Why this error happen?

Code:

---

float a=127.2344,c;
int b;

b = abs(a); <====1
c = a - b; <====2

---

1 - Return 127
2 - Return 0.234375 and the correct is 0.2344 - ERROR.

Why this problem happen? I really need the correct value. óÒ


Tks very much ^^

Quote:

---

Originally Posted by phoenixbr

Code:

---

float a=127.2344,c;
int b;

b = abs(a); <====1
c = a - b; <====2

---

1 - Return 127
2 - Return 0.234375 and the correct is 0.2344 - ERROR.

Why this problem happen? I really need the correct value. óÒ


Tks very much ^^

---

here is your answer

Code:

---

double a=127.2344,c;
int b;
b = abs(a);
cout<<b;
c = a - b;
cout<<endl;
cout<<c;

---

due to decleration of float you was getting some thing
because compiler will treat it as double

Tks very much ^^

Quote:

---

Originally Posted by phoenixbr

Tks very much ^^

---

you welcome

Read this FAQ. Hope this helps. :thumb:

Quote:

---

Originally Posted by phoenixbr

Code:

---

float a=127.2344,c;
int b;

b = abs(a); <====1
c = a - b; <====2

---

1 - Return 127
2 - Return 0.234375 and the correct is 0.2344 - ERROR.

Why this problem happen? I really need the correct value. óÒ


Tks very much ^^

---

There are two implicit type conversion occured;
and the complier will transform the type of expression to assigned object.

Good Luck!

Quote:

---

Originally Posted by phoenixbr

Code:

---

float a=127.2344,c;
int b;

b = abs(a); <====1
c = a - b; <====2

---

---

1. It converts from float to integer, then it computes the absolute value of the integer.
   To avoid strange rounding rules, it may be preferable to use fabs which uses a double argument.
   
   Code:

   ---

   b = fabs(a);

   ---

2. It is a precision error, because 127.2344 cannot be correctly represented on a float number (float are binary numbers, not decimal numbers, so 1/5 for example cannot be correctly represented).
   Using double values greatly increase the precision.
   
   Note, that BC++ has a library for BCD (Binary coded decimals) floats, but use them only if you really need them, because it is not portable.

You can use floor and fabs functions - it wouldn't require conversions:

Code:

---

double a=127.2344;
double c=fabs(a)-floor(fabs(a));

---

You could also use boost::rational if you really want decimal fractions represented exactly.

An IEEE float gives 23 significant bits of data (22 bits but there is one assumed 1-bit) so that's about 8 million. Therefore you get just under 7 significant figures in decimal.