Python wierdness

Fredrik Lundh fredrik at pythonware.com
Wed Feb 7 23:06:18 CET 2001


Roger wrote:
> In my Python musings I came across a bit of seeming wierdness

not Python's fault, really.

>     >>> x % 0.1
>     0.1
>     >>>
>     >>> # Obviously, x modulo 0.1 is indeed equal to 0.1.

obviously, x modulo 0.1 *prints* as 0.1 under 1.5.2.
doesn't mean that it contains exactly the same bits,
does it?  let's see:

>>> (0.3 % 0.1) - 0.1
-2.77555756156e-017

nope.

>     ... # If we pass it to foo(), we should see "bar", right?

obviously not.

>     >>> # Hmmm... "no bar."  So foo() doesn't agree that
>     ... # x modulo 0.1 is equal to 0.1, even though the
>     ... # command line shows that it is.  Why not?

full story here:
http://www.python.org/cgi-bin/moinmoin/RepresentationError

also note that Python 2.0 shows more precision:

    >>> x = 0.3
    >>> x % 0.1
    0.099999999999999978

> What's going on here, and how could one get around it?

unless you know exactly what you're doing, don't use "=="
on floating point values.  alternatives:

    if abs(a - b) < some_small_value:
        ...

or even

    if str(a) == str(b):
        ...

Cheers /F





More information about the Python-list mailing list