Comparison of variables of different type with conversion returns incorrect result

Question

I have the following code

var
  Valor1: string;
  Valor2: Double;
begin
  Valor1 := '150.15';
  Valor2 := 150.15;

  If StrToFloat(Valor1) = Valor2 then
     ShowMessage('Iguais');
end;

In this situation, sometimes Delphi understands that the values are different, however, if I use the code below they are considered equal

    var
      Valor1: string;
      Valor2: Double;
  Valor2: Double;
    begin
      Valor1 := '150.15';
      Valor2 := 150.15;
      Valor3 :=  StrToFloat(Valor1);

      If Valor3 = Valor2 then
         ShowMessage('Iguais');
    end;

Note: in the cited example I am setting the values directly to the variable, however, the problem I had occurred from reading a serial data, but even if I use the function FormatFloat It kept happening, and if I gave a debug in the variables the values were equal, and yet were not considered equal in the if

Answer 1

If you need exact equality you can’t use one Double. It is not possible in all cases. It has a binary format that cannot represent all possible numbers, so something that looks the same may not be, it depends on the form it was obtained. Behold What is the correct way to use the float, double and decimal types?.

I’m not sure about Delphi because there’s a difference. Either this code is a little different from the real where it got the result, or the language has some optimization that brings a different result, which is strange, after all the fact of storing in a variable should not change the semantics, but each language with its preference, We’ve seen on this page that there’s weird stuff on Delphi. You may be misreading something.

Extendend does not solve the problem, after all he is a Double more accurately, but still inaccurate, at best disguises it, which is worse. It will work, but it won’t be right. It may never go wrong, but by definition it can.

It has to solve if to spread roundabouts for the whole code, but it is Gambi of the thick.

The only real solution is a decimal type instead of binary, the binary exponent does not allow all values (2) that we are used to (10).

I am not familiar with the implementation of Currency of Delphi, he would be the solution, but by the observation of Bacchus he was poorly done (we are not both sure). A decimal type should be in essence an integer, not necessarily Integer with a built-in exponent indicated to decimals (has other ways to implement this). There it works and is right.

I have a reply which is kind of a compendium on the subject.

Correct example in Java.

See more.

Answer 2

Whether to use the function StrToFloat consider passing it a string with ',' and not '.' to decimal separator.

Another detail is the Variable Type that will receive the result.

In the System.SysUtils we have:

function StrToFloat(const S: string): Extended;

var
  Valor1: string;
  Valor2: Extended;
begin
  Valor1 := '150,15';
  Valor2 := 150.15;

  If StrToFloat(Valor1) = Valor2 then
     ShowMessage('Iguais');