22 Documentation Elements

Table of contents

This chapter describes a module which may be used for the documentation of the XML elements and element classes which make up any markup scheme, in particular that described by the TEI Guidelines, and also for the automatic generation of schemas or DTDs conforming to that documentation. It should be used also by those wishing to customize or modify these Guidelines in a conformant manner, as further described in chapters 23.2 Personalization and Customization and 23.3 Conformance and may also be useful in the documentation of any other comparable encoding scheme, even though it contains some aspects which are specific to the TEI and may not be generally applicable.

An overview of the kind of processing environment envisaged for the module described by this chapter may be helpful. In the remainder of this chapter we refer to software which provides such a processing environment as an ODD processor.79 Like any other piece of XML software, an ODD processor may be instantiated in many ways: the current system uses a number of XSLT stylesheets which are freely available from the TEI, but this specification makes no particular assumptions about the tools which will be used to provide an ODD processing environment.
As the name suggests, an ODD processor uses a single XML document to generate multiple outputs. These outputs will include:

The input required to generate these outputs consists of running prose, and special purpose elements documenting the components (elements, classes, etc.) which are to be declared in the chosen schema language. All of this input is encoded in XML using elements defined in this chapter. In order to support more than one schema language, these elements constitute a comparatively high-level model which can then be mapped by an ODD processor to the specific constructs appropriate for the schema language in use. Although some modern schema languages such as RELAX NG or W3C Schema natively support self-documentary features of this kind, we have chosen to retain the ODD model, if only for reasons of compatibility with earlier versions of these Guidelines. We do however use the ISO standard XML schema language RELAX NG (http://www.relaxng.org) as a means of declaring content models, rather than inventing a completely new XML-based representation for them. We also use the ISO Schematron language to define additional constraints beyond those expressed in the content model, as further discussed in 22.4.4.2 Additional constraints below.

In the TEI system, a schema is built by combining element and attribute declarations, more or less as required. Each element is documented by an appropriate specification element and has an identifier unique across the whole TEI scheme. For convenience, these specifications are grouped into a number of discrete modules, which can also be combined more or less as required. Each major chapter of these Guidelines defines a distinct module. Each module declares a number of elements specific to that module, and may also populate particular classes. All classes are available globally, irrespective of the module in which they are declared; particular modules extend the meaning of a class by adding elements or attributes to it. Wherever possible, element content models are defined in terms of classes rather than in terms of specific elements. Modules can also declare particular patterns, which act as short-cuts for commonly used content models or class references.

In the present chapter, we discuss the elements needed to support this system. In addition, section 22.1 Phrase Level Documentary Elements discusses some general purpose elements which may be useful in any kind of technical documentation, wherever there is need to talk about technical features of an XML encoding such as element names and attributes. Section 22.2 Modules and Schemas discusses the elements which are used to document XML modules and their high-level components. Section 22.3 Specification Elements discusses the elements which document XML elements and their attributes, element classes, and generic patterns or macros. Finally, section 22.7 Module for Documention Elements provides a summary overview of the elements provided by the module.

22.1 Phrase Level Documentary Elements

22.1.1 Phrase Level Terms

In any kind of technical documentation, the following phrase-level elements may be found useful for marking up strings of text which need to be distinguished from the running text because they come from some formal language:
  • code contains literal code from some formal language such as a programming language.
    lang(formal language) a name identifying the formal language in which the code is expressed
  • ident (identifier) contains an identifier or name for an object of some kind in a formal language.
Like other phrase-level elements used to indicate the semantics of a typographically distinct string, these are members of the model.emph class. They are available anywhere that running prose is permitted when the module defined by this chapter is included in a schema.
The code and ident elements are intended for use when citing brief passages in some formal language such as a programming language, as in the following example:
<p>If the variable <ident>z</ident> has a value of zero, a statement
such as <code>x=y/z</code> will usually cause a fatal error.</p>

If the cited phrase is a mathematical or chemical formula, the more specific formula element defined by the figures module (14.2 Formulæ and Mathematical Expressions) may be more appropriate.

A further group of similar phrase-level elements is also defined for the special case of representing parts of an XML document:
  • att (attribute) contains the name of an attribute appearing within running text.
  • gi (element name) contains the name (generic identifier) of an element.
  • tag contains text of a complete start- or end-tag, possibly including attribute specifications, but excluding the opening and closing markup delimiter characters.
  • val (value) contains a single attribute value.
These elements constitute the model.phrase.xml class, which is also a subclass of model.phrase. They are also available anywhere that running prose is permitted when the module defined by this chapter is included in a schema.
As an example of the recommended use of these elements, we quote from an imaginary TEI working paper:
<p>The <gi>gi</gi> element is used to tag
element names when they appear in the text; the
<gi>tag</gi> element however is used to show how a tag as
such might appear. So one might talk of an occurrence of the
<gi>blort</gi> element which had been tagged
<tag>blort type='runcible'</tag>. The
<att>type</att> attribute may take any name token as
value; the default value is <val>spqr</val>, in memory of
its creator.</p>
Within technical documentation, it is also often necessary to provide more extended examples of usage or to present passages of markup for discussion. The following special elements are provided for these purposes:
  • eg (example) contains any kind of illustrative example.
  • egXML (example of XML) contains a single well-formed XML fragment demonstrating the use of some XML element or attribute, in which the egXML element itself functions as the root element.

Like the code element, the egXML element is used to mark strings of formal code, or passages of XML markup. The eg element may be used to enclose any kind of example, which will typically be rendered as a distinct block, possibly using particular formatting conventions, when the document is processed. It is a specialised form of the more general q element provided by the TEI core module. In documents containing examples of XML markup, the egXML element should be used for preference, as further discussed below in 22.4.2 Exemplification of Components, since the content of this element can be checked for well-formedness.

The global xml:space is available to indicate that white space should be preserved within the content of these elements.

These elements are added to the class model.egLike when this module is included in a schema. That class is a part of the general model.inter class, thus permitting eg or egXML elements to appear either within or between paragraph-like elements.

22.1.2 Element and Attribute Descriptions

Within the body of a document using this module, the following elements may be used to reference parts of the specification elements discussed in section 22.3 Specification Elements, in particular the brief prose descriptions these provide for elements and attributes.
  • specList (specification list) marks where a list of descriptions is to be inserted into the prose documentation.
  • specDesc/ (specification description) indicates that a description of the specified element or class should be included at this point within a document.
TEI practice requires that a specList listing the elements under discussion introduce each subsection of a module's documentation. The source for the present section, for example, begins as follows:
<div3>
 <head>Element and attribute descriptions</head>
 <p>Within the body of a document using this module, the following
   elements may be used to reference parts of the specification elements
   discussed in section <ptr target="#TDcrystals"/>, in particular the
   brief prose descriptions these provide for elements and attributes.
 <specList>
   <specDesc key="specList"/>
   <specDesc key="specDesc"/>
  </specList>
 </p>
 <p>TEI practice requires that a <gi>specList</gi> listing the elements
   ...
 </p>
<!-- ... -->
</div3>

When formatting the ptr element in this example, an ODD processor might simply generate the section number and title of the section referred to, perhaps additionally inserting a link to the section. In a similar way, when processing the specDesc elements, an ODD processor must recover relevant details of the elements being specified (specList and specDesc in this case) from their associated declaration elements: typically, the details recovered will include a brief description of the element and its attributes. These, and other data, will be stored in a specification element elsewhere within the current document, or they may be supplied by the ODD processor in some other way, for example from a database. For this reason, the link to the required specification element is always made using a TEI-defined key rather than an XML IDREF value. The ODD processor uses this key as a means of accessing the specification element required. There is no requirement that this be performed using the XML ID/IDREF mechanism, but there is an assumption that the identifier be unique.

A specDesc generates in the documentation the identifier, and also the contents of the desc child of whatever specification element is indicated by its key attribute, as in the example above. Documentation for any attributes specified by the atts attribute will also be generated as an associated attribute list, .

22.2 Modules and Schemas

As mentioned above, the primary purpose of this module is to facilitate the documentation and creation of an XML schema derived from the TEI Guidelines. The following elements are provided for this purpose:
  • schemaSpec (schema specification) generates a TEI-conformant schema and documentation for it.
  • moduleSpec (module specification) documents the structure, content, and purpose of a single module, i.e. a named and externally visible group of declarations.
  • moduleRef (module reference) references a module which is to be incorporated into a schema.
    includesupplies a list of the elements which are to be copied from the specified module into the schema being defined.
    exceptsupplies a list of the elements which are not to be copied from the specified module into the schema being defined.
  • specGrp (specification group) contains any convenient grouping of specifications for use within the current module.
  • specGrpRef/ (reference to a specification group) indicates that the declarations contained by the specGrp referenced should be inserted at this point.
  • attRef/ (attribute pointer) points to the definition of an attribute or group of attributes.
  • elementRef/ points to the specification for some element which is to be included in a schema
A module is a convenient way of grouping together element and other declarations, and associating an externally-visible name with the resulting group. A specification group performs essentially the same function, but the resulting group is not accessible outside the scope of the ODD document in which it is defined, whereas a module can be accessed by name from any TEI schema. Elements, and their attributes, element classes, and patterns are all individually documented using further elements described in section 22.3 Specification Elements below; part of that specification includes the name of a module to which the component belongs.
An ODD processor generating XML DTD or schema fragments from a document marked up according to the recommendations of this chapter will generate such fragments for each moduleSpec element found. For example, the chapter documenting the TEI module for names and dates contains a module specification like the following:
<moduleSpec ident="namesdates">
 <altIdent type="FPI">Names and Dates</altIdent>
 <desc>Additional elements for names and dates</desc>
</moduleSpec>
together with specifications for all the elements, classes, and patterns which make up that module, expressed using elementSpec, classSpec, or macroSpec elements as appropriate. (These elements are discussed in section 22.3 Specification Elements below.) Each of those specifications carries a module attribute, the value of which is namesdates. An ODD processor encountering the moduleSpec element above can thus generate a schema fragment for the TEI namesdates module that includes declarations for all the elements (etc.) which reference it.

In most realistic applications, it will be desirable to combine more than one module together to form a complete schema. A schema consists of references to one or more modules or specification groups, and may also contain explicit declarations or redeclarations of elements (see further 22.5 Building a Schema). Any combination of modules can be used to create a schema 80

A schema can combine references to TEI modules with references to other (non-TEI) modules using different namespaces, for example to include mathematical markup expressed using MathML in a TEI document. By default, the effect of combining modules is to allow all of the components declared by the constituent modules to coexist (where this is syntactically possible: where it is not — for example, because of name clashes — a schema cannot be generated). It is also possible to over-ride declarations contained by a module, as further discussed in section 22.5 Building a Schema

It is often convenient to describe and operate on sets of declarations smaller than the whole, and to document them in a specific order: such collections are called specGrps (specification groups). Individual specGrp elements are identified using the global xml:id attribute, and may then be referenced from any point in an ODD document using the specGrpRef element. This is useful if, for example, it is desired to describe particular groups of elements in a specific sequence. Note however that the order in which element declarations appear within the schema code generated from an ODD file element is not in general affected by the order of declarations within a specGrp.

An ODD processor will generate a piece of schema code corresponding with the declarations contained by a specGrp element in the documentation being output, and a cross-reference to such a piece of schema code when processing a specGrpRef. For example, if the input text reads
<p>This module contains three red elements:
<specGrp xml:id="RED">
  <elementSpec ident="beetroot">
<!-- ... -->
  </elementSpec>
  <elementSpec ident="east">
<!-- ... -->
  </elementSpec>
  <elementSpec ident="rose">
<!-- ... -->
  </elementSpec>
 </specGrp>
and two blue ones:
<specGrp xml:id="BLUE">
  <elementSpec ident="sky">
<!-- ... -->
  </elementSpec>
  <elementSpec ident="bayou">
<!-- ... -->
  </elementSpec>
 </specGrp>
</p>
then the output documentation will replace the two specGrp elements above with a representation of the schema code declaring the elements <beetroot>, <east>, and <rose> and that declaring the elements <sky> and <bayou> respectively. Similarly, if the input text contains elsewhere a passage such as
<div>
 <head>An overview of the imaginary module</head>
 <p>The imaginary module contains declarations for coloured things:
 <specGrpRef target="#RED"/>
  <specGrpRef target="#BLUE"/>
 </p>
</div>
then the specGrpRef elements may be replaced by an appropriate piece of reference text such as ‘The RED elements were declared in section 4.2 above’, or even by a copy of the relevant declarations. As stated above, the order of declarations within the imaginary module described above will not be affected in any way. Indeed, it is possible that the imaginary module will contain declarations not present in any specification group, or that the specification groups will refer to elements that come from different modules. Specification groups are always local to the document in which they are defined, and cannot be referenced externally (unlike modules).

22.3 Specification Elements

The following elements are used to declare elements, classes, and patterns:
  • elementSpec (element specification) documents the structure, content, and purpose of a single element type.
  • classSpec (class specification) contains reference information for a TEI element class; that is a group of elements which appear together in content models, or which share some common attribute, or both.
    generateindicates which alternation and sequence instantiations of a model class may be referenced. By default, all variations are permitted.
  • macroSpec (macro specification) documents the function and implementation of a pattern.

Unlike most elements in the TEI scheme, each of these elements has a fairly rigid internal structure consisting of a large number of child elements which are always presented in the same order. For this reason, we refer to them metaphorically as ‘crystals’. Furthermore, since these elements all describe markup objects in broadly similar ways, they have several child elements in common. In the remainder of this chapter, we discuss first the elements which are common to all the specification elements, and then those which are specific to a particular type.

Specification elements may appear at any point in an ODD document, both between and within paragraphs as well as inside a specGrp element, but the specification element for any particular component may only appear once (except in the case where a modification is being defined; see further 22.5 Building a Schema). The order in which they appear will not affect the order in which they are presented within any schema module generated from the document. In documentation mode, however, an ODD processor will output the schema declarations corresponding with a specification element at the point in the text where they are encountered, provided that they are contained by a specGrp element, as discussed in the previous section. An ODD processor will also associate all declarations found with the nominated module, thus including them within the schema code generated for that module, and it will also generate a full reference description for the object concerned in a catalogue of markup objects. These latter two actions always occur irrespective of whether or not the declaration is included in a specGrp.

22.4 Common Elements

This section discusses the child elements common to all of the specification elements. These child elements are used to specify the naming, description, exemplification, and classification of the specification elements.

22.4.1 Description of Components

  • remarks contains any commentary or discussion about the usage of an element, attribute, class, or entity not otherwise documented within the containing element.
  • listRef (list of references) supplies a list of significant references to places where this element is discussed, in the current document or elsewhere.
One or more desc elements defined by the core module may be used to provide a brief characterization of the intended function of the element, class, value etc. being documented, as in the following example:
<elementSpec module="drama" ident="actor">
 <desc>Name of an actor appearing within a cast list.</desc>
 <desc xml:lang="ja"> 登場人物リスト中にある役者名を示す．</desc>
 <desc xml:lang="it">nome di un attore che appare nella lista dei personaggi.</desc>
<!-- ... -->
</elementSpec>
The remarks element contains any additional commentary about how the item concerned may be used, details of implementation-related issues, suggestions for other ways of treating related information etc., as in the following example:
<elementSpec module="core" ident="foreign">
<!--... -->
 <remarks>
  <p>This element is intended for use only where no other element
     is available to mark the phrase or words concerned. The global
  <att>xml:lang</att> attribute should be used in preference to this element
     where it is intended to mark the language of the whole of some text
     element.</p>
  <p>The <gi>distinct</gi> element may be used to identify phrases
     belonging to sublanguages or registers not generally regarded as true
     languages.</p>
 </remarks>
<!--... -->
</elementSpec>
A specification element will usually conclude with a list of references, each tagged using the standard ptr element, and grouped together into a listRef element: in the case of the foreign element discussed above, the list is as follows:
<listRef>
 <ptr target="#COHQHF"/>
</listRef>
where the value COHQF is the identifier of the section in the Guidelines where this element is fully documented.

22.4.2 Exemplification of Components

  • exemplum groups an example demonstrating the use of an element along with optional paragraphs of commentary.
  • eg (example) contains any kind of illustrative example.
  • egXML (example of XML) contains a single well-formed XML fragment demonstrating the use of some XML element or attribute, in which the egXML element itself functions as the root element.

The exemplum element is used to combine a single illustrative example with an optional paragraph of commentary following or preceding it. The illustrative example itself may be marked up using either the eg or the egXML element.

If an example contains XML markup, it should be marked up using the egXML element. In such a case, it will clearly be necessary to distinguish the markup within the example from the markup of the document itself. In an XML schema environment, this is easily done by using a different name space for the egXML element. For example:
<p>The <gi>term</gi> element may be used to mark any technical term, thus : <egXML xmlns="http://www.tei-c.org/ns/Examples"> This <term>recursion</term> is giving me a headache.</egXML></p>
Alternatively, the XML tagging within an example may be ‘escaped’, either by using entity references, or by wrapping the whole example in a CDATA marked section:
<p>The <gi>term</gi> element may be used to mark any technical term, thus : <egXML xmlns="http://www.tei-c.org/ns/Examples"> This &lt;term&gt;recursion&lt;/term&gt; is giving me a headache.</egXML></p>
or, equivalently:
<p>The <gi>term</gi> element may be used to mark any technical term, thus : <egXML xmlns="http://www.tei-c.org/ns/Examples"><![CDATA[ This <term>recursion</term> is giving me a headache.]]></egXML></p>
However, escaping the markup in this way will make it impossible to validate, and should therefore generally be avoided.

If the XML contained in an example is not well-formed then it must either be enclosed in a CDATA marked section, or ‘escaped’ as above: this applies whether the eg or egXML is used. If it is well-formed but not valid, then it should be enclosed in a CDATA marked section within an egXML.

An egXML element should not be used to tag non-XML examples: the general purpose eg or q elements should be used for such purposes.

22.4.3 Classification of Components

In the TEI scheme elements are assigned to one or more classes, which may themselves have subclasses. The following elements are used to indicate class membership:
  • classes specifies all the classes of which the documented element or class is a member or subclass.
  • memberOf specifies class membership of the parent element or class.
    keyspecifies the identifier for a class of which the documented element or class is a member or subclass

The classes element appears within either the elementSpec or classSpec element. It specifies the classes of which the element or class concerned is a member by means of one or more memberOf child elements. Each such element references a class by means of its key attribute. Classes themselves are defined by the classSpec element described in section 22.4.6 Element Classes below.

For example, to show that the element gi is a member of the class model.phrase.xml, the elementSpec which documents this element contains the following classes element:
<classes>
 <memberOf key="model.phrase.xml"/>
</classes>

22.4.4 Element Specifications

The elementSpec element is used to document an element type, together with its associated attributes. In addition to the elements listed above, it may contain the following subcomponents:
  • content (content model) contains the text of a declaration for the schema documented.
    autoPrefixcontrols whether or not pattern names generated in the corresponding RELAXNG schema source are automatically prefixed to avoid potential nameclashes.
  • constraint (constraint rules) the formal rules of a constraint
  • attList contains documentation for all the attributes associated with this element, as a series of attDef elements.
    org(organization) specifies whether all the attributes in the list are available (org="group") or only one of them (org="choice")
22.4.4.1 Content models

The content of the element content may be expressed in one of two ways. It may use a schema language of some kind, as defined by a pattern called macro.schemaPattern, which is provided by the module defined in this chapter. Alternatively, the legal content for an element may be fully specified using the valList element, described in 22.4.5 Attribute List Specification below.

In the case of the TEI Guidelines, element content models are defined using RELAX NG patterns, although the user may over-ride this by redefining the macro.schemaPattern pattern.

Here is a very simple example
<content>
 <rng:text/>
</content>
The element within whose specification element this content element appears will have a content model which is expressed in RELAX NG as text, using the RELAX NG namespace. This model will be copied unchanged to the output when RELAX NG schemas are being generated. When an XML DTD is being generated, an equivalent declaration (in this case (#PCDATA)) will be output.
Here is a more complex example:
<content>
 <rng:group>
  <rng:ref name="fileDesc"/>
  <rng:zeroOrMore>
   <rng:ref name="model.teiHeaderPart"/>
  </rng:zeroOrMore>
  <rng:optional>
   <rng:ref name="revisionDesc"/>
  </rng:optional>
 </rng:group>
</content>
This is the content model for the teiHeader element, expressed in the RELAX NG syntax, which again is copied unchanged to the output during schema generation. The equivalent DTD notation generated from this is (fileDesc, (%model.teiHeaderPart;)*, revisionDesc?).

The RELAX NG language does not formally distinguish element names, attribute names, class names, or macro names: all names are patterns which are handled in the same way, as the above example shows. Within the TEI scheme, however, different naming conventions are used to distinguish amongst the objects being named. Unqualified names (fileDesc, revisionDesc) are always element names. Names prefixed with model. or att. (e.g. model.teiHeaderPart and att.typed) are always class names. In DTD language, classes are represented by parameter entities (%model.teiHeaderPart; in the above example); see further 1 The TEI Infrastructure.

The RELAXNG pattern names generated by an ODD processor by default include a special prefix, the default value for which is set using the prefix attribute on schemaSpec. The purpose of this is to ensure that the pattern name generated is uniquely identified as belonging to a particular schema, and thus avoid name clashes. For example, in a RELAXNG schema combining the TEI element ident with another element called ident from some other vocabulary, the former will be defined by a pattern called TEI_ident rather than simply ident. Most of the time, this behaviour is entirely transparent to the user; the one occasion when it is not will be where a content model (expressed using RELAXNG syntax) needs explicitly to reference either the TEI ident or the other one. In such a situation, the autoPrefix attribute on content may be used. For example, suppose that we wish to define a content model for term which permits either a TEI ident or the ident defined by some other vocabulary. A suitable content model would be generated from the following content element:
<content autoPrefix="false">
 <rng:choice>
  <rng:ref name="TEI_ident"/>
  <rng:ref name="ident"/>
 </rng:choice>
</content>
22.4.4.2 Additional constraints

In addition to the content element, a set of identified general constraint elements can be provided where rules about the validity of an element can be expressed. They are identifiable in order that a TEI customization may override, delete or change them individually. These elements follow the content element, are permitted as siblings of datatype in attDef, and as children of schemaSpec. The constraints can be expressed in any notation which is found useful; the scheme must be recorded using the scheme attribute of constraint.

The TEI Guidelines themselves provide constraints using the ISO Schematron language. These are normative, and changes to them may affect conformance, just as for content. Although not all processors will be able to process all constraints, they should follow as many as they can.

A complete Schematron document consists of a <schema> element containing <ns> and <pattern> elements; each pattern specifies a rule and a context. In a normal TEI specification it is expected that <ns> and <pattern> elements will be placed wherever suitable for documentation, and extracted into a single Schematron schema, or embedded in another schema language. As a convenience for readers, however, TEI processors should also support the direct placement of Schematron <report> and <assert> elements inside the constraint element within elementSpec; the <pattern> and <rule> containers should then be generated automatically.

Constraints are generally used to model local rules which are outside the scope of conventional schema languages. For example, the TEI element relation has a note that ‘Only one of the attributes active and mutual may be supplied; the attribute passive may be supplied only if the attribute active is supplied.’, which can be written in Schematron as
<constraintSpec ident="activemutual" scheme="isoschematron">
 <desc>Check mutually incompatible attributes</desc>
 <constraint>
  <sch:report test="@active and @mutual">Only one of the attributes
     'active' and 'mutual' may be supplied</sch:report>
  <sch:report test="@passive and not(@active)">the attribute 'passive'
     may be supplied only if the attribute 'active' is
     supplied</sch:report>
 </constraint>
</constraintSpec>
We may also look at rules to apply to a TEI document which is going to be rendered into accessible HTML. Since the rules refer to TEI elements, we need to declare the TEI namespace in Schematron. We can check that some sort of content is available from which the alt attribute of an HTML <img> can be created:
<constraintSpec ident="alt" scheme="isoschematron">
 <constraint>
  <sch:ns prefix="tei" uri="http://www.tei-c.org/ns/1.0"/>
  <sch:pattern id="Alt tags">
   <sch:rule context="tei:figure">
    <sch:report test="not(tei:figDesc or tei:head)"> You should
         provide information in a figure from which
         we can construct an alt attribute in HTML </sch:report></sch:rule></sch:pattern>
 </constraint>
</constraintSpec>
or we might enforce other HTML accessibility rules about tables; note here the use of a report and an assertion within one pattern:
<constraintSpec ident="tables" scheme="isoschematron">
 <constraint>
  <sch:ns prefix="tei" uri="http://www.tei-c.org/ns/1.0"/>
  <sch:pattern id="Tables">
   <sch:rule context="tei:table">
    <sch:assert test="tei:head">A <table> should have a caption, using a <head> element</sch:assert>
    <sch:report test="parent::tei:body">Do not use tables to lay out the document body</sch:report></sch:rule></sch:pattern>
 </constraint>
</constraintSpec>
We can also model contextual constraints in Schematron:
<constraintSpec ident="subclauses" scheme="isoschematron">
 <constraint>
  <sch:rule context="tei:div">
   <sch:assert test="not(tei:div) or count(tei:div)>1">a division must contain
       at least two subdivisions</sch:assert></sch:rule>
 </constraint>
</constraintSpec>
and base requirements on attribute values:
<constraintSpec ident="introtitle" scheme="isoschematron">
 <constraint>
  <sch:assert
    test="tei:fileDesc/tei:titleStmt/tei:title[@type='introductory']"> an introductory component of the title is expected
  </sch:assert>
 </constraint>
</constraintSpec>
<constraintSpec ident="maintitle" scheme="isoschematron">
 <constraint>
  <sch:assert
    test="tei:fileDesc/tei:titleStmt/tei:title[@type='main']"> a main title must be supplied
  </sch:assert>
 </constraint>
</constraintSpec>

22.4.5 Attribute List Specification

The attList element is used to document information about a collection of attributes, either within an elementSpec, or within a classSpec. An attribute list can be organized either as a group of attribute definitions, all of which are understood to be available, or as a choice of attribute definitions, of which only one is understood to be available. An attribute list may also contain nested attribute lists.

The attDef element is used to document a single attribute, using an appropriate selection from the common elements already mentioned and the following which are specific to attributes:
  • attDef (attribute definition) contains the definition of a single attribute.
    usagespecifies the optionality of an attribute or element.
  • datatype specifies the declared value for an attribute, by referring to any datatype defined by the chosen schema language.
  • defaultVal (default value) specifies the default declared value for an attribute.
  • valDesc (value description) specifies any semantic or syntactic constraint on the value that an attribute may take, additional to the information carried by the datatype element.
  • valList (value list) contains one or more valItem elements defining possible values for an attribute.
  • valItem documents a single attribute-value within a list of possible or mandatory items.

The attList within an elementSpec is used to specify only the attributes which are specific to that particular element. Instances of the element may carry other attributes which are declared by the classes of which the element is a member. These extra attributes, which are shared by other elements, or by all elements, are specified by an attList contained within a classSpec element, as described in section 22.4.6 Element Classes below.

22.4.5.1 Datatypes
The datatype element is used to state what kind of value an attribute may have, using whatever facilities are provided by the underlying schema language. For the TEI scheme, expressed in RELAX NG, elements from the RELAX NG namespace may be used, for example
<datatype>
 <rng:text/>
</datatype>
permits any string of Unicode characters not containing markup, and is thus the equivalent of CDATA in DTD language.
The RELAX NG language also provides support for a number of primitive datatypes which may be specified here, using the <rng:data> element: thus one may write
<datatype>
 <rng:data type="boolean"/>
</datatype>
to specify that an element or attribute's contents should conform to the W3C definition for boolean.
Although only one child element may be given, this might be a selector such as rng:choice to indicate multiple possibilities:
<datatype>
 <rng:choice>
  <rng:data type="Date"/>
  <rng:data type="Float"/>
 </rng:choice>
</datatype>
which would permit either a date or a real number. In fact, the child element might be a rng:list element to indicate that a sequence of values is required, a rng:param element to specify a regular expression, or even a list of explicit rng:values. Such usages are permitted by the scheme documented here, but are not recommended when it is desired to remain independent of a particular schema language, since the full generality of one schema language cannot readily be converted to that of another. In the TEI abstract model, datatyping should preferably be carried out either by explicit enumeration of permitted values (using the TEI-specific valList element described below), or by definition of an explicit pattern, using the TEI-specific macroSpec element discussed further in section 22.4.7 Pattern Documentation.
22.4.5.2 Value Specification
The valDesc element may be used to describe constraints on data content in an informal way: for example
<valDesc>must point to another <gi>align</gi>
element logically preceding this
one.</valDesc>
<valDesc>Values should be Library of Congress subject
headings.</valDesc>
<valDesc>A bookseller's surname,
taken from the list in <title>Pollard and Redgrave</title>
</valDesc>

As noted above, the datatype element constrains the possible values for an attribute. The valDesc element can be used to describe further constraints. For example, to specify that an attribute age can take positive integer values less than 100, the datatype data.numeric might be used in combination with a valDesc such as ‘values must be positive integers less than 100’.

More usually, however, where constraints on values are explicitly enumerated, the valList element is used, as in the following example:
<valList type="closed">
 <valItem ident="req">
  <gloss>required</gloss>
 </valItem>
 <valItem ident="mwa">
  <gloss>mandatory when applicable</gloss>
 </valItem>
 <valItem ident="rec">
  <gloss>recommended</gloss>
 </valItem>
 <valItem ident="rwa">
  <gloss>recommended when applicable</gloss>
 </valItem>
 <valItem ident="opt">
  <gloss>optional</gloss>
 </valItem>
</valList>
Since this value list specifies that it is of type closed, only the values enumerated and glossed above are legal, and an ODD processor will typically enforce these constraints in the schema fragment generated.

The valList element is also used to provide illustrative examples of the kinds of values expected. In such cases the type attribute will have the value open and the datatype will usually be data.enumerated.

Note that the gloss element is needed to explain the significance of the identifier for an item only when this is not apparent, for example because it is abbreviated, as in the above example. It should not be used to provide a full description of the intended meaning (this is the function of the desc element), nor to comment on equivalent values in other schemes (this is the purpose of the equiv element), nor to provide alternative versions of the ident attribute value in other languages (this is the purpose of the altIdent element).

22.4.5.3 Examples
The following attList demonstrates some of the possibilities; for more detailed examples, consult the tagged version of the reference material in these Guidelines.
<attList>
 <attDef ident="type">
  <desc>describes the form of the list.</desc>
  <datatype>
   <rng:text/>
  </datatype>
  <defaultVal>simple</defaultVal>
  <valList type="semi">
   <valItem ident="ordered">
    <desc>list items are numbered or lettered. </desc>
   </valItem>
   <valItem ident="bulleted">
    <desc>list items are marked with a bullet or other
         typographic device. </desc>
   </valItem>
   <valItem ident="simple">
    <desc>list items are not numbered or bulleted.</desc>
   </valItem>
   <valItem ident="gloss">
    <desc>each list item glosses some term or
         concept, which is given by a label element preceding
         the list item.</desc>
   </valItem>
  </valList>
  <remarks>
   <p>The formal syntax of the element declarations allows
   <gi>label</gi> tags to be omitted from lists tagged <tag>list
         type="gloss"</tag>; this is however a semantic error.</p>
  </remarks>
 </attDef>
</attList>
In the following example, the org attribute is used to indicate that instances of the element concerned may bear either a bar attribute or a baz attribute, but not both. The bax attribute is always available:
<attList>
 <attDef ident="bax">
<!-- ... -->
 </attDef>
 <attList org="choice">
  <attDef ident="bar">
<!-- ... -->
  </attDef>
  <attDef ident="baz">
<!-- ... -->
  </attDef>
 </attList>
</attList>

22.4.6 Element Classes

The element classSpec is used to document either an attribute class or a ‘model class’, as defined in section 1.3 The TEI Class System. A corresponding classRef element may be used to select a specific named class from those available.
  • classSpec (class specification) contains reference information for a TEI element class; that is a group of elements which appear together in content models, or which share some common attribute, or both.
    typeindicates whether this is a model class or an attribute class
  • classRef/ points to the specification for an attribute or model class which is to be included in a schema
  • attList contains documentation for all the attributes associated with this element, as a series of attDef elements.
A model class specification does not list all of its members. Instead, its members declare that they belong to it by means of a classes element contained within the relevant elementSpec. This will contain a memberOf element for each class of which the relevant element is a member, supplying the name of the relevant class. For example, the elementSpec for the element hi contains the following:
<classes>
 <memberOf key="model.hiLike"/>
</classes>
This indicates that the hi element is a member of the class with identifier model.hiLike. The classSpec element that documents this class contains the following declarations:
<classSpec type="model" ident="model.hiLike">
 <desc>groups phrase-level elements related to highlighting that have
   no specific semantics </desc>
 <classes>
  <memberOf key="model.highlighted"/>
 </classes>
</classSpec>
which indicate that the class model.hiLike is actually a member (or subclass) of the class model.highlighted.

The function of a model class declaration is to provide another way of referring to a group of elements. It does not confer any other properties on the elements which constitute its membership.

The attribute type is used to distinguish between ‘model’ and ‘attribute’ classes. In the case of attribute classes, the attributes provided by membership in the class are documented by an attList element contained within the classSpec. In the case of model classes, no further information is needed to define the class beyond its description, its identifier, and optionally any classes of which it is a member.

When a model class is referenced in the content model of an element (i.e. in the content of an elementSpec), its meaning will depend on the name used to reference the class.

If the reference simply takes the form of the class name, it is interpreted to mean an alternated list of all the current members of the class. For example, suppose that the members of the class model.hiLike are elements hi, <it>, and <bo>. Then a content model such as
<content>
 <rng:zeroOrMore>
  <rng:ref name="model.hiLike"/>
 </rng:zeroOrMore>
</content>
would be equivalent to the explicit content model:
<content>
 <rng:zeroOrMore>
  <rng:choice>
   <rng:ref name="hi"/>
   <rng:ref name="it"/>
   <rng:ref name="bo"/>
  </rng:choice>
 </rng:zeroOrMore>
</content>
(or, to use RELAX NG compact syntax, (hi|it|bo)*). However, a content model referencing the class as model.hiLike_sequence would be equivalent to the following explicit content model:
<content>
 <rng:zeroOrMore>
  <rng:ref name="hi"/>
  <rng:ref name="it"/>
  <rng:ref name="bo"/>
 </rng:zeroOrMore>
</content>
(or, in RELAX NG compact syntax, (hi,it,bo)*.
The following suffixes, appended with an underscore, can be given to a class name when it is referenced in a content model:
alternation
members of the class are alternatives
sequence
members of the class are to be provided in sequence
sequenceOptional
members of the class may be provided, in sequence, but are optional
sequenceOptionalRepeatable
members of the class may be provided one or more times, in sequence, but are optional.
sequenceRepeatable
members of the class must be provided one or more times, in sequence
Thus a reference to model.hiLike_sequenceOptional in a content model would be equivalent to:
<rng:zeroOrMore>
 <rng:optional>
  <rng:ref name="hi"/>
 </rng:optional>
 <rng:optional>
  <rng:ref name="it"/>
 </rng:optional>
 <rng:optional>
  <rng:ref name="bo"/>
 </rng:optional>
</rng:zeroOrMore>
A reference to model.hiLike_sequenceRepeatable would however be equivalent to:
<rng:zeroOrMore>
 <rng:oneOrMore>
  <rng:ref name="hi"/>
 </rng:oneOrMore>
 <rng:oneOrMore>
  <rng:ref name="it"/>
 </rng:oneOrMore>
 <rng:oneOrMore>
  <rng:ref name="bo"/>
 </rng:oneOrMore>
</rng:zeroOrMore>
and a reference to model.hiLike_sequenceOptionalRepeatable would be equivalent to:
<rng:zeroOrMore>
 <rng:zeroOrMore>
  <rng:ref name="hi"/>
 </rng:zeroOrMore>
 <rng:zeroOrMore>
  <rng:ref name="it"/>
 </rng:zeroOrMore>
 <rng:zeroOrMore>
  <rng:ref name="bo"/>
 </rng:zeroOrMore>
</rng:zeroOrMore>

The ‘sequence’ in which members of a class appear in a content model when one of the sequence options is used is that in which the elements are declared.

In principal, all these possibilities are available to any element making reference to any class. The classSpec element defining the class may however limit the possibilities by means of its generate attribute, which can be used to say that this particular model may only be referenced in a content model with the suffixes it specifies. For example, if the classSpec for model.hiLike took the form <classSpec ident="model.hiLike" generate="sequence sequenceOptional"> then a content model referring to (say) model.hiLike_sequenceRepeatable would be regarded as invalid by an ODD processor.

An attribute class (a classSpec of type atts) contains an attList element which lists the attributes that all the members of that class inherit from it. For example, the class att.interpLike defines a small set of attributes common to all elements which are members of that class: those attributes are listed by the attList element contained by the classSpec for att.interpLike. When processing the documentation elements for elements which are members of that class, an ODD processor is required to extend the attList (or equivalent) for such elements to include any attributes defined by the classSpec elements concerned. There is a single global attribute class, att.global, to which some modules contribute additional attributes when they are included in a schema.

22.4.7 Pattern Documentation

The macroSpec element is used to declare and document predefined strings or patterns not otherwise documented by the elements described in this chapter. A corresponding macroRef element may be used to select a specific named pattern from those available. Patterns are used as a shorthand chiefly to describe common content models and datatypes, but may be used for any purpose. The following elements are used to represent patterns:
  • macroSpec (macro specification) documents the function and implementation of a pattern.
    typeindicates which type of entity should be generated, when an ODD processor is generating a module using XML DTD syntax.
  • macroRef/ points to the specification for some pattern which is to be included in a schema
    keythe identifier used for the required pattern within the source indicated.
  • remarks contains any commentary or discussion about the usage of an element, attribute, class, or entity not otherwise documented within the containing element.
  • stringVal contains the intended expansion for the entity documented by a macroSpec element, enclosed by quotation marks.

22.5 Building a Schema

The specification elements, and some of their children, are all members of the att.identified class, from which they inherit the following attributes:
  • att.identified provides the identifying attribute for elements which can be subsequently referenced by means of a key attribute.
    identSupplies the identifier by which this element may be referenced.
    predeclareSays whether this object should be predeclared in the tei infrastructure module.
    moduleSupplies a name for the module in which this object is to be declared.
This attribute class is a subclass of the att.combinable class from which it (and some other elements) inherits the following attribute:
  • att.combinable provides attributes indicating how multiple references to the same object in a schema should be combined
    modespecifies the effect of this declaration on its parent object.

These attributes are used by an ODD processor to determine how declarations are to be combined to form a schema or DTD, as further discussed in this section.

As noted above, a TEI schema is defined by a schemaSpec element containing an arbitrary mixture of explicit declarations for objects (i.e. elements, classes, patterns, or macro specifications) and references to other objects containing such declarations (i.e. references to specification groups, or to modules). A major purpose of this mechanism is to simplify the process of defining user customizations, by providing a formal method for the user to combine new declarations with existing ones, or to modify particular parts of existing declarations.

In the simplest case, a user-defined schema might simply combine all the declarations from two nominated modules:
<schemaSpec ident="example">
 <moduleRef key="core"/>
 <moduleRef key="linking"/>
</schemaSpec>
An ODD processor, given such a document, should combine the declarations which belong to the named modules, and deliver the result as a schema of the requested type. It may also generate documentation for the elements declared by those modules. No source is specified for the modules, and the schema will therefore combine the declarations found in the most recent release version of the TEI Guidelines known to the ODD processor in use.

The value specified for the source attribute, when it is supplied as a URL, specifies any convenient location from which the relevant ODD files may be obtained. For the current release of the TEI Guidelines, a URL in the form http://www.tei-c.org/Vault/P5/x.y.z/xml/tei/odd/p5subset.xml may be used, where x.y.z represents the version number, e.g. 1.3.0. Alternatively, if the ODD files are locally installed, it may be more convenient to supply a value such as ../ODDs/p5subset.xml".

The value for the source attribute may any form of URI. A set of TEI-conformant specifications in a form directly usable by an ODD processor must be available at the location indicated. When no source value is supplied, an ODD processor may either raise an error or assume that the location of the current release of the TEI Guidelines is intended.

If the source is specified in the form of a private URI, the form recommended is tei:x.y.z, where x.y.z indicates the version number, e.g. tei:1.2.1 for release 1.2.1 of the current TEI Guidelines. When such a URI is used, it will usually be necessary to translate it before such a file can be used in blind interchange.

The effect of a moduleRef element is to include in the schema all declarations provided by that module. This may be modified by means of the attributes include and except which allow the encoder to supply an explicit lists of elements from the stated module which are to be included or excluded respectively. For example:
<schemaSpec ident="example">
 <moduleRef key="core" except="add del orig reg"/>
 <moduleRef key="linking" include="ptrGroup link"/>
</schemaSpec>
The schema specified here will include all the elements supplied by the core module except for add, del, orig, and reg. It will also include only the <ptrGroup> and link elements from the linking module.
Alternatively, the element elementRef may be used to indicate explicitly which elements are to be included in a schema. The same effect as the preceding example might thus be achieved by the following:
<schemaSpec ident="example">
 <moduleRef key="core" except="add del orig reg"/>
 <elementRef key="ptrGroup"/>
 <elementRef key="link"/>
</schemaSpec>
Note that in this last case, there is no need to specify the name of the module from which the two element declarations are to be found; in the TEI scheme, element names are unique across all modules. The module is simply a convenient way of grouping together a number of related declarations.
A schema may also include declarations for new elements, as in the following example:
<schemaSpec ident="example">
 <moduleRef key="teiheader"/>
 <moduleRef key="verse"/>
 <elementSpec ident="soundClip">
  <classes>
   <memberOf key="model.pPart.data"/>
  </classes>
 </elementSpec>
</schemaSpec>
A declaration for the element <soundClip>, which is not defined in the TEI scheme, will be added to the output schema. This element will also be added to the existing TEI class model.pPart.data, and will thus be available in TEI conformant documents.
A schema might also include re-declarations of existing elements, as in the following example:
<schemaSpec ident="example">
 <moduleRef key="teiheader"/>
 <moduleRef key="teistructure"/>
 <elementSpec ident="head" mode="change">
  <content>
   <rng:ref name="macro.xtext"/>
  </content>
 </elementSpec>
</schemaSpec>
The effect of this is to redefine the content model for the element head as plain text, by over-riding the content child of the selected elementSpec. The attribute specification mode="change" has the effect of over-riding only those children elements of the elementSpec which appear both in the original specification and in the new specification supplied above: content in this example. Note that if the value for mode were replace, the effect would be to replace all children elements of the original specification with the the children elements of the new specification, and thus (in this example) to delete all of them except content.
A schema may not contain more than two declarations for any given component. The value of the mode attribute is used to determine exactly how the second declaration (and its constituents) should be combined with the first. The following table summarizes how a processor should resolve duplicate declarations; the term identifiable refers to those elements which can have a mode attribute:
mode value existing declaration effect
add no add new declaration to schema; process its children in add mode
add yes raise error
replace no raise error
replace yes retain existing declaration; process new children in replace mode; ignore existing children
change no raise error
change yes process identifiable children according to their modes; process unidentifiable children in replace mode; retain existing children where no replacement or change is provided
delete no raise error
delete yes ignore existing declaration and its children

22.6 Combining TEI and Non-TEI Modules

In the simplest case, all that is needed to include a non-TEI module in a schema is to reference its RELAX NG source using the url attribute on moduleRef. The following specification, for example, creates a schema in which declarations from the non-TEI module svg11.rng (defining Standard Vector Graphics) are included. To avoid any risk of name clashes, the schema specifies that all TEI patterns generated should be prefixed by the string "TEI_".
<schemaSpec prefix="TEI_" ident="testsvg" start="TEI svg">
 <moduleRef key="header"/>
 <moduleRef key="core"/>
 <moduleRef key="tei"/>
 <moduleRef key="textstructure"/>
 <moduleRef url="svg11.rng"/>
</schemaSpec>
This specification generates a single schema which might be used to validate either a TEI document (with the root element TEI), or an SVG document (with a root element <svg:svg>), but would not validate a TEI document containing <svg:svg> or other elements from the SVG language. For that to be possible, the <svg:svg> element must become a member of a TEI model class (1.3 The TEI Class System), so that it may be referenced by other TEI elements. To achieve this, we modify the last moduleRef in the above example as follows:
<moduleRef url="svg11.rng">
 <content>
  <rng:define name="TEI_model.graphicLike" combine="choice">
   <rng:ref name="svg"/>
  </rng:define>
 </content>
</moduleRef>

This states that when the declarations from the svg11.rng module are combined with those from the other modules, the declaration for the model class model.graphicLike in the TEI module should be extended to include the element <svg:svg> as an alternative. This has the effect that elements in the TEI scheme which define their content model in terms of that element class (notably figure) can now include it. A RELAX NG schema generated from such a specification can be used to validate documents in which the TEI figure element contains any valid SVG representation of a graphic, embedded within an <svg:svg> element.

22.7 Module for Documention Elements

The module described in this chapter makes available the following components:
Module tagdocs: Documentation of TEI modules
The selection and combination of modules to form a TEI schema is described in 1.2 Defining a TEI Schema.

The elements described in this chapter are all members of one of three classes: model.oddDecl, model.oddRef, or model.phrase.xml, with the exceptions of schemaSpec (a member of model.divPart) and both eg and egXML (members of model.common and model.egLike). All of these classes are declared along with the other general TEI classes, in the basic structure module documented in 1 The TEI Infrastructure.

In addition, some elements are members of the att.identified class, which is documented in 22.5 Building a Schema above, and make use of the macro.schemaPattern pattern, which is documented in 22.4.4 Element Specifications above.

↑ Contents « 21 Certainty, Precision, and Responsibility » 23 Using the TEI

Notes
79.
ODD is short for ‘One Document Does it all’, and was the name invented by the original TEI Editors for the predecessor of the system currently used for this purpose. See further Burnard and Sperberg-McQueen (1995) and Burnard and Rahtz (2004).
↵
80.
The distinction between base and additional tagsets in earlier versions of the TEI scheme has not been carried forward into P5.
↵

[English] [Deutsch] [Español] [Italiano] [Français] [日本語] [한국어] [中文]

TEI Consortium | Feedback
Copyright TEI Consortium 2010 Licensed under the GPL. Copying and redistribution is permitted and encouraged.
Version 1.9.1. Last updated on March 5th 2011.This page generated on 2011-04-29T09:39:14Z