[boost] Re: yet another utility class(es)

From: Aleksey Gurtovoy (alexy_at_[hidden])
Date: 2000-02-16 00:38:46

Ed Brey <brey_at_[hidden]> wrote:
> IHMO, this looks very good, indeed good enough for detailed review and
> comments:
>
> 1. I really like two class breakdown and the names of the classes, except
> check the spelling of deferred.
>
Opps.. of course I'll change it :)

> 2. Check the detail namespace.  IIRC, we decided on "detail" rather than
> "details".
>
Will be fixed.

> 3. The "::boost::" in "::boost::details::" seems unnecessary.  Any reason
> this the lookup would ever go awry using the terse form?
>
I think there is no one.. just hadn't thought enough when wrote the code.
I'll fix it.

> 4. Any reason that auto_restore doesn't inherit from noncopyable instead
of
> the derived classes?  It would simplify the code a notch.  One could argue
> that you lose documentation, but that seems to be a weak argument since
the
> user shouldn't be looking at private base class specifications (or private
> anything) for documentation.  The html docs should state the copy
> capabilities of the public classes.
>
Good point. I overlooked this refactoring...

> 5. I think you may need "typename" on the "const T value_" and "T&
> variable_", and maybe the typedefs too.  I've never been able to figure
out
> what the requirements are.  Does anyone know of a good (easily
> understandable) reference for learning where typename is required?
>

I don't think so. From the section 14.6 [temp.res] of the standard ( para 5,
p. 258):

... 'The keyword typename shall only be applied to qualified names, but
those names need not be dependent.'..

'T&' and 'const T' are certainly not a 'qualified names' :)

-Alexy