xslt: Multiple output of copy-of when using template recursively - xslt

Question

Following up on this question XSLT: Copy child elements of a complex type only once, i have a problem using that template recursivlely.
Lets say i have a structure like that:
<complexType name="InvoiceType">
<xs:element name="AdressIn" type="AdressType"/>
<xs:element name="AdressOut" type="AdressType" />
<xs:element name="Partner" type="PartnerType" />
<xs:element name="Date" type="DateType"/>
</complexType>
<complexType name="AdressType">
<xs:element name="Person" type="PersonType"/>
<xs:element name="City" type="string" />
<xs:element name="Nr" type="int" />
</complexType>
<complexType name="PartnerType">
<xs:element name="Person" type="PersonType"/>
<xs:element name="Age" type="int"/>
</complexType>
<complexType name="PersonType">
<xs:element name="Name" type="string"/>
</complexType>
And this template:
<xsl:template match="xs:complexType" mode="copy">
<xsl:apply-templates select="/xs:schema/xs:complexType[#name=current()/xs:sequence/xs:element/#type]"
<xsl:copy-of copy-namespaces="no" select="."></xsl:copy-of>
</xsl:template>
I am trying to copy all referenced complexType with this template. My problem now is, PersonType
is referenced both by AdressType and PartnerType (see example). In my outputfile, this type is printed twice. (simplified example)
How can i work around that?
Thanx for any help!

Answer 1

What do you actually want to do?
(1) Display all the complex types in the schema document? (that's easy).
(2) Display all the global complex types that are actually referenced from somewhere?
You could do
select(//xs:complexType[#name = //xs:element/#type])
(It's not correct in the presence of namespaces, but neither is your original)
(3) Display a complex type the first time it's encountered in some kind of graph traversal, but not subsequently?
That's much harder and involves some fairly tough recursive coding.
Generally, accessing schema information from source XSD files is difficult; it gets harder the more you want to handle "any schema in the wild" rather than just a few schemas written in the way you would write them. It might be worth standing back and asking what the purpose of this exercise is.