Problems using Delphi's round function - ROUND(39994.5)=39994, WHY?

Hi, try to use this function instead:

function NormalRound(X:Extended):Int64;
begin
  if Frac(X)>0.5 then Result:=Floor(X) else Result:=Ceil(X);
end;

regards,
Kate

Well rather use this one :)

function NormalRound(X:Extended):Int64;
begin
  if Frac(X)<0.5 then Result:=Floor(X) else Result:=Ceil(X);
end;

Kate

or with
SetRoundMode(rmNearest)

as wavget suggested, then you can keep using the Round function

...or not... did the same thing after I used rmNearest and rmUp... use Katka's method...

I use this:

function Round5(X: Extended): Int64;
begin
  if abs(Frac(x)) <0.5
    then result:= Trunc(x)
    else begin
      if x < 0
        then result:= Trunc(x) -1
        else result:= Trunc(x) +1;
    end;
end;

regards
Pierre

The above function will round as follows:

decimals >= 0.5    -> round up
decimals < 0.5      -> round down

Katka,

I don't think Floor and Ceiling are built in Delphi functions (I think they're visual basic). Well, I can't find it anywere anyway...

Hi PierreC, they're in Math unit :)

Kate

Thx

Hi,

Use this simle function:

function RoundUp(X: Extended): Int64;
begin
  Result := Trunc(X) + Trunc(Frac(X) * 2);
end;

Regards, Geo

hi my 2 cents

http://www.delphibasics.co.uk/RTL.asp?Name=Round

12.4 rounds to  12
12.5 rounds to  12 // Round down to even
12.6 rounds to  13
   
13.4 rounds to  13
13.5 rounds to  14 // Round up to even
13.6 rounds to  14

if you can follow it - i think this could explain part of the problem (if its the same)
https://www.experts-exchange.com/questions/21091818/Rounding-numbers.html

after contacting borland it seems that this issue was resolved in the service pack for Delphi 7

DAvid

Kate, try your function with '-3.1', please. It doesn't work correctly for negative numbers.
wavget, I'm not able to find SetRoundMode function in my D5.

Regards, Geo

SimpleRoundTo from D7 uses asymmetric arithmetic rounding.

function SimpleRoundTo(const AValue: Double; const ADigit: TRoundToRange = -2): Double;

One thing when rounding floating point numbers. Sometimes a simple number 0.15 cannot be represented as 0.15 in floating point so its true representation is 0.149999999999999 This of course will round to 0.1 no mater which rounding function you use. Simple Test:

SimpleRoundTo(0.05, -1); = 0.1
SimpleRoundTo(0.15, -1); = 0.1 should = 0.2
SimpleRoundTo(0.25, -1); = 0.3
SimpleRoundTo(0.35, -1); = 0.3 should = 0.4
SimpleRoundTo(0.45, -1); = 0.5
SimpleRoundTo(0.55, -1); = 0.6

I work for an inventory company and this kind of floating point error is not acceptable. So we had to write our own that added .000001 before rounding to compensate for this.