Re: [Boost-users] [date_time] dst_local_start_time

Subject: Re: [Boost-users] [date_time] dst_local_start_time
From: Roland Bock (rbock_at_[hidden])
Date: 2008-11-06 09:05:34

Hi,

regarding 2:00 or 3:00: I can live with both. Maybe there is an official 
definition? The more important thing is to document the exact meaning of 
the return value. Thus, a bug report for the documentation would be useful.

Regarding "unnatural" behaviour: The short-term change would be to add 
more documentation. In the long run, I would vote for modifying the API. 
Again, a bug report would be useful.

Regards,

Roland

Jean-Pierre Bergamin wrote:
> Hello
> 
> We ended up with an explanation like this as well. But in my opinion, 
> 2:00 in summertime does not exist, since 2:00 is 3:00. I'd expect that 
> the function dst_local_start_time returns a local_date_time of 
> 2007-Mar-25 03:00:00.
> 
> As you say, some functions with local_date_times behave very "unnatural" 
> and counterintuitive. Does it make sense to file a bug-report?
> 
> 
> Regards
> 
> James
> 
> 
>> forgot to write some kind of conclusion after my code sample:
>>
>> My guess: The dst_local_start_time() function returns a ptime object 
>> which actually represents wall time at the location (Zurich in your 
>> case).
>>
>> I would say that 2007-Mar-25 02:00:00 is the start of the DST just 
>> like 2007-Mar-25 03:00:00, because those two times are exactly the 
>> same. It's not like "when the clock reaches 02:00, we've reached end 
>> of wintertime, then we switch to 03:00 and have beginning of 
>> summertime". On that day, in the infinitesimal time interval when it 
>> is 02:00, it is also 03:00.
>>
>> So, personally, I would say, the confusing part is that local times 
>> are represented by ptimes in some cases (see also the example I sent). 
>> The result is OK, though.
>>
>> If you look at the dst_local_end_time(), it is consistent with the 
>> start function: In both cases, it returns the time which would be 
>> shown on the clock without the switch.
>>
>>
>> Regards,
>>
>> Roland
>>
>>
>> Roland Bock wrote:
>>> Hi,
>>>
>>> I had a similar problem in getting the current wall time in New York. 
>>> I wrote the following working function and added a bunch of comments, 
>>> since the library is counter-intuitive here IMHO:
>>>
>>>
>>> // Get local time of day, based on time zone
>>> const boost::posix_time::time_duration 
>>> DateTime::localizedTimeOfDayAsDuration(const 
>>> boost::local_time::time_zone_ptr timeZone)
>>> {
>>>    /* This looks a bit perverted (and it is in my eyes :-))
>>>
>>>       Calling time_of_day on a local_date_time object implicitly 
>>> turns it into a UTC ptime object first, thus yielding UTC time_of_day 
>>> instead of the wall time (I do not really understand why it works 
>>> this way, but it does).
>>>
>>>       Calling local_time on a local_date_time object returns a ptime 
>>> object which in this case represents wall time within the given time 
>>> zone (again, I don't know why the behavior is this way, but it is).
>>>    */
>>>    return 
>>> local_microsec_clock::local_time(timeZone).local_time().time_of_day();
>>> }
>>>
>>> Please let me know if that helped or if you found a better 
>>> solution/explanation.
>>>
>>> Regards,
>>>
>>> Roland
>>>
>>> Jean-Pierre Bergamin wrote:
>>>> Hello
>>>>
>>>> I'm confused about the returned time of the dst_local_start_time() 
>>>> function. First of all I don't understand why a ptime is returned 
>>>> and not a local_date_time, buy anyway.
>>>>
>>>> The ptime I get returned for the timezone Europe/Zurich for 2007 is: 
>>>> 2007-Mar-25 02:00:00
>>>> Assuming that this is a UTC time, the local time in Zurich would 
>>>> then be 2007-Mar-25 04:00:00 (because during summertime, Zurich is 
>>>> 2h ahead of UTC), but the correct local dst start time would be 
>>>> 2007-Mar-25 03:00:00 (because the clocks are shifted from 2 to 3).
>>>> To correctly calculate the dst start time, I have to add the 
>>>> dst_offset to the returned time_of_day.
>>>>
>>>> The returned ptime 2007-Mar-25 02:00:00 seems to be the endtime of 
>>>> wintertime as a local time.
>>>>
>>>> time_zone_ptr zh_tz = tz_db.time_zone_from_region("Europe/Zurich");
>>>> ptime zh_dst_start_time = zh_tz->dst_local_start_time(2007);
>>>> std::cout << zh_dst_start_time << std::endl; // 2007-Mar-25 02:00:00
>>>> local_date_time zh_dst_local_start_time = 
>>>> local_date_time(zh_dst_start_time, zh_tz);
>>>> std::cout << zh_dst_local_start_time << std::endl; // 2007-Mar-25 
>>>> 04:00:00 CEST which is wrong!
>>>>
>>>> ptime zh_dst_start_time_ok = zh_dst_start_time + zh_tz->dst_offset();
>>>> local_date_time zh_dst_local_start_time_ok = 
>>>> local_date_time(zh_dst_start_time_ok.date(), 
>>>> zh_dst_start_time_ok.time_of_day(), zh_tz, true);
>>>> std::cout << zh_dst_local_start_time_ok << std::endl; // 2007-Mar-25 
>>>> 03:00:00 CEST which is correct
>>>>
>>>> So what time exactly is local_start_time() returning?
>>>> Is this a bug or am I missing something?
>>>>
>>>>
>>>> Regards
>>>>
>>>> James
>>>> _______________________________________________
>>>> Boost-users mailing list
>>>> Boost-users_at_[hidden]
>>>> http://lists.boost.org/mailman/listinfo.cgi/boost-users
>>> _______________________________________________
>>> Boost-users mailing list
>>> Boost-users_at_[hidden]
>>> http://lists.boost.org/mailman/listinfo.cgi/boost-users
>> _______________________________________________
>> Boost-users mailing list
>> Boost-users_at_[hidden]
>> http://lists.boost.org/mailman/listinfo.cgi/boost-users
>>
> 
> _______________________________________________
> Boost-users mailing list
> Boost-users_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/boost-users