Re: [Boost-users] [Chrono] IO formatting doc minimal/ missing/ appears not to match the actual code base

Subject: Re: [Boost-users] [Chrono] IO formatting doc minimal/ missing/ appears not to match the actual code base
From: Vicente J. Botet Escriba (vicente.botet_at_[hidden])
Date: 2015-09-03 18:09:51

Le 03/09/15 23:55, Vicente J. Botet Escriba a écrit :
> Le 03/09/15 17:03, Hickman, Steve (AdvTech) a écrit :
>
> Hi and thanks for catching all these issues.
>> As of at least Boost 1.58 / 1.59 (possibly earlier), the 
>> documentation for Boost.Chrono V2 IO appears to be out of synch with 
>> the actual libraries in the following ways:
>>
>> 1)      Header file names to include as described in the doc don't 
>> match the file names actually available to include. I was unable to 
>> find <boost/chrono/io/ios_state.hpp>. I did find 
>> <boost/chrono/io/ios_base_state.hpp> but it appears to include things 
>> not described in the doc for <boost/chrono/io/ios_state.hpp>.
> I will change the filename. Does the file includes everything included 
> in the doc? If not, what is missing?
> Is there something in the file that you want to be documented as public?
>>
>> 2)      Some format tags are listed as 'unsupported', but there is no 
>> description of what that means.  Does it mean they are ignored? Does 
>> a string still match (which means the input reader will move past 
>> those bytes)?
> I don't reach to locate "unsupported". Please could you be more precise?
>> 3)      I was able to find some examples in chrono\test\io . It would 
>> be nice if the tutorial made reference to these as a source of 
>> additional info.
> I will try it.
What is on the test/io directory is already on the tutorial. I recognize 
that not all the interface is in the tutorial, but writing tutorials is 
not my best card.
>>
>> 4)      There appear to be a number of interfaces whose doc appears 
>> to consist only of what is embedded in the header files (or I just 
>> didn't find the doc for them): time_point_get, etc. Not being 
>> familiar with how facets work, I'm not certain if I should be using 
>> these directly or not.
>  a facet that can be used the same way as std::num_get. I will add 
> some examples on how to use it.

Next follows an example of how I use it from the implementation of the 
extraction operator<<.  I'm not sure this is the best way to implement 
it but anyway this will give you an idea.

     template <class CharT, class Traits, class Clock, class Duration>
     std::basic_istream<CharT, Traits>&
     operator>>(std::basic_istream<CharT, Traits>& is, time_point<Clock, 
Duration>& tp)
     {
       std::ios_base::iostate err = std::ios_base::goodbit;

       BOOST_TRY
       {
         typename std::basic_istream<CharT, Traits>::sentry ipfx(is);
         if (bool(ipfx))
         {
           if (!std::has_facet<time_point_get<CharT> >(is.getloc()))
           {
             time_point_get<CharT> ().get(is, 
std::istreambuf_iterator<CharT, Traits>(), is, err, tp);
           }
           else
           {
             std::use_facet<time_point_get<CharT> >(is.getloc()).get(is, 
std::istreambuf_iterator<CharT, Traits>(), is,
                 err, tp);
           }
         }
       }
       BOOST_CATCH (...)
       {
         bool flag = false;
         BOOST_TRY
         {
           is.setstate(std::ios_base::failbit);
         }
         BOOST_CATCH (std::ios_base::failure )
         {
           flag = true;
         }
         BOOST_CATCH_END
         if (flag) throw;
       }
       BOOST_CATCH_END
       if (err) is.setstate(err);
       return is;
     }