Re: patch: function xmltable

On 15 September 2016 at 19:31, Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com> wrote:

> b_expr enforces shift/reduce conflict :(

No problem then. I just thought it'd be worth allowing more if it
worked to do so.

> I found other opened question - how we can translate empty tag to SQL value?
> The Oracle should not to solve this question, but PostgreSQL does. Some
> databases returns empty string.

Oracle doesn't solve the problem? it ERRORs?

> I prefer return a empty string - not null in this case.

I agree, and that's consistent with how most XML is interpreted. XSLT
for example considers <x></x> and <x/> to be pretty much the same
thing.

> The reason is simple
> - Empty string is some information - and NULL is less information. When it
> is necessary I can transform empty string to NULL - different direction is
> not unique.

Yep, I definitely agree. The only issue is if people want a DEFAULT to
be applied for empty tags. But that's something they can do in a
post-process pass easily enough, since XMLTABLE is callable as a
subquery / WITH expression / etc.

--
Craig Ringer http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services