[Boost-users] [date_time] new local_date_time constuctor

Subject: [Boost-users] [date_time] new local_date_time constuctor
From: a k (akrzemi1_at_[hidden])
Date: 2009-01-28 10:21:50

Hi,

This is a request to add a third constructor of the class
boost::local_time::local_date_time.

THE PROBLEM:

The following piece of code throws an exception.

    date d(2008, Oct, 26);
    time_duration td( 2, 30, 0 ); // 2:30 a.m.
    time_zone_ptr tz(
        new posix_time_zone("CET-1CEST,M3.5.0,M10.5.0/3") );
    local_date_time myLtime( d, td, tz, EXCEPTION_ON_ERROR );
    // last line throws

This is because the date and time I chose happens to be located in the
"vulnerble" moment when the wall clock gets moved back by an hour. In
human terms, time 2:30am appears twice on 26-Oct-2008 in CET time zone
because at 3:00 am we go back to 2:00 am and arrive at point 2:30 once
more. In terms of boost::date_time library, the following condition
appears:

    (local_date_time::check_dst(d, td, tz) == ambiguous);

I would like to be able to write something like "On creating date/time
infer the DST flag; on ambiguity assume no DST". In fact I can do that
currently by replacing the last line from my above example with:

bool inDST =
    (local_date_time::check_dst(d, td, tz) == is_in_dst);
local_date_time myLtime(d, td, tz, inDST);

This has two drawbacks however. First, check_dst gets called twice:
first by me, second by the constructor, to check it once more. Second,
it is too (IMO) verbose.

THE REQUEST:

Would it be possible to add a third constructor of local_date_time
class that would accept a hint how to solve an ambiguity (if we have
one). It would read like:

    local_date_time myLtime( d, td, tz, ASSUME_NOT_DST_ON_ERROR );

In the boost code I found it was already thought of, but got into a
TODO state. Is there any chance that it gets implemented?

Regards,
&rzej