Re: [lmi] Delegating ctor uncalled for?

[Greg Chicares]

On 2017-03-03 14:18, Vadim Zeitlin wrote:
> On Wed, 1 Mar 2017 12:52:58 +0000 Greg Chicares <address@hidden> wrote:
> 
> GC> What do you think of the patch below[0]? The question I'm struggling with
> GC> is whether it's worth an extra fifty lines of code just to say "don't use
> GC> class SequenceParser directly"
[...]
>  I'm afraid I can't say anything really original about this, but just
> repeat what I had already written before: yes, I do think the patch below
> improves things and I also think that [...snip...]

Focusing only on the 2017-03-01 patch...I just set out to apply it, and
started by trying to write a commit message describing its purpose, but
I'm stuck. What useful purpose does it accomplish?

Omitting the minor details (which may worth applying in their own right),
the patch changes this:

  class parser {
    public:
      parser(argument_type); // This is the only public ctor.
      result_type results() const;
  };

into this:

  class parser {
    friend parser make_parser(argument_type);
    public:
      result_type results() const;
    private:
      parser(argument_type);
  };

  parser make_parser(argument_type args) {return parser(args);}

In what way is that an improvement?