Loading an XML with multiple nested tags

Question

I've got some problems dealing with loading a nested tags XML file.
Let's suppose I have such a very simple myxml.xml file:
<ROWDATA>
<ROW>
  <EMPNO>7369</EMPNO>
  <ENAME>SMITH</ENAME>
   </ROW>
<ROW>
  <EMPNO>7902</EMPNO>
  <ENAME>FORD</ENAME>
</ROW>
</ROWDATA>
I can create the following table:
create table EMP
empno NUMBER(4) not null,
ename VARCHAR2(10),
and then inserting the XML file in this way:
insert into EMP
    (empno, ename)
  select extractvalue (column_value, '/ROW/EMPNO'),
      extractvalue (column_value, '/ROW/ENAME'),
  from   table
           (xmlsequence
          (extract
           (xmltype
             (bfilename ('MY_DIR', 'myxml.xml'),
               nls_charset_id ('AL32UTF8')),
             '/ROWDATA/ROW')))
so as to get inserted two rows into my table:
EMPNO   ENAME
7369         SMITH
7902         FORD
Now, and this is my question, let's suppose I have such a “more difficult” XML:
<ROWDATA>
<ROW>
  <COMPANY>
    <EMPNO>7369</EMPNO>
    <ENAME>SMITH</ENAME>
    <EMPNO>1111</EMPNO>
    <ENAME>TAYLOR</ENAME>
  </COMPANY>
</ROW>
<ROW>
  <COMPANY>
   <EMPNO>7902</EMPNO>
   <ENAME>FORD</ENAME>
  </COMPANY>
</ROW>
<ROW>
  <COMPANY>
  </COMPANY>
</ROW>
</ROWDATA>
In this case it seems to me things look harder 'cause for every row that I should insert into my table, I don't know how many “empno” and “ename” I'll find for each /ROW/COMPANY and so how could I define a table since the number of empno and ename columns are “unknown”?
According to you, in that case should I load the whole XML file in an unique XMLType column and than “managing” its content by using EXTRACT and EXTRACTVALUE built-in funcions? But this looks a very difficult job.
My Oracle version is 10gR2 Express Edition
Thanks in advance!

Answer 1

Here's a possible solution using a single pass through the XML data :
with sample_data as (
  select xmltype(
    '<ROWDATA>
    <ROW>
      <COMPANY>
        <EMPNO>7369</EMPNO>
        <ENAME>SMITH</ENAME>
        <EMPNO>1111</EMPNO>
        <ENAME>TAYLOR</ENAME>
      </COMPANY>
    </ROW>
    <ROW>
      <COMPANY>
       <EMPNO>7902</EMPNO>
       <ENAME>FORD</ENAME>
      </COMPANY>
    </ROW> 
  </ROWDATA>'
  ) doc
  from dual
select min(case when node_name = 'EMPNO' then node_value end) as empno
     , min(case when node_name = 'ENAME' then node_value end) as ename
from (
  select trunc((rownum-1)/2) as rn
       , extractvalue(value(x), '.') as node_value
       , value(x).getrootelement() as node_name
  from sample_data t
     , table(
         xmlsequence(extract(t.doc, '/ROWDATA/ROW/COMPANY/*'))
       ) x
group by rn ;
I would be cautious with this approach though, as I'm not sure the ROWNUM projection is entirely deterministic.
As Jason said, it's probably safer to first apply a transformation in order to get a more friendly format.
For example :
<xsl:stylesheet version="1.0" xmlns:xsl="http://www.w3.org/1999/XSL/Transform">
  <xsl:template match="node()|@*">
    <xsl:copy>
      <xsl:apply-templates select="node()|@*"/>
    </xsl:copy>
  </xsl:template>
  <xsl:template match="COMPANY">
    <xsl:apply-templates select="EMPNO"/>
  </xsl:template>
  <xsl:template match="EMPNO">
    <EMP>
      <xsl:copy-of select="."/>
      <xsl:copy-of select="following-sibling::ENAME[1]"/>
    </EMP>
  </xsl:template>
</xsl:stylesheet>