On Thu, Nov 22, 2018 at 03:16:11PM +0100, Patrick Francelle wrote:
> On 11/15/18 00:02, Tom Lane wrote:
> > I think this could be improved some more. Perhaps something like this
> > (I've not bothered with markup...)
> >
> > This is a little verbose maybe, but as the text stands, it sounds like
> > using a trigger is enough to solve all the consistency problems that
> > a cross-row CHECK has. Which it's not of course.
>
> Thank you for the rewriting, this is much more clear and explicit that way.
>
> > I'm also wondering whether it's better to put this in the CREATE TABLE
> > reference page instead of here. While there are certainly benefits in
> > having the caveat here, I'm a bit troubled by the number of forward
> > references to concepts that are described later. OTOH, a lot of people
> > who need the warning might never see it if it's buried in the reference
> > material.
>
> To address your remark, I added a small message in the CREATE TABLE
> reference page to be more explicit about the topic, so that it would be
> a warning for the users reading the section. And then a reference to the
> CHECK constraint page where the full explanation is to be located.
>
> That way, the caveat is mentioned in both pages, but the full
> explanation is located only on a single page.
>
> Please, let me know if this is good enough or maybe if I missed
> something.
>
> Patrick Francelle
I believe that features F671 (subqueries in CHECK constraints) and
possibly F673 (reads SQL-data routine invocations in CHECK
constraints) from the standard should be referred to here.
We haven't implemented either one of them, but we might some day.
Best,
David.
--
David Fetter <david(at)fetter(dot)org> http://fetter.org/
Phone: +1 415 235 3778
Remember to vote!
Consider donating to Postgres: http://www.postgresql.org/about/donate