Re: [boost] [complex] Feedback and Potential Review Manager

Subject: Re: [boost] [complex] Feedback and Potential Review Manager
From: Christopher Kormanyos (e_float_at_[hidden])
Date: 2012-05-07 16:27:40

<snip>
>> I believe the form of the constructor is actually:

>> complex(const T& re = T(), const T& im = T());

> I think we need someone to get a rule-book here,
> because I think you're wrong.

Well, it won't be the first time or the last that I'm wrong.

> The "T()" syntax should provide value-initialization,
> which will be zero for the built-in numeric types.
> Asking any mathematical type that isdefault-constructible
> to have said default state be equivalent to zero
> (or empty set, or some other analog) should be a requirement.
> (And for such math types, they should have a Boolean conversion
> where said zero/empty state is the only regular one that maps to False.
> Math types that do not map to (a subset of) real numbers should have
> said conversion be explicit.)
<snip>

> Daryle W.

Daryle, this issue has puzzled me for some time now
I guess there are two choices here for a potential complex<T>
in boost:

        1. complex(const T& re = T(), const T& im = T());
        2. complex(const T& re = T(0), const T& im = T(0));
Let's keep in mind that we are talking about a template parameter
that is *not* required to be one of the built-in types float,
douuble or long double.

It's hard to interpret the right thing to do here.

ISO/IEC 14882:2011 specifies a generalized template for
complex<T> with the first form of the constructor
(the one with T() default parameters).

But ISO/IEC14882:2011goes on to explicitly specify
the forms of each template specialization for float, double
and long double. In addition, ISO/IEC 14882:2011 clearly
states that complex<T> for any type other than float,
double or long double is not specified.

According to ISO/IEC 14882:2011, the constructor
for the specialization of complex<float>, for example, is:

template<>

complex<float>(float re = 0.0F, float im = 0.0F);

The input parameters are non-constant and non-references.
Furthermore, the default parameters are specified to be 0.0F,
not float().
So should a potential Boost.Complex use the default
parameter T() or rather T(0)?

It may be more "in the spirit" of ISO(IEC 14882:2011
to use T(). On the other hand, writers and users of
a specialized type might be confused when all of
a sudden complex<user_type> might potentially be
creatable with non-zero junk.

In the end, there is no right or wrong. It's just up to
boost to set a policy here and document it.
In my opinion, this should be sorted out in a
potential review.

Best regards, Chris.