Converting a float to int (Visual Studio 2003)

[Code4u]

In the course of writing numerical code I needed to convert a float to
an int with a defined behavior: if the float is great than INT_MAX,
set the int to INT_MAX, otherwise assign directly. The problem I ran
into is a float with value INT_MAX assigned to an int results in the
value -2147483648 being assigned, but if the conversion takes place in
an expression INT_MAX is assigned as I would expect:

int temp1=float(std::numeric_limits<int>::max()); // 2147483647

float temp2=std::numeric_limits<int>::max(); //2.1474836e+009
int temp3=temp2; // -2147483648

My question: how do I write a function to trim a float to an int's
limits?

 

[Alf P. Steinbach]

* Code4u:

In the course of writing numerical code I needed to convert a float to
an int with a defined behavior: if the float is great than INT_MAX,
set the int to INT_MAX, otherwise assign directly. The problem I ran
into is a float with value INT_MAX assigned to an int results in the
value -2147483648 being assigned, but if the conversion takes place in
an expression INT_MAX is assigned as I would expect:

int temp1=float(std::numeric_limits<int>::max()); // 2147483647

I think this may be computed at compile-time with 'double' precision.

float temp2=std::numeric_limits<int>::max(); //2.1474836e+009
int temp3=temp2; // -2147483648

Consider that in many current implementations an 'int' is 32 bits, and a
'float' is 32 bits. In the 'float' some bits are used for the exponent, so
there are far fewer than 32 bits in the mantissa, i.e. not enough to
represent an arbitrary 'int' value exactly.

My question: how do I write a function to trim a float to an int's
limits?

Off the cuff,

int intFromFloat( float f )
{
static int const max = std::numeric_limits<int>::max();
return 0?0
: f > max? max
: f < -max? -max
: static_cast<int>( f );
}

 

[Code4u]

* Code4u:

I think this may be computed at compile-time with 'double' precision.



Consider that in many current implementations an 'int' is 32 bits, and a
'float' is 32 bits. In the 'float' some bits are used for the exponent, so
there are far fewer than 32 bits in the mantissa, i.e. not enough to
represent an arbitrary 'int' value exactly.



Off the cuff,

int intFromFloat( float f )
{
static int const max = std::numeric_limits<int>::max();
return 0?0
: f > max? max
: f < -max? -max
: static_cast<int>( f );
}

Thanks, yep, I should have spotted the compile-time conversion to
double.

One question, I don't quite understand the syntax of the ternary
conditional operator you used in the function, I've never seen 0?0 in
code before, could you elucidate?

 

[__PPS__]

return 0?0
: f > max? max
: f < -max? -max
: static_cast<int>( f );

means

if(0) {
return 0;
} else if(f>max) {
return max;
}else if(f<-max){
return -max;
}else{
return static_cast<int>( f );
}

 

[Code4u]

return 0?0
: f > max? max
: f < -max? -max
: static_cast<int>( f );

means

if(0) {
return 0;
} else if(f>max) {
return max;
}else if(f<-max){
return -max;
}else{
return static_cast<int>( f );
}

I understand how the conditional operator works. But what purpose does
if (0) serve? It is always false and is therefore redundant. Obviously
it must have a virtue, but I can't figure it out.

 

[Alf P. Steinbach]

* Code4u:

I understand how the conditional operator works. But what purpose does
if (0) serve? It is always false and is therefore redundant. Obviously
it must have a virtue, but I can't figure it out.

It's just a formatting issue. I prefer the colons lined up neatly on the
left.