XML Schema: Component Designators

2.1 Candidate Requirements

Basic Functionality

1. Designators should be unambiguous, designating exactly one component within a schema. However, certain abbreviation and wildcarding constructs may designate collections of components. In any case, each component should have at least one unambiguous designator that designates that component and no other.

2. It should be possible to designate any schema component within a schema. However, some exceptions will be made for certain of the helper components.

3. Designators should provide (or assure) a URI for the unique identification of a schema.

4. If there is more than one designator for a construct, there should be exactly one canonical form of the designator.

Properties of the Designator

1. Designators should be human-readable.

2. Given a designator, it should be possible to get the extended name for the designated component, if one exists for that component.

3. It should be possible to algorithmically and consistently generate the canonical designator for each component, either against an assembled schema or as the schema is assembled.

4. Designators should be parsable with compositional semantics (or, in other words, they should have expressive notation).

5. There should be a formal specification of what constitutes a legal designator.

Compatibility

1. Designators should be URI references.

2. Designators should work well in a RDDL environment.

Non-goals

1. It is not a requirement to designate particle components as distinct from terms.

2. It is not a requirement to tie schema component validity to namespace validity.

3. It is not a requirement that it be possible to construct designators to refer to arbitrary schema components without any knowledge of schema internals.

4. It is not a requirement to be able to tell, from a schema component designator for an element declaration, whether instances of that element declaration must appear in document instances with qualified or unqualified names.

5. It is not a requirement to uniquely assign schema component designators (element, attribute, and type designators) to information items in a document instance given only a set of schema component designators.

2.2 Use Cases

Type references

In general Schema Component Designators can be used to provide references to arbitrary types, whether they are named global types or local or anonymous types. These use cases all benefit from being able to refer to any type declaration.

• Describing the type of an expression, where the type might be an anonymous or local type and a QName is therefore inadequate. There are a number of examples where this could be valuable:

  • Naming the type of something that has been selected, as in XQuery, XPath 2.0, or XSLT 2.0

  • For use wherever types are named, for example to specify a type in an XPath 2.0 or XQuery expression. e,g. element-of(type-t)

  • Enabling DOM3 to expose anonymous types

• Identifying types for TREAT AS (in Query) specifically wrt anonymous and local types.

• Identifying types for function signatures for XQuery, XSLT 2.0 and related specifications.

Element declaration references

These use cases benefit from being able to refer to any element declaration.

• Naming an element declaration which is matched by an arbitrary expression. e.g. Given XPath /one/two/foo, naming that it only matches local element 'foo'.

Stand-off Schema Annotation

Schema component designators can be used to refer to specific parts of a schema to provide an out-of-band annotation capability, for commentary, error reporting, or the association of layered semantics with schema components.

• Writing error messages. There are two classes of error messages in view: errors in an instance and errors in the schema. For example, if an instance has a quantity which violates the constraints of a particular type, such as a range constraint, it is valuable for the error message to be able to name and refer to the specific type whose constraints were violated, even if that is a local element with an anonymous types. An example of a schema error would be that one type is not a valid restriction of another type. This use case benefits from being able to name and refer to type, element, and attribute declarations, as well as element and attribute uses, groups, and possibly facets and particles. It is unclear whether being able to separately refer to particles versus groups is important. Being able to separately refer to attribute and element uses versus attribute and element declarations may be important.

• Schema documentation. The transfer syntax for W3C XML Schema allows for documentation to be attached to components in-line. In many cases it may be preferable to keep the documentation separate from the active definitions, to save bandwidth, to provide for alternative documentation for different communities, and so on. Being able to refer to any schema component that can support an annotation (all but annotation itself) supports this use case.

• Commentary on schemas, such as best practices documents, reviews and comparisons of particular schemas, and so on.

• Associating an additional layer of semantics with schema components, for example, providing information to programming language or application environments to identify code to execute. For example, certain XML data binding frameworks today reference complex types and elements using their own path syntax for the purpose.

Miscellaneous text-based uses within schema-based tools

Having a simple textual way to refer to the abstract components of a schema enables certain kinds of text-based schema-related processing.

• A tool that creates interlinked HTML pages describing a schema.

• String-based tests of type equivalence. Such tests can be used for comparing serialized PSVIs to provide for interoperable tests of type equivalence and processor comparisons.

• String-based comparisons of processors generally, making it possible, for example, to ask and answer the question "Did two validations of the same document on different machines use the same type for a given element?"

• input/output

• user reports

• glue/interfaces among

Other

• Simple selection queries, such as "the type that is the base type of X" without knowledge of its name.

• Creation or initial drafting of a new schema or schema document by selecting components from existing schema.

• RDF assertions about types, etc.

• Formal description requires unique identifier for each declaration and definition component used within the context of the validation episode.

• Enumerating dependencies among schema components. For example, where one complex type restricts another, the local elements in one depend on the local elements of the other.

3.1 Schema Component Designator Syntax

EBNF

3.2 Canonical Schema Component Designators

[Definition: A canonical schema component designator is an absolute schema component designator that is a URI that has been canonicalized according to the rules given in the update to RFC2396 and where the relative schema component designator consists of an xmlns XPointer pointer part (if required) followed by a canonicalized xscd XPointer pointer part. ]

[Definition: A canonicalized xscd XPointer pointer part is an xscd XPointer pointer part where the pointer data is a canonical schema component path. ]

3.3 Equality of Schema Component Designators

Many use cases for schema component designators call for them to be compared for equality. A simple string comparison cannot be used with this scheme because namespace prefixes may vary. However, comparison is still straightforward.

[Definition: Two schema component designators are equal if they are absolute and their URIs are equal or if they are relative to the same schema, and their schema component paths are equal.]

4.1 Schema Component Graph Traversal

An assembled schema consists of a graph of schema components. The following schema component properties have as their values (or a part of their values) other schema components, thus creating the links in the graph which may be traversed to construct a schema component path.

Traversal through the schema component graph occurs through any of the properties by taking a member of one of these lists as the basis for the next [step]:

{type definitions}
{attribute declarations}
{element declarations}
{attribute group definitions}
{model group definitions}
{notation declarations}
{identity constraint definitions}
{facets}
{fundamental facets}
{member type definitions}
{attribute uses}
{particles}

Traversal through the schema component graph occurs through any of the following properties by taking that schema component as the basis for the next [step].

{type definition}
{item type definition}
{attribute wildcard}
{model group}
{attribute declaration}
{term}
{base type definition}
{primitive type definition}
{substitution group affiliation}
{referenced key}

Traversal through the schema component graph occurs to the following properties, but in a somewhat special way, as referring to a collection of annotations as a group, rather than to an individual annotation component. This is because we cannot distinguish between individual annotations.

{annotations}
{annotation}

Traversal through the schema component graph occurs through the following properties, but in different ways depending on what kind of value it has:

{content type}
{scope}

Traversal will proceed through the particle of the {content type} property if a content model is present, or through the {content type} property directly if it is a simple type definition. Traversal will proceed through the {scope} property unless it has the value "global".

4.2 Schema Component Path Syntax

In general the syntax of a step of the schema component path has the following form:

where ns-prefix is bound to [target namespace]. In the context of schema component designators the namespace prefixes will be bound via the xmlns XPointer scheme; in the context of an XML document the namespace prefixes will be bound in the conventional way (using the [in-scope namespaces] of the element information item); other host languages (such as XQuery) will define their own namespace binding rules.

The syntax provides various abbreviations as defined below.

Steps in the path are separated by a slash:

except in the case of the schema-as-a-whole component, which is represented by a bare slash already, so no additional separator is necessary.

/type(my:title)

/element(my:title)

/attribute(my:title)

/attributeGroup(my:title)

/type(my:title)/facet(pattern)

<xs:complexType name="articleType">
   <xs:sequence>
      <xs:element ref="my:section"/>
      <xs:element name="appendix" type="my:sectionType"/>
   </xs:sequence>
</xs:complexType>

<xs:element name="chapter">
   <xs:complexType>
      <xs:sequence>
         <xs:element ref="my:title" minOccurs="0" maxOccurs="unbounded"/>
         <xs:any namespace="##other" minOccurs="0" maxOccurs="unbounded"/>
      </xs:sequence>
      <xs:attribute name="name" type="xs:string"/>
      <xs:anyAttribute namespace="##other" use="optional"/>
   </xs:complexType>
</xs:element>

/type(my:articleType)/sequence()/element(my:section)

/type(my:articleType)/sequence()/element(my:appendix)

/element(my:chapter)/type()

/element(my:chapter)/type()/sequence()/any()

/element(my:chapter)/type()/attribute(name)

4.3 Canonical Schema Component Paths

[Definition: The canonical schema component path of a component is a distinguished valid component path that uniquely identifies that particular component, that has as few steps as possible, and that can be deterministically constructed. ]

The remainder of this section consists of a constructive definition of canonical schema component paths: a path produced by this construction will be the canonical one.

[Definition: The current schema component is a schema component for which there is a valid canonical path through the schema component graph.]

[Definition: The target schema component is a schema component linked to the current schema component via one of the schema component properties defined previously.]

[Definition: The component relationship is the name of the schema component property on the current schema component which references the target schema component. ]

[Definition: The ancestor set of the target schema component is the set of schema components on the valid canonical path to the current schema component together with the current schema component.]

[Definition: The base type set of a schema component is the set consisting of the {base type definition} of that schema component and the {base type definition} of every member of the set. ]

Constructively, this set can be computed by adding the {base type definition} to the set, and iterating on that type component, until you see a component whose {target namespace} property is "http://www.w3.org/2001/XMLSchema" and whose {name} property is "anyType".

[Definition: The extended base type set of a schema component is the set consisting of the {base type definition} of that schema component if its {derivation} method is "extension" and the {base type definition} of every member of the set whose {derivation} method is "extension". ]

[Definition: The base facet set of a schema component is the set consisting of all members of the {facets} property of members of the base type set of that schema component.]

[Definition: Two facets are the same if they are the same kind of component (e.g. both are length facets) and their {value} properties have identical values.]

[Definition: The base attribute use set of a schema component is the set consisting of all members of the {attribute uses} property of members of the base type set of that schema component.]

[Definition: Two attribute uses are the same if the {name} and {target namespace} of their {attribute declaration} properties are equal. ]

(Note that this is not equality which requires checking the {type definition} as well. However, the rules for component consistency will ensure that checking the {name} and {target namespace} suffices for our purposes here.)

[Definition: The particle set of a schema component is the set consisting of the {content type} of that schema component with the members of the {particles} property of the {term} of any member of the set.]

[Definition: The base particle set of a schema component is the union of all particle sets of all the members of the extended base type set of that schema component.]

[Definition: Two particles are the same if the implementation reports them as the same.] Informally, implementations will report particles that are "inherited" from base types as "the same" as the corresponding particle in the base type. Future revisions of the XML Schema recommendation will specify more precisely the conditions under which particles should be regarded as "the same".

Traversal where the current schema component is the schema-as-a-whole component to any target schema component is always permitted in the construction of a canonical path. If the current schema component is not the schema-as-a-whole schema component, then traversal from the current schema component to the target schema component is forbidden in the construction of a canonical path if any of the following conditions is met:

1. the target schema component is an identity constraint definition

2. the target schema component is a simple or complex type definition whose {name} property is not absent

3. the target schema component is an element or attribute declaration whose {scope} property is global

4. the target schema component has a {scope} property whose value is a schema component and that schema component is not a member of the ancestor set of the target schema component

5. the component relationship is {scope}, {substitution group affiliation}, {base type definition}, {primitive type definition}, or {referenced key}

6. the target schema component is a facet and there exists some facet in its base facet set of the target schema component that is the same as the facet

7. the current schema component is a complex type, the target schema component is the {attribute declaration} of an attribute use component and there exists some attribute use in the base attribute use set of the current schema component that is the same as the attribute use whose {attribute declaration} is the target schema component

8. the target schema component is the {term} of a particle and there exists some particle in the base particle set of the current schema component that is the same as the target schema component

Informally, the first two conditions ensure that the canonical path to an identity constraint or a named type is the one that flows directly from the schema-as-a-whole component, the third condition ensures that the canonical path to element and attribute declarations is likewise the global one (if there is one), the last condition accounts for element and attribute declarations stemming from named model groups and attribute groups, the fifth ensures that the canonical path doesn't follow "backwards" paths, and the final three conditions ensure the canonical path to "inherited" components goes through their component of origin.

In general, it is not possible to obtain the canonical path for a schema component without access to the schema component graph. In particular, canonicalizing some non-canonical path, or determining whether a schema component path is canonical cannot be performed without such access.

For example, consider the path /type(x:foo)/element(x:bar).

For this schema fragment, this represents a canonical path to the local element bar:

<complexType name="foo">
  <sequence>
     <element name="bar" type="x:barType"/>
  </sequence>
</complexType>

But for this schema fragment, it represents a non-canonical path. The canonical path would be /group(x:barGroup)/element(x:bar).

<group name="barGroup">
  <sequence>
     <element name="bar" type="x:barType"/>
  </sequence>
</group>
<complexType name="foo">
  <sequence>
    <group ref="x:barGroup"/>
  </sequence>
</complexType>

4.4 Equality of Schema Component Paths

[Definition: Two schema component paths are equal if they have the same number of steps, and each step in one path is equal to the corresponding step in the other. Steps are equal if all of their properties have the same value.]

Note: This definition of equality neither relies on the notion of 'the same component' nor defines component equivalence. A schema component may have multiple schema component paths. Although these schema component paths address 'the same component', they are not equal. It is true, however, that two canonical paths to the same component will be equal.

5.1 Schema Component Path Conformance

Schema component paths may be used by other specifications. For such usage, this specification does not define any criteria for conformance and relies on other specifications to specify criteria for conformance of implementations.

5.2 Schema Component Designator Conformance

Conforming XPointer processors claiming to support the xscd() scheme must conform to the behavior defined in this specification and XPointer xmlns() scheme specification.