What to do about this piece of dangerous code?

[mike3]

Hi.

I was reworking this program I had done in C++ and I came across this
little number:

---
template<class T>

class RandomGeneratorRanged : public RandomGeneratorRangedBase<T> {

protected:

T minVal, maxVal;

public:

RandomGeneratorRanged() : minVal(0), maxVal(1) {}



RandomGeneratorRanged(const T minVal_, const T maxVal_)

: minVal(minVal_), maxVal(maxVal_)

{

if(minVal > maxVal)

std::swap(minVal, maxVal);



if((static_cast<long>(minVal_) < c_rangeMin) ||

(static_cast<long>(maxVal_) > c_rangeMax))
// This is where the problem is
throw XXXXXException(ERROR_RANDOM_RANGE_OUT_OF_RANGE);

}


...


};
---

The class is supposed to wrap a random number generator and allow you
to set up a variable that can output random numbers in a preset range
(set up on construction). The T can be one of various integer types,
and that's where we run into problems. Namely, I've only ever used
this with signed types, but I wasn't sure how to make it go with
unsigned, because the "max check" there has c_RangeMin which is a
_negative_ (signed!!!!) number. So if you tried a
"RandomGeneratorRanged<unsigned int> foo(<smth>, <smth>)" _you'd get
into trouble_ and I realized I better do something about this before
it bites me in the ass. Is there a superior way to achieve this, that
will allow for both signed and unsigned use? As making classes for
each signed and unsigned integer type sounds like a way to make a mess
with duplicated code. Hence why there's a template there anyway.

[Ian Collins]

On 05/21/12 11:05 AM, mike3 wrote:

<snip>

> The class is supposed to wrap a random number generator and allow you
> to set up a variable that can output random numbers in a preset range
> (set up on construction). The T can be one of various integer types,
> and that's where we run into problems. Namely, I've only ever used
> this with signed types, but I wasn't sure how to make it go with
> unsigned, because the "max check" there has c_RangeMin which is a
> _negative_ (signed!!!!) number. So if you tried a
> "RandomGeneratorRanged<unsigned int> foo(<smth>,<smth>)" _you'd get
> into trouble_ and I realized I better do something about this before
> it bites me in the ass. Is there a superior way to achieve this, that
> will allow for both signed and unsigned use? As making classes for
> each signed and unsigned integer type sounds like a way to make a mess
> with duplicated code. Hence why there's a template there anyway.

Does this help?

#include <iostream>
#include <limits>

template <typename T>
void wibble( T t )
{
if( std::numeric_limits<T>::is_signed )
{
std::cout << "signed" << std::endl;
}
else
{
std::cout << "unsigned" << std::endl;
}
}

int main()
{
wibble(42);
wibble(42U);
}

--
Ian Collins

[mike3]

On May 20, 5:25 pm, Ian Collins <ian-n...@hotmail.com> wrote:
<snip>

>    if( std::numeric_limits<T>::is_signed )

Thanks for the answer!