Re: Have I found an interval arithmetic bug?

On Sun, Jul 25, 2021 at 11:56:54AM -0700, Bryn Llewellyn wrote:
> As far as I’ve been able, the PG documentation doesn’t do a good job of
> defining the semantics of any of these operations. Some (like the “justify”

This is because fractional interval values are not used or asked about
often.

> functions” are sketched reasonably well. Others, like interval multiplication,
> are entirely undefined.

Yes, the “justify” functions were requested and implemented because they
met a frequently-requested need unrelated to fractional values, though
they do have spill-up uses.

> This makes discussion of simple test like the two I showed immediately above
> hard. It also makes any discussion of correctness, possible bugs, and proposed
> implementation changes very difficult.

Agreed.  With fractional values an edge use-case, we are trying to find
the most useful implementation.

> As I’ve said, my conclusion is that the only safe approach is to create and use
> only “pure” interval values (where just one of the internal fields is
> non-zero). For this reason (and having seen what I decided was the impossibly
> unmemorable rules that my modeled implementation uses) I didn’t look at the
> rules for the other fields that the interval literal allows (weeks, centuries,
> millennia, and so on).

I think the current page is clear about _specifying_ fractional units,
but you are right that multiplication/division of fractional values is
not covered.

-- 
  Bruce Momjian  <bruce@momjian.us>        https://momjian.us
  EDB                                      https://enterprisedb.com

  If only the physical world exists, free will is an illusion.