Re: BUG #18348: Inconsistency with EXTRACT([field] from INTERVAL);

David Rowley <dgrowleyml(at)gmail(dot)com> writes:
> On Sat, 17 Feb 2024 at 01:27, PG Bug reporting form
> <noreply(at)postgresql(dot)org> wrote:
>> Moreover, the documentation does not mention that the field cannot be
>> extracted from INTERVAL, like it does for isoyear:
>> https://www.postgresql.org/docs/current/functions-datetime.html#FUNCTIONS-DATETIME-EXTRACT

> Maybe that table should specify which type(s) each of the items listed
> is applicable to. Seems better than mentioning which types they're not
> applicable to.

The thing's not laid out as a table though, and converting it seems
like more trouble than this is worth. The rejected cases hardly seem
surprising. I propose just mentioning that not all fields apply for
all data types, as in 0001 attached.

(Parenthetically, one case that perhaps is surprising is
ERROR: unit "week" not supported for type interval
Why not just return the day field divided by 7?)

Unrelated but adjacent, the discussion of the century field seems
more than a bit flippant when I read it now. In other places we
are typically content to use examples to make similar points.
I propose doing so here too, as in 0002 attached.

Lastly, the entire page is quite schizophrenic about whether to leave
a blank line between adjacent examples. I could go either way on
whether to have that whitespace or not, but I do think it would be
better to make it uniform. Any votes on what to do there?

regards, tom lane