Re: [office] Drafts for date-time functions

[Michael Stahl <michael.stahl@allotropia.de>]

hi Regina,

On 17/10/2023 19:17, Regina Henschel wrote:
> Hi all,
>
> as suggested in the meeting I sent you the current state of my work. It 
> shows my proposed solution for the case, that we decide to use new 
> functions for 'rounding before extracting'.
>
> The idea of using a dot in the function name comes from Michael. I like 
> that idea. It makes the function names good readable.

yes, and there are already several functions with a . in the name.

> I have added a new pseudotype DateTimeParam. That makes it possible to 
> give YEAR.ROUND a different behavior than YEAR, for example.
>
> The proposal does not introduce a DATETIMEVALUE.ROUND function as 
> suggested by Michael, but uses the VALUE function directly.
>
> Comments are welcome.

thanks, mostly looks good to me!


> Evaluators shall support extracting the year from a date beginning in 1900. Three-digit year numbers precede adoption of the Gregorian calendar, and may return either an Error or the year number. Four-digit year numbers preceding 1582 (inception of the Gregorian Calendar) may return either an Error or the year number. Four-digit year numbers following 1582 should return the year number.

the new text does not allow implementations to return an Error for a 
pre-Gregorian date, do we know if that is a problem for any existing 
implementation?

> If the serial number given in D represents a date before 1852-10-15 (adoption of the Gregorian calendar) then Evaluators may return an implementation-defined year number.

the "1852-10-15" has a typo, the year should be 1582.


> A formal definition of term âserial numberâ is missing.

not sure, i think it becomes obvious enough if you search the 
specification for "serial", but maybe that means it's not obvious enough.

> A text can describe a date in content, but use a non-Gregorian calendar. How is that handled?

i don't want to open that can of worms :)

> Should the specification force, that the serial number of a date is interpreted using the Gregorian calendar, if the date represented by the serial number is after 1582-10-15 (inclusive)?

no, because there is already an optional "Capability" for this:

> 7.4  Year 1583
> Evaluators claiming to implement âYear 1583â can correctly calculate dates correctly starting from the January 1 of the (ISO) year 1583. 

(editorial: one "correctly" is enough?)


regards,
 michael

--
Michael Stahl
Senior Software-Entwickler LibreOffice
âââ
allotropia software GmbH
Versmannstr. 4
20457 Hamburg
Germany
âââ
michael.stahl@allotropia.de
https://www.allotropia.de
âââ
Registered office: Hamburg, Germany
Registration court Hamburg, HRB 165405
Managing director: Thorsten Behrens
VAT-ID: DE 335606919
âââ