On Thu, May 6, 2010 at 10:10 AM, Mike Fowler <mike@mlfowler.com> wrote:
> Hi hackers,
>
> Although this is a very small change I figured I'd practice the policy of
> outlining your change before you write the code and attempt a patch
> submission. Essentially I see the function as a convenience function that
> exposes the results of the xpath built in exists() function as a boolean for
> easier SQL. The syntax will match the xpath function already present:
>
> xpath_exists(xpath, xml[, nsarray])
>
> The implementation will check that the xpath value contains 'exists(.....)'
> and add it if not present for added usability. I can't blindly wrap the
> value with exists(...) as exists(exists(.....)) will always return true in
> xpath. I've not allocated the oid yet, but will try to earn my bonus points
> for getting it close to xpath()'s oid. :) Appropriate regression tests will
> be added after I've ensured that 'make check' is happy I've not regressed
> anything.
I'm not sure I understand how this more convenient than just using
xpath() with exists()?
> Can I confirm that contrib/xml2 is deprecated and I should be carrying out
> my work in backend/utils/adt/xml.c?
Yes.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise Postgres Company