Create a form with 3 edit boxes and a button. Edit1 = float such as
12.3456789.  Edit 2= number of decimals to round to such as 3.  Edit 3
= result of this button click.

procedure TForm1.Button1Click(Sender: TObject);
var
   t1, t2, t3: Single;
begin
   t1 := StrToFloat(Edit1.Text);
   t2 := StrToInt(Edit2.Text);
   t2 := exp(t2*ln(10));
   t3 := t1*t2;
   t3 := Round(t3);
   t3 :=t3/t2;
   Edit3.Text := FloatToStr(t3);
end;

edit3 reads the following on my PC:  12.3459997177124

i.e. Round doesnt work or FloattoStr has floating point math error.

Any Comments????

Eric

On 27 May 1996 12:49:01 -0700, morg...@primenet.com (G. Eric Morgan)
wrote:

Actually, the problem is not with either function but when I divide
12345 by 1000 I get 12.3459997177124.  How do I get around this
problem?

Eric

On 27 May 1996 12:49:01 -0700, morg...@primenet.com (G. Eric Morgan)
wrote:

Duncan Murdoch

On Mon, 27 May 1996 20:47:48 GMT, dmurd...@mast.queensu.ca (Duncan

Any clues?

Eric

12.346-12.3459997177124=0.00000029 (approx), which is less than 1 in
1000000 (actually 0.00000029/12.34=2e-8). This does NOT mean that the 6
first digits are correct.

--
Svante Granqvist        Speech, Music and Hearing
Phone +46-8-790 7561    Box 700 14
Fax   +46-8-790 7854    S-100 44 Stockholm
sva...@speech.kth.se    http://www.speech.kth.se/~svante

morg...@primenet.com (G. Eric Morgan) wrote:

Change your Single's to Extended and it'll work fine. I've had never
ending problems with this - Delphi use's Extended for all floating
point calculations, so if you use something else ( eg. Singles or
Doubles ) then you get weird rounding problems as it converts your
values back and forth. The only way I've got around this is to always
use the Extended type.

Hope that helps,

Simon.

On 27 May 1996 18:39:02 -0700, morg...@primenet.com (G. Eric Morgan) wrote:

Components are (or should be) available that can do fixed point arithmetic to
almost any precision. If you only need relatively small numbers within the range
of 15 or 31 bits then you can use normal binary numbers rather than float and
get accurate results. It is only when you need the capability of exceeding that
range that you need floating point. If you need longer numbers then you should
round to the required precision and you should get accurate results.

dick
{*word*106} MacDonald

Use FloatToStrF(Number,ffGeneral,8,8) - or something like that
(not exactly sure about the parameters); and specify number of significant
digits in your converted string LESS than precision of the type you are
using.

Dmitry

Floating point numbers are represented in a BINARY form (not decimals)...=
 so =

a number that can be represented precisly with a by a power of 10 may not=
 be =

represented exactly with a power of 2...

This is an old problem with any application that need really precise =

result... The only overcome you can do is to use a larger precision type =
and =

then round it... Or is to build your own routine that do the calculation =
with =

a a decimal base ... The later will give a MUCH slower result...

Even Supercomputer have the problem ;-) And this can give headache to =

scientist who really need fast precise calculation in decimals...

There is an article in Wired (somewhere in the Autumn '95) that is talkin=
g =

about the implication of that fact...

Hope it help...

-- =

----------
Emmanuel Pirsch
Visitez la Chaumi=E8re Virtuelle : http://www.odyssee.net/~tasm
-------------------------------------------------------------
"Gravitation cannot be held responsible for people falling
in love."
        - Albert Einstein.

On 27 May 1996 12:49:01 -0700, morg...@primenet.com (G. Eric Morgan)
wrote:

Then look up the FloatToStrF function.
--
Ray Lischner                              li...@tempest-sw.com
Tempest Software, Corvallis, Oregon, USA  http://www.tempest-sw.com