比较浮点数和常数时的奇怪行为

Question

I was struck by this strange behavior:

float pi = 3.14;

if(pi == 3.14)
    cout << "OK";
else
    cout << "How is it possible?";

Could anyone explain this?

Answer 1

The constant in the if statement is (double)3.14 . It is very close to its float version, but not exactly equals it, because 3.14 is not represented exactly.

Try the same trick with 1.25 or any other number that can be represented exactly, and you will get an OK .

float num = 1.25;
if(num == 1.25)
    cout << "OK";
else
    cout << "How is it possible?";

You could also cast 3.14 to float to get an OK :

float pi = 3.14;

if(pi == (float)3.14)
    cout << "OK";
else
    cout << "How is it possible?";

Answer 2

Because by default values like 3.14 are treated as double, but storing one of this value in a float variable causes a loss of precision; and so there are microscopic differences related to the different precisions. Here's why floating point numbers shouldn't be compared directly most of the time.