Looking for help in enabling smarter models

[mcbeth]

I'm trying to figure out to put a little intelligence in my models so that the impedance mismatch between the xml and our internal classes is lessened.

Here is an example greatly simplified XSD

<?xml version="1.0" encoding="UTF-8"?>
<xsd:schema elementFormDefault="qualified" xmlns:xsd="http://www.w3.org/2001/XMLSchema">
    <xsd:complexType name="TimeDelta">
        <xsd:annotation>
            <xsd:documentation>
                Offset from Jan 1, 2000 00:00:00Z
            </xsd:documentation>
        </xsd:annotation>
        <xsd:attribute name="sec" type="xsd:unsignedInt"    use="required"/>
        <xsd:attribute name="micros" type="xsd:unsignedInt" use="required"/>
    </xsd:complexType>

    <xsd:element name="Two">
        <xsd:complexType>
            <xsd:sequence>
                <xsd:element name="Before" type="TimeDelta"/>
                <xsd:element name="After" type="TimeDelta"/>
            </xsd:sequence>
        </xsd:complexType>
    </xsd:element>

</xsd:schema>

If we pretend for a second to have a class that stores things in days instead

class MyClass:
  def __init__(self, days):
      self.days = days

I was hoping to put a shim in between the two objects so I could step between the two more cleanly

The Converter infrastructure appears to be for DATA rather than attributes.
I think my worst case is having to do something like

a = example.Two(before=MyClass(end).render_to_time_delta(), after=example.TimeDelta(sec=10, micros=100))

But, I'm hoping I can declare a type-mapping or converter or... so that I tell xsdata "when you see MyClass", always render that as TimeDelta. parsing is much less important, but still a consideration.

Thanks for your time.

[mcbeth]

I'm also looking for ways to ensure that the xsd namespace always appears only in the root node and not on and sub-nodes, but that's a very different question