Hrounding error

[cooperq]

Hi,
I am new to python. I was messing around in the interperator checking
out the floating point handling and I believe I may have found a
rounding bug:

>>234 - 23234.2345

-23000.234499999999
This is not correct by my calculations.

I am using python 2.5.2 in ubuntu 8.04. I am wondering if this is a
known bug, or if I am just not understanding some feature of python.
Thanks for the help!

-Cooper Quintin
http://www.bitsamurai.net

[casevh]

On Jun 17, 10:47*pm, [e-mail] wrote:

Hi,
I am new to python. *I was messing around in the interperator checking
out the floating point handling and I believe I may have found a
rounding bug:

>234 - 23234.2345

-23000.234499999999

This is not correct by my calculations.

I am using python 2.5.2 in ubuntu 8.04. *I am wondering if this is a
known bug, or if I am just not understanding some feature of python.
Thanks for the help!

-Cooper Quintinhttp://www.bitsamurai.net

It is a side-effect of binary floating-point.

http://www.python.org/doc/faq/genera...-so-inaccurate

casevh

[Aidan]

co*****@gmail.com wrote:

Hi,
I am new to python. I was messing around in the interperator checking
out the floating point handling and I believe I may have found a
rounding bug:

>>>234 - 23234.2345

-23000.234499999999
This is not correct by my calculations.

I am using python 2.5.2 in ubuntu 8.04. I am wondering if this is a
known bug, or if I am just not understanding some feature of python.
Thanks for the help!

-Cooper Quintin
http://www.bitsamurai.net

If you need to round before you display the result, I find the following
works:

>>'%.4f' % (234 - 23234.2345)

'-23000.2345'

The %.4f means that the interpolated value will be treated as a float,
rounded to 4 digits after the decimal point.

[Jerry Hill]

On Wed, Jun 18, 2008 at 1:47 AM, <co*****@gmail.comwrote:

>>>234 - 23234.2345

-23000.234499999999

This is not correct by my calculations.

Python floating point operations use the underlying C floating point
libraries which, in turn, usually rely on the hardware's floating
point implementations. This Wikipedia article talks about how those
values are usually stored and manipulated:
http://en.wikipedia.org/wiki/IEEE_fl...point_standard

So, which IEEE double precision floating point value would you like
instead? As far as I understand it, these are your two choices:

'ba490c020f76d6c0' = -23000.234499999999
'bb490c020f76d6c0' = -23000.234500000002

Alternatively, investigate the Decimal module, but keep in mind that
it can have the same sorts of issues, just on different numbers.
Compare, for instance:

>>(1.0/3.0) * 3.0

1.0

>>from decimal import Decimal
( Decimal('1.0')/Decimal('3.0') ) * Decimal('3.0')

Decimal("0.9999999999999999999999999999")

>>>

--
Jerry

[cooperq]

On Jun 18, 8:02*am, "Jerry Hill" <[e-mail]:

On Wed, Jun 18, 2008 at 1:47 AM, *<[e-mail]:

>>234 - 23234.2345

-23000.234499999999

This is not correct by my calculations.

Python floating point operations use the underlying C floating point
libraries which, in turn, usually rely on the hardware's floating
point implementations. *This Wikipedia article talks about how those
values are usually stored and manipulated:http://en.wikipedia.org/wiki/IEEE_fl...point_standard

So, which IEEE double precision floating point value would you like
instead? *As far as I understand it, these are your two choices:

'ba490c020f76d6c0' = -23000.234499999999
'bb490c020f76d6c0' = -23000.234500000002

Alternatively, investigate the Decimal module, but keep in mind that
it can have the same sorts of issues, just on different numbers.
Compare, for instance:

>(1.0/3.0) * 3.0

1.0

>from decimal import Decimal
( Decimal('1.0')/Decimal('3.0') ) * Decimal('3.0')

Decimal("0.9999999999999999999999999999")

--
Jerry

So it seems then that python might not be very good for doing
precision floating point work, because there is a good chance its
floating points will be off by a (very small) amount? Or is there a
way to get around this and be guaranteed an accurate answer?

[casevh]

>

So it seems then that python might not be very good for doing
precision floating point work, because there is a good chance its
floating points will be off by a (very small) amount? *Or is there a
way to get around this and be guaranteed an accurate answer?- Hide quotedtext -

It's not a Python problem. That is just the behavior for floating-
point arithmetic.

casevh