 -> 1, -0.5 -> -1, etc). The behavior of round(float) is
platform-dependent, but round-to-nearest-even is the rule used by IEEE
compliant platforms.
regards, tom lane