RE: [Dev-C++] Overloading & reliability

[Ioannis V. <no...@ho...>]

> -----Original Message-----
> From: dev...@li... 
> [mailto:dev...@li...] On Behalf 
> Of rchavez
> Sent: Saturday, May 04, 2002 4:13 AM
> To: dev...@li...
> Subject: [Dev-C++] Overloading & reliability
> 
> 
> Hi,
> 
> The thread on 'Begginer's question, Function overloading' 
> made me wonder about
> the following:
> 
> Following Brian's example (I'm not criticizing his code), 
> reproduced partially
> below, it's clear how overloading works.  But it seems to me 
> it could be
> a very dangerous feature of the C++ language.  If you meant to call
> function with 2 parameters 1.1 and 2.2 but wrote a comma 
> instead you wouldn't
> get a error message, you would get a totally different function.
> 
> I would imagine the 2 functions would be similar or else they 
> would have
> different names right?  So it might be very difficult to 
> detect that the
> wrong function has been called in a large program.
> 
> When a support programmer changes the 2,2 to 2,3 really bad 
> things might happen.
> 
> Overloading seems very risky in critical applications with 
> rare corner cases.
> It's the opposite of the graceful failure I would want in an 
> airplane control
> system for example.  After all, you never test every possibility.
> 
> Comments?


In cases where the same processing of data takes place, templates should
be used. Compiler cannot protect us from implicit conversions in
function overloading. Consider the code:

void something(int x)
{
 // ...
}

void something(char c)
{
 // ..
}

int main()
{
 double x=7.3;

// void something(int) is called, data truncated to 7.
// No compiler warnings or errors.
 something(x);

}


Also maintenance can be a nightmare in the above.

The superior alternative which should be used:


template <class T>
void something(const T &x)
{
 // ...
}



Ioannis

* Ioannis Vranos
* Programming pages: http://www.noicys.cjb.net
* Alternative URL: http://run.to/noicys