Re: xpath improvement suggestion

From: Scott Bailey <artacus(at)comcast(dot)net>
To: Arie Bikker <arie(at)abikker(dot)nl>
Cc: pgsql-hackers(at)postgresql(dot)org
Subject: Re: xpath improvement suggestion
Date: 2010-01-06 00:34:46
Message-ID: 4B43DAA6.3070109@comcast.net
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Arie Bikker wrote:
> Hi all,
>
> Well I had to burn some midnight oil trying to figure out why a
> construct like
> SELECT xpath('name()','<a/>');
> doesn't give the expected result. Kept getting an empty array:
> xpath
> -------------
> {}
> instead of the expected "{a}"
> BugID 4294 and the TODO item "better handling of XPath data types"
> pointed in the right direction.
> whithin src/backend/utils/adt/xml.c in the function xpath the result of
> the call to xmlXPathCompiledEval is not handled optimally. In fact, the
> result is assumed to be a nodeset without consulting the ->type member
> of the result. I've made some minor changes to xml.c to handle some
> non-nodeset results of xmlXPathCompiledEval.
> Essentially, the revised code makes an array of all the nodes in the
> xpathobj result in case this is a nodeset, or an array with a single
> element in case the reult is a number/string/boolean. The problem cases
> mentioned in
> http://archives.postgresql.org/pgsql-hackers/2008-06/msg00616.php now
> work as expected.
> Revision of the code involves:
> - A switch statement to handle the result type of xmlXPathCompiledEval.
> - an additional function xmlpathobjtoxmltype.
>
> diff of the revisioned code with respect to original is in attached file.
>
> kind regards, Arie Bikker

Well that's interesting. I was getting ready to dig into libxml2 to see
what it would take to return scalar values from xpath expressions. Thanks.

Scott

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message u235sentinel 2010-01-06 01:00:53 Auto-extending table partitions?
Previous Message Tom Lane 2010-01-06 00:06:35 Re: Status of plperl inter-sp calling