Am 01.10.21 um 15:32 schrieb Sven Schreiber:
Am 01.10.2021 um 15:23 schrieb Artur T.:
> Hi
>
> Am 30.09.21 um 23:03 schrieb Sven Schreiber:
>> Maybe there's a case for using an integer arg similar to $pd here,
>> allowing values 4 and 12.
>
> Ah, good point. I've changed the signature to
>
> <>
> function scalar gdate_to_iso8601 (string date "Date string",
> string frequency[null] "Frequency of
> observation")
> </>
>
> Thus, the "frequency" string parameter is optional. If not passed, the
> frequency string is determined by the underlying periodicity as
> retrived by the $pd accessor.
It's good to give the user the possibility to override the default. Note
that $pd has a rather different meaning in a panel dataset. So even if
you declare the panel-time to be quarterly, $pd will give you the total
time dimension length instead. (Which IMHO is quite confusing, but it is
clearly documented.) So I think internally the function must not use $pd
if $datatype gives you 3 (panel data code) [or 0 (no data)].
True, the panel case is considered now, too.
But to avoid misunderstanding: The idea was that (perhaps) the
frequency
arg might be changed from string to integer. But it's clearly a matter
of taste as well.
Yes, I guess it's a matter of taste. Personally I think that a string is
more intuitive for users not dealing with that frequency aspect that often.
Ok, then we have it I think. Let me update the help text and upload a
new version later.
Artur