Re: DateTime segments in series: Is there prior art?

cc list

On 9/9/13 4:54 AM, Flavio S. Glock wrote:
> Hi James,
>
> Yes, you are using DateTime::SpanSet correctly, but maybe this is not the
> right data structure for your problem.
> The problem that DateTime::SpanSet solves is exactly about using Spans as a
> single data entity - and the Spans are merged whenever possible.
> This is more complex than what you need, and it gets in the way.
>
> There are some workarounds - you could define the spans with
> 'after'/'before' - this would effectively create a "hole" in the data when
> the spans are merged.
>
> Maybe your problem is better solved by iterating over the list of Spans
> that you already have.
>
>
>
> 2013/9/9 James E Keenan<[email protected]>
>
>> On 8/11/13 1:25 PM, Flavio S. Glock wrote:
>>
>>> http://search.cpan.org/dist/**DateTime-Set/lib/DateTime/**SpanSet.pm<http://search.cpan.org/dist/DateTime-Set/lib/DateTime/SpanSet.pm>
>>>
>>> next() and previous() can be used to determine whether a given point
>>> in time falls between, before, or after the segments.
>>>
>>> intersects() can tell whether a new segment would intersect existing
>>> segments.
>>>
>>>
>> Flavio, thank you for pointing me to that distribution.  I have been
>> experimenting with DateTime::Span and find it helpful, but there's a
>> problem with DateTime::SpanSet -- or perhaps with my understanding of that
>> module.
>>
>> Once again, my original problem:
>>
>>
>>   2013/8/11 James E Keenan<[email protected]>:
>>>
>>>> I need to use DateTime functionality and would like to know
>>>> if the particular functionality I need already exists on
>>>> CPAN.
>>>>
>>>> I need to construct a series of time segments which do not
>>>> overlap but which do not need to be contiguous.  Each
>>>> segment is defined by a start date which is part of the
>>>> segment and by an end date which is the first moment in time
>>>> after the segment's conclusion.
>>>>
>>>>
>> I can use a start date and an end date to compose a DateTime::Span object
>> for each segment.  (It turns out not to matter so much whether the the
>> segment is, in DT::S parlance, 'semi-open' or 'closed', i.e., whether I
>> define the object with the second date with 'before' or with 'end'.)
>>
>> However, I see that when (a) the end date of one segment is identical to
>> the start date of the next segment; and (b) I use DateTime::SpanSet to
>> create a spanset; then those two segments are, in effect, merged -- which
>> seems counterintuitive.  The only way I can get a spanset with the same
>> number of elements as the spans going in is to make sure that the end date
>> of one segment is one unit of time less than the start date of the second
>> segment.
>>
>> The attached program, bspanset.pl, illustrates this problem.  As a
>> convenience, I have divided the program into 7 sections named with roman
>> numerals.
>>
>> In Section I, I create 4 pairs of DateTime objects.  The end date of the
>> first pair is identical to the start date of the second pair.  Likewise for
>> the third and fourth pairs.  But there is a gap between the second and
>> third pairs.  The output:
>>
>> # I:
>>
>> start:  2012-12-15T00:00:00     end:    2013-01-01T00:00:00
>> start:  2013-01-01T00:00:00     end:    2013-02-01T00:00:00
>> start:  2013-03-01T00:00:00     end:    2013-04-01T00:00:00
>> start:  2013-04-01T00:00:00     end:    2013-04-15T00:00:00
>>
>> In Section II, I create DateTime::Span objects from each pair.  I use
>> 'before' to create semi-open objects.
>>
>> # II:
>>
>> start: 2012-12-15T00:00:00      end: 2013-01-01T00:00:00
>> start: 2013-01-01T00:00:00      end: 2013-02-01T00:00:00
>> start: 2013-03-01T00:00:00      end: 2013-04-01T00:00:00
>> start: 2013-04-01T00:00:00      end: 2013-04-15T00:00:00
>>
>> In Section III, I create a DateTime::SpanSet object from the 4
>> DateTime::Span objects.  But this SpanSet object only contains two
>> elements.  The first and second have been merged, as have the third and
>> fourth.
>>
>> # III:
>>
>> start: 2012-12-15       end:   2013-02-01
>> start: 2013-03-01       end:   2013-04-15
>>
>> In Sections IV an V, I repeat II and III, only with 'end' to create closed
>> objects.  I still end up with only two elements in the DateTime::SpanSet
>> object.
>>
>> # IV:
>>
>> start: 2012-12-15T00:00:00      end: 2013-01-01T00:00:00
>> start: 2013-01-01T00:00:00      end: 2013-02-01T00:00:00
>> start: 2013-03-01T00:00:00      end: 2013-04-01T00:00:00
>> start: 2013-04-01T00:00:00      end: 2013-04-15T00:00:00
>>
>> # V:
>>
>> start: 2012-12-15       end:   2013-02-01
>> start: 2013-03-01       end:   2013-04-15
>>
>> In Section VI, I subtract one second from the latter date in each pair,
>> then use that to create a closed object.
>>
>> # VI:
>>
>> start: 2012-12-15T00:00:00      end: 2012-12-31T23:59:59
>> start: 2013-01-01T00:00:00      end: 2013-01-31T23:59:59
>> start: 2013-03-01T00:00:00      end: 2013-03-31T23:59:59
>> start: 2013-04-01T00:00:00      end: 2013-04-14T23:59:59
>>
>> In Section VII, I created a DateTime::SpanSet object from the four
>> DateTime::Span objects.  Only now do I get what I intuitively thought I
>> should get, namely, a SpanSet object with four elements.
>>
>> # VII:
>>
>> start: 2012-12-15       end:   2012-12-31
>> start: 2013-01-01       end:   2013-01-31
>> start: 2013-03-01       end:   2013-03-31
>> start: 2013-04-01       end:   2013-04-14
>>
>> Am I using DateTime::SpanSet correctly?
>>
>>
>> Thank you very much.
>> Jim Keenan
>>
>