Subject: Re: [boost] [review] status of Boost.Conversion?
From: Vicente J. Botet Escriba (vicente.botet_at_[hidden])
Date: 2012-06-29 02:49:06
Le 29/06/12 02:16, Gordon Woodhull a écrit :
> On Jun 28, 2012, at 7:20 AM, "Vicente J. Botet Escriba"<vicente.botet_at_[hidden]> wrote:
>> I would prefer to avoid a dependency from Boost.Chrono to Boost.DateTime.
> Understood. Boost.Conversion has/had a nice directory structure to put these things in.
> But for the sake of argument, what is the difference between an optional header dependency between Chrono and DateTime, and an optional header dependency from Conversion to both libraries? Is your objection because Chrono is mostly standard?
Yes and not. Boost.Chrono contains already features that are not
standard (e.g. Boost.Chrono/IO/Rounding). and I pretend to add more
(e.g. Boost.Chrono/Stopwatches). Of course if a conversion was added it
should not be included as a member of the duration or time_point classes
as I try to maintain them close to the standard, so a non member
functions are needed. I don't know a good name for these non members
functions conversions. Has someone considered to add these conversions
> A few reviewers said they'd rather see individual libraries implement their own conversions rather than putting them all in Conversion.
I'm not against adding these conversions as a sub-library of
Boost.Chrono (eg. Boost.Chrono/Conversions) if the Boost community reach
a consensus on the interface and someone takes care of the authoring and
maintenance of this sub-library. Has someone considered to add these
conversions to Boost.DateTime?