How to write string (float) to file?

[Grant Edwards]

In article <1Vbj7.71653$54.3810366 at e420r-atl1.usenetserver.com>, Steve Holden wrote:

> "Rainer Deyke" <root at rainerdeyke.com> wrote in message
> news:G89j7.54140$c8.24141594 at news1.denver1.co.home.com...
> [ ... ]
>>
>> > >>>> foo(10000000000000000000000000000000000000000l, 2)
>> > > '5000000000000000200000000000000000000000.00'
>>                      ^
>>
>> Not just incorrect, but the incorrect digit is right in the middle of the
>> number, not in the least significant place.
>
> Well by my count it occurs in the seventeenth decimal place,
> which given the precision on floats is about what you ought to
> expect. Clearly if you need to deal with numbers of arbitrary
> precision you are going to *have* to represent them as longs
> and keep a scaling factor (or use some other similar
> technique).
> 
> However, I would say that for most normal accounting purposes
> (i.e. numbers up to the tens of millions, significant to 2 DP,
> or ten significant figures) floating point numbers will be
> perfectly satisfactory *provided roundoff error is controlled
> at each step*. [This will also be good enough for the OMB,
> since they seem to round everything to the nearest million
> anyway :-)]

For accounting stuff I think you'd be far better off using
integer math in cents or mils or something like that.  Or
find/write a BCD floating point package.

But, I am not an accountant, and my head almost exploded the
time somebody explained what "good will" meant...

-- 
Grant Edwards                   grante             Yow!  My mind is making
                                  at               ashtrays in Dayton....
                               visi.com