Nested XML output in relational format

Question

I have serious performance issue in converting the relational xml in to relational table.
Below is my query which works good if number of nodes are around <100.
I have issue if number of node is more then 500.I have requirement where the number of nodes can be >5000.
I am using Oracle 11g R2.
Appreciate the help in tuning this or any alternative for this.
Thanks,
WITH t1
       AS (SELECT xmltype('<Id_359328 prodId="101265" prodName="Computer">
  <Id_359329  prodId="101788" prodName="C1">
    <Id_359330 prodId="175" prodName="C2"/>
    <Id_359331 prodId="101766" prodName="C3"/>
  </Id_359329>
  <Id_388909 prodId="100654" prodName="C4"/>
  <Id_388191 prodId="100672" prodName="C5"/>
  <Id_388927 prodId="100625" prodName="C6"/>
  <Id_359332 prodId="101766" prodName="C7">
    <Id_359380 prodId="100651" prodName="C8" />
    <Id_359334 prodId="100643" prodName="C9">
      <Id_359337 prodId="101788" prodName="C10">
        <Id_359338 prodId="175" prodName="C11"/>
        <Id_359339 prodId="101766" prodName="C12"/>
      </Id_359337>
    </Id_359334>
    <Id_359336 prodId="101860" prodName="C13"/>
  </Id_359332>
</Id_359328>')
                     t1_val
           FROM DUAL)
SELECT ROWNUM rn, id, prod_id, prnt_id
FROM t1,
     XMLTABLE ('for $i in $doc/descendant::*
                         return <ROW><Id>{name($i)}</Id>
                         <PROD_ID>{data($i/@prodId)}</PROD_ID>
                          <PRNT_ID>{name($i/..)}</PRNT_ID>
                                                </ROW>'
                  PASSING t1.t1_val AS "doc"
                  COLUMNS id VARCHAR2 (100) PATH '/ROW/Id',
                          prod_id NUMBER PATH '/ROW/PROD_ID',
                          prnt_id VARCHAR2 (100) PATH '/ROW/PRNT_ID'
ORDER BY ROWNUM
Output
RN     ID             PROD_ID     PRNT_ID
1     Id_359328     101265     
2     Id_359329     101788     Id_359328
3     Id_359330     175     Id_359329
4     Id_359331     101766     Id_359329
5     Id_388909     100654     Id_359328
6     Id_388191     100672     Id_359328
7     Id_388927     100625     Id_359328
8     Id_359332     101766     Id_359328
9     Id_359380     100651     Id_359332
10     Id_359334     100643     Id_359332
11     Id_359337     101788     Id_359334
12     Id_359338     175     Id_359337
13     Id_359339     101766     Id_359337
14     Id_359336     101860     Id_359332Edited by: ravit on Apr 20, 2010 4:43 PM

Answer 1

Hi,
I agree with damorgan, XQuery processing requires some ressources and it could help to have some of your system specs and settings.
Anyway, tuning is not my specialty but I can give you an alternative.
You could move the XQuery logic to an XSLT stylesheet and transform the document before passing it to XMLTable.
For example :
Creating sample data like yours...
CREATE TABLE test_tree AS
SELECT n,
       power(10,p)*(1+trunc(n/power(10,p))) p
FROM (
  SELECT level n,
         nvl(length(regexp_substr(level,'0+$')),0)+1 p
  FROM dual
  CONNECT BY level < 10000
UNION ALL
SELECT 10000, null FROM dual
CREATE TABLE test_xml (
doc1 xmltype,
doc2 xmltype
DECLARE
ctx dbms_xmlgen.ctxHandle;
doc xmltype;
xsl xmltype := xmltype(
'<?xml version="1.0"?>
<xsl:stylesheet version="1.0" xmlns:xsl="http://www.w3.org/1999/XSL/Transform">
<xsl:output method="xml"/>
<xsl:template match="/">
<root>
  <xsl:apply-templates/>
</root>
</xsl:template>
<xsl:template match="*">
<ROW Id="{name(.)}">
  <PROD_ID><xsl:value-of select="@prodId"/></PROD_ID>
  <PRNT_ID><xsl:value-of select="name(..)"/></PRNT_ID>
</ROW>
<xsl:apply-templates/>
</xsl:template>
</xsl:stylesheet>'
BEGIN
-- hierarchical XML generation :
ctx := dbms_xmlgen.newContextFromHierarchy(
'select level,
         xmlelement(evalname(''Id_''||to_char(n,''fm09999'')),
           xmlattributes( to_char(n) as "prodId",
                          ''C''||to_char(n) as "prodName" )
  from test_tree
  connect by prior n = p
  start with p is null'
doc := dbms_xmlgen.getXMLType(ctx);
dbms_xmlgen.closeContext(ctx);
INSERT INTO test_xml VALUES(doc, doc.transform(xsl));
COMMIT;
END;
/Now, TEST_XML.DOC1 holds the original document, DOC2 holds the result of the transformation :
<root>
  <ROW Id="Id_10000">
    <PROD_ID>10000</PROD_ID>
    <PRNT_ID>#document</PRNT_ID>
  </ROW>
  <ROW Id="Id_01000">
    <PROD_ID>1000</PROD_ID>
    <PRNT_ID>Id_10000</PRNT_ID>
  </ROW>
  <ROW Id="Id_00100">
    <PROD_ID>100</PROD_ID>
    <PRNT_ID>Id_01000</PRNT_ID>
  </ROW>
  <ROW Id="Id_00010">
    <PROD_ID>10</PROD_ID>
    <PRNT_ID>Id_00100</PRNT_ID>
  </ROW>
  <ROW Id="Id_09997">
    <PROD_ID>9997</PROD_ID>
    <PRNT_ID>Id_10000</PRNT_ID>
  </ROW>
  <ROW Id="Id_09998">
    <PROD_ID>9998</PROD_ID>
    <PRNT_ID>Id_10000</PRNT_ID>
  </ROW>
  <ROW Id="Id_09999">
    <PROD_ID>9999</PROD_ID>
    <PRNT_ID>Id_10000</PRNT_ID>
  </ROW>
</root>Then we can execute the query on DOC2:
SELECT rn, id, prod_id, prnt_id
FROM test_xml x,
     XMLTable(
      '/root/ROW'
      passing x.doc2
      columns
        id      VARCHAR2(100) path '@Id',
        prod_id NUMBER        path 'PROD_ID',
        prnt_id VARCHAR2(100) path 'PRNT_ID',
        rn      FOR ORDINALITY
;On my environment, 10000 records fetched in about 18s.