XDB annotations on complex XML Schema

Question

I am currently testing this in 11gr2 Express, while I wait for our DBA to upgrade our main AIX dev instance to 11gr2 enterprise edition.
I have a fairly complex xmlschema, provided by the IPTC and trying to annotate it to override some of the default sqltype mappings, but running into problems with how/where to annotate.
From what I have learned so far, annotations to override the sqltype mappings can only be done on complexType and simpleTypes. ComplexTypes must map to a sql object type, and simpleTypes map to a basic atomic sqlType.
How would you handle an instance where a complex type extends an xs:string.. but need to have this string map to a CLOB and not a varchar2(4000)?
<xs:element name="inlineData">
                                                  <xs:annotation>
                                                       <xs:documentation>A rendition of the content using plain-text or encoded inline data</xs:documentation>
                                                  </xs:annotation>
                                                  <xs:complexType>
                                                       <xs:simpleContent>
                                                            <xs:extension base="xs:string">
                                                                 <xs:attributeGroup ref="newsContentAttributes" />
                                                                 <xs:attributeGroup ref="newsContentTypeAttributes" />
                                                                 <xs:attribute name="encoding" type="QCodeType">
                                                                      <xs:annotation>
                                                                           <xs:documentation>The encoding applied to the content before inclusion</xs:documentation>
                                                                      </xs:annotation>
                                                                 </xs:attribute>
                                                                 <xs:attributeGroup ref="newsContentCharacteristics" />
                                                                 <xs:attributeGroup ref="i18nAttributes" />
                                                                 <xs:anyAttribute namespace="##other" processContents="lax" />
                                                            </xs:extension>
                                                       </xs:simpleContent>
                                                  </xs:complexType>
                                             </xs:element>Adding the xdb: mapUnboundedStringToLob annotation to element "inlineData" fails... not valid... and can't add it to the extension either.. so would a viable valid option in this case, be to create a simpleType, call it say "LargeStringToClob"..with a restriction of type xs:String... then annotate this simpleType to map it to a CLOB.
Then change the inlineData element extension from xs:string to "LargeStringToClob" ? I plan to use an XMLType column with BinaryXML storage...
I was hoping to not have to modify the xml schema too much, but I guess its to be expected when trying to map it into Oracle.
I am faced with a similar issue when trying to annotate for Timestamp with time zone...
some elements are declared as unions of other types. From what I have read, Oracle maps these to varchar2(4000)
Example:
<xs:simpleType name="DateOptTimeType">
          <xs:annotation>
               <xs:documentation>The type of a date (required) and a time (optional).</xs:documentation>
          </xs:annotation>
          <xs:union memberTypes="xs:date xs:dateTime" />
     </xs:simpleType>Even though this is a simpleType... I cannot annotate it to map it to a TimeStamp with Timezone ... would I have to again... create simpleTypes for each of the memberTypes in the union, annotate each one to map to Timestampt with time zone.. then use these new simpleTypes in the union?
Thanks for any tips or advice!

Answer 1

odie_63 wrote:
so if using Binary XML, you are stuck with what Oracle gives you? for instance, the inlineData element is an unbounded string in the schema... Oracle maps it to varchar2(4000).No, Binary XML stores data differently.
Consider this simple example :
SQL> begin
2   dbms_xmlschema.registerSchema(
3   schemaURL => 'test.xsd',
4   schemaDoc => '<?xml version="1.0"?>
5  <xs:schema xmlns:xs="http://www.w3.org/2001/XMLSchema" xmlns:xdb="http://xmlns.oracle.com/xdb">
6  <xs:element name="note" xdb:defaultTable="NOTES_TABLE">
7    <xs:complexType>
8      <xs:sequence>
9        <xs:element name="content" type="xs:string"/>
10        <xs:element name="dt" type="xs:dateTime"/>
11      </xs:sequence>
12    </xs:complexType>
13  </xs:element>
14  </xs:schema>',
15   local => true,
16   genTypes => false,
17   genTables => false,
18   enableHierarchy => dbms_xmlschema.ENABLE_HIERARCHY_NONE,
19   options => dbms_xmlschema.REGISTER_BINARYXML
20   );
21  end;
22  /
PL/SQL procedure successfully completed
SQL> CREATE TABLE notes_table OF XMLTYPE
2  XMLTYPE STORE AS binary xml
3  XMLSCHEMA "test.xsd"
4  ELEMENT "note"
5  ;
Table created
SQL> insert into notes_table values(xmltype(
2  '<note>
3  <content>'||lpad(to_clob('X'),8000,'X')||'</content>
4  <dt>2011-11-16T11:56:23+01:00</dt>
5  </note>'));
1 row inserted
SQL> select x.dt
2       , x.content
3       , length(x.content)
4  from notes_table t
5     , xmltable('/note' passing t.object_value
6       columns dt      timestamp with time zone  path 'dt'
7             , content clob                      path 'content'
8       ) x
9  ;
DT                                CONTENT                                                                          LENGTH(X.CONTENT)
16/11/11 11:56:23,000000 +01:00   XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX              8000No issue in storing large strings.Very cool... that works... .. the string limit exceeded expcetion I got earlier was no fault to Binary XML.. but me pasting in too much text in SQLDeveloper insert statement.
I think I may have been on this for too long... brain is getting fuzzy ;-)
The approach you outline shows promise and I will go with BinaryXML storage option.
I need to decide whether to use virtual columns or non-trivial columns at this point, to enforce integrity constraints and possibly some partitioning. We do have a few relational lookup tables to tie into this as well... and will also need to implement VPD.