pricision of string.atof?

[Fernando Pérez]

Elvis Chen wrote:

> 
> Greetings,
> 
> I'm working on some numerical analysis that requires rather good precision
> on calculation.  In my project, I need to read some numbers from a file
> and subsequently process it to give other results.  However, I found that
> string.atof doesn't always convert the read string to equivalent number.
> 
> For example, if I have
> 
> s = 'A 0.424\n'
> splitline = string.split( s )
> A = string.atof( splitline[1] )
> 
> then,
> 
>>>> A
> 0.42399999999999999
>>>> print A
> 0.424

I know this topic gets beaten to death about once a week, but here it goes 
once more. It has nothing to do with atof():

In [1]: a=0.424

In [2]: a
Out[2]: 0.42399999999999999

In [3]: print a
0.424

It's simply that repr() gives a more faithful representation of the internal 
form of a value while str() does some rounding for visual convenience. And 
since not _all_ real numbers can be represented via 32 bit floats (2^32 << 
infinity), some values by necessity have an approximate representation. 

Any decent text on numerical analysis discusses this in detail.

cheers,

f.