Problem with casting a double to int

[adonissiu]

Hi every c++ expert, today i've seen a strange case when i was debugging.
When I cast a double to an int, the value is altered but it doesn't look like neither round up nor round down.

Here is the code:

double temp = (double )(pQinQplOdds[iOddsCount-1]*100); //temp = 820.00000
int odds = static_cast<int>(temp);

odds should get 820 but it doesn't. It turns out getting a value of 819 !!

Anyone could tell me why is it?

P.S. It fixed after i use type float instead of double.

Thanks x 1000

[Meetee]

Hi every c++ expert, today i've seen a strange case when i was debugging.
When I cast a double to an int, the value is altered but it doesn't look like neither round up nor round down.

Here is the code:

double temp = (double )(pQinQplOdds[iOddsCount-1]*100); //temp = 820.00000
int odds = static_cast<int>(temp);

odds should get 820 but it doesn't. It turns out getting a value of 819 !!

Anyone could tell me why is it?

P.S. It fixed after i use type float instead of double.

Thanks x 1000

Hi,

I did this way

double d = 820.00000;
int i = static_cast<int>(d);
 

and it gives i = 820. So I think (double )(pQinQplOdds[iOddsCount-1]*100) here is some problem? What are the values of pQinQplOdds and iOddsCount?

Regards

[adonissiu]

Thanks for replying, i believe that the problem isn't in
(double )(pQinQplOdds[iOddsCount-1]*100).

And pQinQplOdds is an array of double and

(pQinQplOdds[iOddsCount-1]*100) == 820.00000000 is true.

I believe that the problem is in the casting part.

[weaknessforcats]

What type is pQinQplOdds?

If it's a float, then it has only 6 significant digits. When the division is cast to a double, the float is converted to double, the 100 is converted to double, the calcaulation is done as a double and the result is a double.

A double has 15 significant digits. Only 6 have any meaning. The other digits are bogus. Now you cast and chop off the decimals.

You can't say (pQinQplOdds[iOddsCount-1]*100) == 820.00000000 is true.
If may test as true but you can't use the == operator with floating point. Actually, you can't use any of the operators like <,> <=, >=, !=, etc with floating point.

Floating point numbers only close in value may test equal. You need to do the compare based on absolute value by subtracting one float from the other an testing the ressult be within a specified sigma error tolerance.

Google floating point arithmetic.

As a general practice:
1) never mix floating point an integers.
2) avoid floating point at all costs. It's very slow. It's inaccurate. It has automatic rounding that can't be turned off.
3) if you have to use floating point, use only double. Do not use float unless you can write down on paper the precise technical reason you need a float.

Finally, do not cast in C++.