From: David B. Held (dheld_at_[hidden])
Date: 2003-03-18 12:28:21
"David B. Held" <dheld_at_[hidden]> wrote in message
news:b57jvh$bts$1_at_main.gmane.org...
> "David Abrahams" <dave_at_[hidden]> wrote in message
> news:u7kaw7h88.fsf_at_boost-consulting.com...
> > [...]
> > That's a pretty major problem, though. Your idea also cuts off
> > implicit conversions.
>
> Do you mean user-defined conversions, because it consumes one
> from by_ref->T? It's more awkward, but wouldn't calling a named
> member to get T solve that?
> [...]
I wonder if you mean the implicit conversion from T to by_ref in the
first place instead. It turns out that I really don't understand how
argument deduction goes, because Comeau doesn't like this code,
which I expected to work:
template <typename T, bool Big = (sizeof(T) > 8)>
class by_ref
{
public:
by_ref(T const& val) : val_(&val) { }
operator T const&() const { return *val_; }
private:
T* val_;
};
template <typename T>
class by_ref<T, false>
{
public:
by_ref(T val) : val_(val) { }
operator T() const { return val_; }
private:
T val_;
};
void g(long l) { }
template <typename T1>
void f(by_ref<T1> v1)
{
g(v1);
}
int main()
{
f(5);
}
"ComeauTest.c", line 31: error: no instance of function template "f"
matches the argument list
The argument types that you used are: (int)
Also, I was suprised that I had to put parens around the sizeof(T) > 8
expression, as I thought that the parser had to try to match the largest
possible token sequence to the template parameter. But maybe that
only applies for instantiations, and not definitions?
Dave