Remove domain from numericValue

[uscholdm]

@dylan-sa @rjyounes
The domain of gist:numericValue is gist:Magnitude.
If I want to have an inexact duration such as 5 months, I want to model it as if it was a Magnitude:

ex:_InexactDuration_5_month 
	a ex:InexactDuration ;
	gist:numericValue "5"^^xsd:decimal ;
	skos:prefLabel "5 month"^^xsd:string ; 
	ex:hasInexactDurationUnit ex:_InexactDurationUnit_Month ;
	.

Because of the domain axiom, this instance is inferred to be a Magnitude, but it is not a Magnitude because Month not a proper UnitOfMeasure for a duration -- a month is a pseudo-duration. Math would not work.

This can be solved by removing the domain axiom and using gist:domainIncludes instead.

[rjyounes]

There may be other ways to model the month case, such as with precision. A month is based on the lunar calendar, which doesn't mean it's not an exact duration, only a duration that is out of alignment with the solar calendar we use to define days and years.

Removing the domain from numericValue is a significant downgrading of the semantics of the predicate, so we need to consider it carefully.

[uscholdm]

A month is based on the lunar calendar, which doesn't mean it's not an exact duration, only a duration that is out of alignment with the solar calendar we use to define days and years.

I'm talking about a calendar month in common parlance. Exactly 3 months can in the number of days it is, depending on which months. But I can say exactly 3 months after Jan27, 2021 and it always means May 27, 2021. The enterprise is much less concerned with astronomical months based on the moon.

[Jamie-SA]

A calendar month that is not from the first day to the last day, is not a good concept. What is one month from Jan 31, 2023?
In most cases, that gets restated as 30 (or 60, or 90) days.

[uscholdm]

A calendar month that is not from the first day to the last day, is not a good concept. What is one month from Jan 31, 2023?

That's an interesting question, you had me stumped for a while. In that case Ione might say one month would last until the day before the last day in February - or some agreed sensible convention for such cases. I don't know how warranty periods work, maybe no-one ever has a N month warranty period, only in days, for the exactness. But there are 1 year warranty periods.

Similar issue arises for birthdays and anniversaries or purchases that happen on Feb 29. When is the end of the year that started on Feb 29, 2020? One reasonable convention would be Feb 28, 2021.

[stevenchalem]

@uscholdm Here are some examples of how Excel's EDATE() function handles this:
1/31/2020 + 1 month = 2/29/2020
2/29/2000 + 1 month = 3/29/2020
2/29/2000 + 12 mos. = 2/28/2020

[rjyounes]

A calendar month that is not from the first day to the last day, is not a good concept.

We use this all the time. "We are having our anniversary party one month from today." "I last saw you one month ago yesterday." Etc.

If I say the first sentence on Oct 3, the party takes place on Nov 3. If I say it on Oct 31, the party takes place on ...? This is the difficulty. But it's not that we can't use the notion of calendar month for anything but Oct 1 - Oct 31.

[uscholdm]

What is one month from Jan 31, 2023?

Excel EDATE gives one answer to @Jamie-SA 's question. Here are the results for end January.

[rjyounes]

Makes sense to me. But according to what Steven shows, Excel knows whether the year is a leap year or not. I assume for Oct 31 it does Nov 30, right?

[stevenchalem]

@rjyounes Yes, EDATE("10/31/2023", 1) = 11/30/2023

[rjyounes]

@philblackwood Is there impact from the new units and measures model?