why error double + double ???

[Nguyen Chau Son]

Code:

double total = 0;
double a1 = 28.69;
double a2 = 14.21;
total = a1 + a2;

I builde it in Vs2008,and get the result as:
total = 42.900000000000 006

[William Di Luig]

It's because of the precision. In fact, when you do operations with doubles you can't expect to have perfect results.

It is often useful to compare doubles using a particular technique, instead of writing:

double a, b;
if (a == b)

you use an "eps" constant (very very little) and do the inverse:

const double eps = 1e-9;
double a, b;
if ( fabs(a-b) < eps)

[Nguyen Chau Son]

But when I use the two operators is the decimal it for the right result. The comparison with how you say it will only be used when applied to some cases but if so when you have to compare like to see an addition is correct?

[William Di Luig]

I dont' understand you.

If you have to check the correctness of an addition you can use EPS.

if (total == a1+a2)

becomes

if ( fabs(total-(a1+a2)) < EPS)

[Nguyen Chau Son]

my mind to ask why there is no exact results when added two double numbers, but exactly when add two decimal numbers.

You have a comprehensive solution for the case added two double numbers this?

[code green]

The problems you are encountering are part of basic computer science.
That is how a computer stores floating point numbers.
I recommend some background reading on this subject

When comparing floats it is general practice to set a tolerance, which is what Willian explained to you.

You define a constant EPS (epsilon) as the acceptable tolerance say
0.0000001.

Then subtract actual and expected results and compare to EPS as William showed.
if outside the tolerance the result is rejected.

[William Di Luig]

Look here to understand how double works.

[Nguyen Chau Son]

If the comparison as you say, when I perform operations that are intermediate calculations over to do it the way you say?

[code green]

Interesting link William.
I'm afraid I could not understand the OP last question

[bvrwoo]

Read on conversion. No percision is exact with double (and even floats). You have to remember that you are using 8 bytes of memory with 64 bits of 1 and 0's in machine code for your number. It can only get so precise with floating points of exponents (include 32bit (float) exponent of 8).