sprintf & round (was RE: More random python observations from a perl programmer)

[Tim Peters]

[Tom Christiansen]
>     Python has a round() function, but it seems to ignore the
>     IEEE rules of rounding.  However, its sprintf operator doesn't:
>         >>> print "int of %f is %.0f" % (1.5, 1.5)
>         int of 1.500000 is 2
>         >>> print round(1.5,0)
>         2.0
>         >>> print "int of %f is %.0f" % (2.5, 2.5)
>         int of 2.500000 is 2
>         >>> print round(2.5,0)
>         3.0

Python's sprintf rounding is inherited from the platform libc's.  Virtually
all systems will round 1.5 up to 2.0, but there's wide variation (even among
Unices) in what %f.0 does to 2.5.  The "round" function is Python's own.

[while Christian Tismer gets ...]
> I had a look at this again. Despite the fact that the text in
> the above example is a bit misleading by "int of" where
> a round() takes place, which version of Python/architecture
> did you use? I get the expected result, see below.
>
> >>> print "round of %f is %.0f" % (1.5, 1.5)
> round of 1.500000 is 2
> >>> print round(1.5,0)
> 2.0
> >>> print "round of %f is %.0f" % (2.5, 2.5)
> round of 2.500000 is 3
> >>> print round(2.5,0)
> 3.0
> >>>

My guess is you're running on Windows.  MS rounds all halfway cases up;
Tom's complaint is that under IEEE-754 to-nearest/even rules, 1.5 and 2.5
should both round to 2.  Luckily, Python's will do that the instant MS's
does <wink>.

not-sure-there's-a-winner-in-arguing-over-which-language-has-the-
    lamest-754-support-ly y'rs  - tim