XML to hierarchical table

Question

I would like to convert XML in to a simple table structure, with:
- 1 row per tag
- a tag_id for each tag
- a parent tag_id for each tag
- the attribute for each tag
E.g. for the following
DROP TABLE test_xml;
CREATE TABLE test_xml OF XMLType;
INSERT INTO test_xml
VALUES
  (XMLType
    ('<tag1>
  <tag2_1>
    <tag2_1_1 Value1="A">
      <tag2_1_1_1 Value2="B" Value3="C"/>
    </tag2_1_1>
    <tag2_1_2 Value1="D">
      <tag2_1_2_1 Value2="E" Value3="F"/>
    </tag2_1_2>
    <tag2_1_3 Value1="G">
      <tag2_1_3_1 Value2="H" Value3="I"/>
    </tag2_1_3>
    <tag2_1_4 Value1="J">
      <tag2_1_4_1 Value2="K" Value3="L"/>
    </tag2_1_4>
  </tag2_1>
  <tag2_2>
    <tag2_2_1 Value4="M"/>
  </tag2_2>
  <tag2_3>
    <tag2_3_1 Value5="N">
      <tag2_3_1_1>
        <tag2_3_1_1_1>
          <tag2_3_1_1_1_1 Value6="O" Value7="P">
            <tag2_3_1_1_1_1_1 Value2="Q" Value3="R"/>
          </tag2_3_1_1_1_1>
        </tag2_3_1_1_1>
      </tag2_3_1_1>
      <tag2_3_1_2>
        <tag2_3_1_2_1 Value8="S" Value9="T"/>
      </tag2_3_1_2>
    </tag2_3_1>
  </tag2_3>
</tag1>'
set long 1000
select * from test_xml;
SYS_NC_ROWINFO$
<tag1>
  <tag2_1>
    <tag2_1_1 Value1="A">
      <tag2_1_1_1 Value2="B" Value3="C"/>
    </tag2_1_1>
    <tag2_1_2 Value1="D">
      <tag2_1_2_1 Value2="E" Value3="F"/>
    </tag2_1_2>
    <tag2_1_3 Value1="G">
      <tag2_1_3_1 Value2="H" Value3="I"/>
    </tag2_1_3>
    <tag2_1_4 Value1="J">
      <tag2_1_4_1 Value2="K" Value3="L"/>
    </tag2_1_4>
  </tag2_1>
  <tag2_2>
    <tag2_2_1 Value4="M"/>
  </tag2_2>
  <tag2_3>
    <tag2_3_1 Value5="N">
      <tag2_3_1_1>
        <tag2_3_1_1_1>
          <tag2_3_1_1_1_1 Value6="O" Value7="P">
            <tag2_3_1_1_1_1_1 Value2="Q" Value3="R"/>
          </tag2_3_1_1_1_1>
        </tag2_3_1_1_1>
      </tag2_3_1_1>
      <tag2_3_1_2>
        <tag2_3_1_2_1 Value8="S" Value9="T"/>
      </tag2_3_1_2>
    </tag2_3_1>
  </tag2_3>
</tag1>I want to generate the following output
tag_id parent_tag_id tag_name             attribute
     1               tag1
     2             1 tag2_1
     3             2 tag2_1_1             Value1="A"
     4             3 tag2_1_1_1           Value2="B" Value3="C"
     5             2 tag2_1_2             Value1="D"
     6             5 tag2_1_2_1           Value2="E" Value3="F"
     7             2 tag2_1_3             Value1="G"
     8             7 tag2_1_3_1           Value2="H" Value3="I"
     9             2 tag2_1_4             Value1="J"
    10             9 tag2_1_4_1           Value2="K" Value3="L"
    11             1 tag2_2              
    12            11 tag2_2_1             Value4="M"
    13             1 tag2_3              
    14            13 tag2_3_1             Value5="N"
    15            14 tag2_3_1_1          
    16            15 tag2_3_1_1_1        
    17            16 tag2_3_1_1_1_1       Value6="O" Value7="P"
    18            17 tag2_3_1_1_1_1_1     Value2="Q" Value3="R"
    19            14 tag2_3_1_2          
    20            19 tag2_3_1_2_1         Value8="S" Value9="T"I can find lots of ways of converting data to XML but not much this other way round
In the actual code, the XML will be in a file - not sure how much of a difference to the solution..
Thanks,
Ben

Answer 1

Hi Ben,
Here's a first "working" solution, but certainly not the optimum considering the way I compute node IDs :
SQL> select * from v$version
  2  ;
BANNER
Oracle Database 10g Enterprise Edition Release 10.2.0.4.0 - 64bi
PL/SQL Release 10.2.0.4.0 - Production
CORE     10.2.0.4.0     Production
TNS for 64-bit Windows: Version 10.2.0.4.0 - Production
NLSRTL Version 10.2.0.4.0 - Production
SQL>
SQL> SELECT x.*
  2  FROM test_xml t,
  3       XMLTable(
  4       'declare function local:getID($e as node()) as xs:integer
  5        {
  6         for $i at $j in $d/descendant::*
  7         where $i is $e
  8         return $j
  9        }
10        ; declare function local:getChildren($e as node()) as element()*
11        {
12         for $i in $e/*
13         return element r
14         {
15          element tag_id {local:getID($i)},
16          element parent_tag_id {if (not($e is $d)) then local:getID($e) else()},
17          element tag_name {local-name($i)},
18          element attr_list {string-join(for $j in $i/@* return concat(local-name($j),''="'',$j,''"'')," ")}
19         }
20          | local:getChildren($i)
21        }; local:getChildren($d)'
22        passing t.object_value as "d"
23        columns tag_id        number        path 'tag_id',
24                parent_tag_id number        path 'parent_tag_id',
25                tag_name      varchar2(30)  path 'tag_name',
26                attr_list     varchar2(100) path 'attr_list'
27       ) x
28  ;
    TAG_ID PARENT_TAG_ID TAG_NAME                       ATTR_LIST
         1               tag1                          
         2             1 tag2_1                        
         3             2 tag2_1_1                       Value1="A"
         4             3 tag2_1_1_1                     Value2="B" Value3="C"
         5             2 tag2_1_2                       Value1="D"
         6             5 tag2_1_2_1                     Value2="E" Value3="F"
         7             2 tag2_1_3                       Value1="G"
         8             7 tag2_1_3_1                     Value2="H" Value3="I"
         9             2 tag2_1_4                       Value1="J"
        10             9 tag2_1_4_1                     Value2="K" Value3="L"
        11             1 tag2_2                        
        12            11 tag2_2_1                       Value4="M"
        13             1 tag2_3                        
        14            13 tag2_3_1                       Value5="N"
        15            14 tag2_3_1_1                    
        16            15 tag2_3_1_1_1                  
        17            16 tag2_3_1_1_1_1                 Value6="O" Value7="P"
        18            17 tag2_3_1_1_1_1_1               Value2="Q" Value3="R"
        19            14 tag2_3_1_2                    
        20            19 tag2_3_1_2_1                   Value8="S" Value9="T"
20 rows selected
Do you really want attributes displayed that way?
Edited by: odie_63 on 23 mars 2011 13:04