Optimize XML parsing query

Question

My developers have come up with a query that is hogging CPU and leading to unacceptable execution times. My initial analysis is that the problem is caused by the fact that the query is heavily parsing XML data with multiple calls to getClobVal() and extract(). I'm looking for suggestions on how to improve the performance of this query-- radical suggestions if necessary.
My first thought is to denormalize the data: pull out the XML fields that are being parsed frequently and put them into their own table columns. This would take quite a while (it's a large table), and there is no guarantee that tomorrow they won't come up with another field that needs to be parsed out. Can anyone offer other ideas, based on the details provided here?
It's an 11g database that consists of little more than a 20-million row table of XML CLOBs with an Oracle Text index (about 2 TB in size). The following is the problematic query. Its purpose is to parse certain fields out of the XML, transform some of them, and then dump those fields to a spreadsheet. It gets called on an ad-hoc basis by a web-based reporting app.
SELECT /*+ FIRST_ROWS( 500 ) */ a.* FROM (
  SELECT /*+ FIRST_ROWS( 500 ) */ SCORE(0) as SCORE ,
  extract( mytable.xmldata , '/document/biblio/doc_num/text()').getClobVal() AS Document,
  extract( mytable.xmldata , '/document/biblio/kind_code/text()').getClobVal() AS Kind_Code,
  extract( mytable.xmldata , '/document/biblio/filing_date/text()').getClobVal() AS Filed,
  extract( mytable.xmldata , '/document/biblio/pub_date/text()').getClobVal() AS Published,
  REPLACE(REPLACE(extract( mytable.xmldata , '/document/biblio/classification/classes[@type="abc"]/class/text()').getClobVal() , '<![CDATA[', ''), ']]>', '') AS ABC,
  REPLACE(REPLACE(extract( mytable.xmldata , '/document/biblio/classification/classes[@type="abcd"]/class/text()').getClobVal() , '<![CDATA[', ''), ']]>', '') AS ABC_D,
  REPLACE(REPLACE(extract( mytable.xmldata , '/document/biblio/title/text/text()').getClobVal() , '<![CDATA[', ''), ']]>', '') AS Title,
  extract( mytable.xmldata , '/document/biblio/author/name/text()').getClobVal() AS Author,
  extract( mytable.xmldata , '/document/biblio/assignees/assignee/name/text()').getClobVal() AS Assignee,
  REPLACE(REPLACE(extract( mytable.xmldata , '/document/biblio/abstract/text/text()').getClobVal() , '<![CDATA[', ''), ']]>', '') AS Abstract
  FROM mytable WHERE contains(xmldata, :yesclause , 0)>0 ORDER BY SCORE(0) DESC
) a WHERE ROWNUM<=500  Thanks for any thoughts on this.

Answer 1

when parse xml try to manipulate character set
like
select <data>
    from (XMLTABLE('*'
                   PASSING (XMLTYPE(bfilename('XMLSTORE','data.xml'),NLS_CHARSET_ID('AL32UTF8')))
                   COLUMNS <data> xmltype PATH '/*'
         )