[Info-ingres] Doing arithmetic with ANSI dates

[Roy Hann]

Karl Schendel wrote:

>> On Oct 28, 2021, at 4:08 AM, Roy Hann <specially at processed.almost.meat> wrote:
>> [snip] 
>> For completeness, it seems like the permitted keywords are DAY, MONTH,
>> and YEAR only...right?
>
> It's a bit more complicated than that.  I don't have the full syntax at hand, but
> the options are YEAR, MONTH, DAY, HOUR, MINUTE, SECOND,
> YEAR TO MONTH, and DAY TO HOUR, MINUTE, or SECOND.
> Intervals can have either year and month fields, or day / hour / minute / second
> fields, but not both for some bizarre reason that I don't know.

I recall an argument that it is because months don't all have the same
number of days so if an expression included months as well as days you
can't legitimately evaluate it. Sometimes, depending in my mood I can
hypnotize myself into agreeing.

> The quoted part in the middle is 'year-month' for a year to month interval,
> and 'day hour[:min[:second]]' for a day to second interval.  You can have
> a leading minus sign (inside the quotes) to indicate a negative interval.

I had wondered why it has to be quoted. 'YEAR-MONTH' explains it. It's
still kinda grotesque-looking and it grates, but OK, fine. 

I don't see why it can't be an expression that evaluates to a string
though.

>> And while I am here, Ingres accepts ANSIDATE + INTEGER. The integer is
>> interpretted as a number of days. Does the standard endorse such an
>> expression?
>
> I don't think it does.

On the one hand that's a pity, because it's unambiguous and useful. On
the other hand Ingres/X/Vector support it so I'm happy.

Roy