针对具有抽象类型的xsd模式验证xml

Question

Can anyone offer any explanation for the tricky behavior related to validating documents with abstract type definition?

This xml/xsd pair validates properly:

basedef.xsd:

<xsd:schema xmlns:xsd="http://www.w3.org/2001/XMLSchema" >
  <xsd:complexType name="abstractType" abstract="true"/>
</xsd:schema>

concrete.xsd:

<xsd:schema xmlns:xsd="http://www.w3.org/2001/XMLSchema" >

  <xsd:include schemaLocation="basedef.xsd"/>

  <xsd:element type="abstractType" name="object" />

  <xsd:complexType name="concreteType" >
      <xsd:complexContent>
          <xsd:extension base="abstractType">
              <xsd:sequence >
                  <xsd:element type="xsd:string" name="field" />
              </xsd:sequence>
          </xsd:extension>
      </xsd:complexContent>
  </xsd:complexType>
</xsd:schema>

concrete.xml:

<object xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:type="concreteType">
  <field>value</field>
</object>

This however, does not:

testoper.xsd:

<xsd:schema xmlns:xsd="http://www.w3.org/2001/XMLSchema">
  <xsd:include schemaLocation="basedef.xsd"/>

  <xsd:complexType name="dbOper">
    <xsd:sequence>
      <xsd:element name="operation-create" type="createOperation"/>
    </xsd:sequence>
  </xsd:complexType>
  <xsd:complexType name="createOperation">
    <xsd:sequence>
      <xsd:element name="object" type="abstractType"/>
    </xsd:sequence>
  </xsd:complexType>
  <xsd:element name="dbOper" type="dbOper"/>
</xsd:schema>

testoper.xml:

<dbOper>
  <operation-create>
    <object xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:type="concreteType" xsi:noNamespaceSchemaLocation="concrete.xsd">
      <field>value</field>
    </object>
  </operation-create>
</dbOper>

The validation tool is xmllint from the libmxl2 package. Errors I'm getting are as follows:

testoper.xml:3: element object: Schemas validity error : Element 'object', attribute '{w3.org/2001/XMLSchema-instance}type': 
    **The QName value 'concreteType' of the xsi:type attribute does not resolve to a type definition**.

testoper.xml:3: element object: Schemas validity error : 
    Element 'object': **The type definition is abstract.** 

Answer 1

Essentially, you can't validate instances against an abstract type. The only use for an abstract type is as a base for defining concrete types by extension or restriction. The concrete subtype can be defined either in the schema (as the type of the element you want to validate), or in the instance (using the xsi:type attribute) but it must be defined somewhere.

Validation of testoper.xml is failing simply because the xsi:type attribute names a nonexistent type. If the attribute is present then it must name a concrete subtype of the declared type of the element.

Further, in response to your comment:

Whether or not the processor takes any account of the xsi:noNamespaceSchemaLocation depends on what processor you are using and how it is configured. Note in particular this clause in XSD Part 1 section 4.3.2:

xsi:schemaLocation and xsi:noNamespaceSchemaLocation [attributes] can occur on any element. However, it is an error if such an attribute occurs after the first appearance of an element or attribute information item within an element information item initially ·validated· whose [namespace name] it addresses. According to the rules of Layer 1: Summary of the Schema-validity Assessment Core (§4.1), the corresponding schema may be lazily assembled, but is otherwise stable throughout ·assessment·. Although schema location attributes can occur on any element, and can be processed incrementally as discovered, their effect is essentially global to the ·assessment·. Definitions and declarations remain in effect beyond the scope of the element on which the binding is declared.

What this means is that an xsi:schemaLocation attribute appearing half way through the instance document is problematic. If the processor takes account of them, then it must take account of them for the whole instance document, it can't change the schema half way through validating. The spec says that what you are doing is "an error", but it's notoriously vague as to what this actually means, and processors vary.

Answer 2

But... If we define:

<xsd:schema xmlns:xsd="http://www.w3.org/2001/XMLSchema">
  <xsd:include schemaLocation="concrete.xsd"/>

  <xsd:complexType name="dbOper">
    <xsd:sequence>
      <xsd:element name="operation-create" type="createOperation"/>
    </xsd:sequence>
  </xsd:complexType>
  <xsd:complexType name="createOperation">
    <xsd:sequence>
      <xsd:element name="object" type="abstractType"/>
    </xsd:sequence>
  </xsd:complexType>
  <xsd:element name="dbOper" type="dbOper"/>
</xsd:schema>

This will validate

xmllint --noout --path . --schema testoper2.xsd test.xml test.xml 

There is a problem with

  xsi:noNamespaceSchemaLocation="concrete.xsd"

But basically, yes, I fully agree using abstract types to define an element looks like a bad idea, using them as base types looks more sensible.