Features of the mql5 language, subtleties and tricks - page 242

[Maxim Kuznetsov]

an old feature of Min/Max in MQL - they work the way the developers think, i.e. correctly :-)

it's the standards that are wrong.

[Nikolai Semko]

Nikolai Semko #:

ahhh..., then why was it so confusing about " Mathematically the same expression (highlighted) " and union
was enough:

but since 0.0 == -0.0, then I can't imagine a situation where this could be a problem and, in fact, affect anything at all

If it wasn't zero, then it would be a problem. And at zero it doesn't affect anything.
For:

x * 0.0 = 0.0
x * -0.0 = -0.0
x + -0.0 = x
x - -0.0 = x
pow(x,-0.0) = 1
log(0.0)  // -inf
log(-0.0)  // -inf

[Nikolai Semko]

c++ - What is the difference between -0 and 0? - Stack Overflow

So, all you need to know is that 0 = -0
That's the whole feature that doesn't affect calculations and logic

[Denis Kirichenko]

fxsaber #:

I have an excellent understanding of the question, so I wrote both versions of MathMin to show that mathematically identical functions in programming languages produce different results.

template <typename T>
T ToType( const double Num )
{
  union UNION
  {
    T Num1;
    double Num2;
  } Union;
  
  Union.Num2 = Num;
  
  return(Union.Num1);
}

void OnStart()
{
  Print(ToType<long>(MathMin(-0.0, 0.0))); // 0
  Print(ToType<long>(MathMin(0.0, -0.0))); // -9223372036854775808
}

It seems to be a bug.

It works fine in C++:

#include <iostream>

using namespace std;
//
template <typename T>
T ToType(const double Num)
{
    union UNION
    {
        T Num1;
        double Num2;
    } Union;
    Union.Num2 = Num;
    return(Union.Num1);
}
//
int main()
{
    double positive_zero_val, negative_zero_val;
    positive_zero_val = 0.0;
    negative_zero_val = -0.0;
    //---
    long long_negative_zero_val = ToType<long>(negative_zero_val);
    printf("\nLong negative zero = %d", long_negative_zero_val);
    cin.get();
}

in the console:

Long negative zero = 0

Question in passing to the developers. Is it normal that in the debugging window the Union variable does not expand and does not react at all to a click?

The fields Union.Num1 and Union.Num2 were added manually. At least this way you can see the values...

[amrali]

string DoubleToHexadecimal(const double value)
  {
   return StringFormat("0x%.16llX", value);
  }

void OnStart()
  {
   Print(DoubleToHexadecimal(0.0) );    // 0x0000000000000000   (i.e, long integer 0)
   Print(DoubleToHexadecimal(-0.0) );   // 0x8000000000000000   (i,e, LONG_MIN -9223372036854775808)
  }
  

Works as expected! The first bit (bit index 0) is the sign bit which is set for -0.0.

These are also more special cases in ieee 754 format. https://www. wikiwand.com/en/IEEE%20754#Special_values

void OnStart()
  {
   const double Nan = (double)"nan";
   const double Inf = (double)"inf";

   Print( MathMin(Nan, Inf));   // inf
   Print( MathMin(Inf, Nan));   // nan

   Print( MathMin(Nan, 0));     // 0.0
   Print( MathMin(0, Nan));     // nan
  }

[Nikolai Semko]

Denis Kirichenko #:

It seems to be a bug.

It works fine in C++:

in the console:

there is no bug. In C++ it is the same.
you just made a mistake with formatting and type

try this:

#include <iostream>

using namespace std;
//
template <typename T>
T ToType(const double Num)
{
    union UNION
    {
        T Num1;
        double Num2;
    } Union;
    Union.Num2 = Num;
    return(Union.Num1);
}
//
int main()
{
    double positive_zero_val, negative_zero_val;
    positive_zero_val = 0.0;
    negative_zero_val = -0.0;
    //---
    long long  long_negative_zero_val = ToType<long long>(negative_zero_val);
    printf("\nLong negative zero = %ll d", long_negative_zero_val);
}

[amrali]

In the same context, this can examine the 64 bit representation of doubles https://www.mql5.com/en/code/20822

//+------------------------------------------------------------------+
//| Returns the bit representation corresponding to a double value . |
//+------------------------------------------------------------------+
long DoubleToLongBits(const double value)
  {
   union _d {double value; long bits;} dbl;
   dbl.value = value;

   return dbl.bits;
  }

or the shorter version

long DoubleToLongBits(const double value)
  {
   union _d {double value; long bits;} dbl = { value };

   return dbl.bits;
  }

[Denis Kirichenko]

Nikolai Semko #:

there is no bug. It is the same in C++.
just made a mistake with formatting and type

try this:

Yes, I was in a hurry. Thanks for the science ))

[fxsaber]

Sometimes you need to quickly put an EA into the Tester. You can do it like this.

Select a date in the future and press CTRL+F5 in ME.

[fxsaber]

If you need fast compilation without a big performance hit, you can use compiled libraries - #import .ex5