STEP 1: Migration from XML DTD (without parameter entities)

$Id: step1e.html 1.8 2000/03/13 12:21:11 murata Exp $

text by MURATA Makoto

html by NAMBA Ryosuke

STEP 1 covers basic features, which allows easy migration from DTD. A DTD-to-RELAX converter (dtd2relax) uses these features only.

1. An example module

To give an idea of RELAX, we recapture a DTD as a RELAX module.

A DTD is shown below. The number attribute of title elements should be integers, but DTD cannot represent this constraint.

<!ELEMENT doc   (title, para*)>

<!ELEMENT para  (#PCDATA | em)*>

<!ELEMENT title (#PCDATA | em)*>

<!ELEMENT em    (#PCDATA)>

<!ATTLIST para
  role NMTOKEN #IMPLIED
>

<!ATTLIST title  
  role   NMTOKEN #IMPLIED
  number CDATA   #IMPLIED
>

Next, we show a RELAX module. The number attribute is specified as an integer.

<module
      moduleVersion="1.2"
      relaxCoreVersion="1.0"
      targetNamespace=""
      xmlns="http://www.xml.gr.jp/2000/relaxCore">

  <interface>
    <export labels="doc"/>
  </interface>

  <elementRule pred="doc">
    <sequence>
      <ref label="title"/>
      <ref label="para" occurs="*"/>
    </sequence>
  </elementRule>
  
  <elementRule pred="para">
    <mixed>
      <ref label="em" occurs="*"/>
    </mixed>
  </elementRule>
  
  <elementRule pred="title">
    <mixed>
      <ref label="em" occurs="*"/>
    </mixed>
  </elementRule>

  <elementRule pred="em" type="string"/>
  
  <tag name="doc"/>

  <tag name="para">
    <attribute name="role" type="NMTOKEN"/>
  </tag>
  
  <tag name="title">
    <attribute name="role" type="NMTOKEN"/>
    <attribute name="number" required="true" type="integer"/>
  </tag>

  <tag name="em"/>

</module>

Subsequent sections explain syntactical constructs appeared in this example.

2. The `module` element

A RELAX grammar is a combination of modules. If the number of namespaces is one and the grammar is not so large, a module provides a RELAX grammar. A module is represented by a module element.

<module
      moduleVersion="1.2"
      relaxCoreVersion="1.0"
      targetNamespace=""
      xmlns="http://www.xml.gr.jp/2000/relaxCore">
  ...
</module>

The moduleVersion attribute shows the version of this module. In this example, it is "1.2".

The relaxCoreVersion attribute shows the version of RELAX Core. At present, it is always "1.0".

The targetNamespace attribute shows the namespace which this module is concerned with. In this example, it is "".

The namespace name for RELAX Core is "http://www.xml.gr.jp/2000/relaxCore".

3. The `interface` element

A module element begins with an interface element. There is at most one interface element in a single module.

<module
      moduleVersion="1.2"
      relaxCoreVersion="1.0"
      targetNamespace=""
      xmlns="http://www.xml.gr.jp/2000/relaxCore">

  <interface>
    ...
  </interface>
  ...
</module>

3.1 The `export` element

An interface element contains export element(s).

<export labels="foo bar"/>

The labels attribute of export elements specifies element types that may become the root. More than one export may appear in an interface element.

Each of the following examples allows element type foo and bar as the root.

<interface>
  <export labels="foo"/>
  <export labels="bar"/>
</interface>

<interface>
  <export labels="foo bar"/>
</interface>

4. Element type declarations

Element type declarations (<!ELEMENT ...>) of XML are represented by elementRule elements. The pred attribute of elementRule specifies an element type name. More than one elementRule may follow the interface element.

<elementRule pred="element-type-name">
  ...hedge model...
</elementRule>

An elementRule element has an hedge model. A hedge is a sequence of elements (and their decendands) as well as character data. A hedge model is a constaint on permissible hedges.

A hedge model is either an element hedge model, datatype reference, or mixed hedge model.

4.1 Element hedge model

Element hedge models are represented by empty, none, ref, choice, sequence elements and the occurs attribute. An element hedge model represents permissible sequences of child elements, which are possibly intervened by whitespace characters.

(1) The `empty` element

empty represents the empty sequence.

Consider an elementRule as below:

<elementRule pred="foo">
  <empty/>
</elementRule>

This elementRule implies that the content of a foo element is the empty sequence. A foo element can be a start tag followed by an end tag, or an empty-element tag.

<foo/>

<foo></foo>

Unlike EMPTY of XML, whitespace characters may intervene between start and end tags.

<foo>  </foo>

empty can be used within choice and sequence (see (3) and (4)). The motivation behind this extension will become clear in STEP 2. If you need exactly the same feature as EMTPY of XML, use the emptyString datatype (shown in STEP 3).

From now on, we assume that foo, foo1, foo2 are declared by elementRules whose hedge models are empty.

(2) The `ref` element

ref references to an element type. For example, <ref label="foo"/> references to an element type foo.

Consider an elementRule as below:

<elementRule pred="bar">
  <ref label="foo"/>
</elementRule>

This elementRule implies that the content of a bar element is an foo element. For example, the next bar element is legitimate against this elementRule.

<bar><foo/></bar>

Whitespace may appear before and after the foo element.

<bar>
  <foo/>
</bar>

ref can have the occurs attribute. Permissible values are "*", "+", and "?" , which indicate "zero or more", "one or more", and "zero or one times", respectively.

An example of "?" as the occurs attribute is as below:

<elementRule pred="bar">
  <ref label="foo" occurs="?"/>
</elementRule>

This elementRule implies that the content of a bar element is either a foo or empty.

<bar><foo/></bar>

<bar></bar>

Whitespace characters may appear before and after the foo element. Even when this bar is empty, it may have whitespace characters.

<bar>
  <foo/>
</bar>

<bar>
</bar>

(3) The `choice` element

choice indicates a choice of the specified hedge models ("|" of XML 1.0). Subordinate elements of choice elements are element hedge models. choice can also have the occurs attribute.

An example of elementRule containing choice is shown below:

<elementRule pred="bar">
  <choice occurs="+">
    <ref label="foo1"/>
    <ref label="foo2"/>
  </choice>
</elementRule>

This elementRule indicates that the content of a bar element is one or more occurrences of either foo1 or foo2 elements.

<bar><foo2/></bar>

<bar>
  <foo2/>
</bar>

<bar>
  <foo1/>
  <foo2/>
  <foo1/>
</bar>

(4) The `sequence` element

sequence is a sequence of the specified hedge models. ("," of XML 1.0). Subordinate elements of sequence are element hedge models. sequence can also have the occurs attribute.

An example of elementRule containing sequence is shown below:

<elementRule pred="bar">
  <sequence occurs="?">
    <ref label="foo1"/>
    <ref label="foo2"/>
  </sequence>
</elementRule>

This elementRule implies that the content of a bar element is either a sequence of a foo1 element and a foo2 element, or empty.

<bar><foo1/><foo2/></bar>

<bar>
  <foo1/>
  <foo2/></bar>

<bar/>

<bar></bar>

<bar>
  </bar>

(5)The `none` element

none is an element hedge model, which does not match anything. none is unique to RELAX.

<elementRule pred="bar">
  <none/>
</elementRule>

This elementRule implies that nothing is permitted as the content of bar elements. The motivation behind none will become clear in STEP 2.

4.2 Datatype reference

The type attribute of elementRule allows a content model that references to a datatype. Character strings in an document are compared with the specified datatype. Permissible datatypes are built-in datatypes of XML Schema Part 2, or datatypes unique to RELAX. Details of datatypes will be covered by STEP 3.

An example of elementRule containing type is shown below:

<elementRule pred="bar" type="integer"/>

This elementRule indicates that the content of a bar element is a character string representing an integer.

<bar>10</bar>

Whitespace characters may not occur before or after the integer. For example, the following is not permitted.

<bar>
  10
</bar>

4.3 Mixed hedge model

mixed significantly extends mixed content models (#PCDATA|a|b|...|z)* of XML.

A mixed element wraps an element hedge model. Recall that an element hedge model allows whitespace characters to intervene between elements. By wrapping it with mixed, any character is allowed to intervene.

As an example, consider elementRule as below:

<elementRule pred="bar">
  <mixed>
    <ref label="foo"/>
  </mixed>
</elementRule>

Element <foo/> matches ref in the mixed element. Thus, the following example is permitted by this contentRule.

<bar>Murata<foo/>Makoto</bar>

As shown in the following example, CDATA sections and character references may appear.

<bar><![CDATA[Murata]]><foo/>Mako&#x74;&#x6F;</bar>

(#PCDATA | foo1| foo2)* of XML can be captured as below:

<elementRule pred="bar">
  <mixed>
    <choice occurs="*">
      <ref label="foo1"/>
      <ref label="foo2"/>
    </choice>
  </mixed>
</elementRule>

There are two ways to capture a content model (#PCDATA). One is to reference to the datatype string by the type attribute. The other is to make an element hedge model that matches the empty sequence only and wrap it with mixed. An example is as below:

<elementRule pred="bar" type="string"/>

<elementRule pred="bar">
  <mixed>
    <empty/>
  </mixed>
</elementRule>

5.Attribute-list declarations

Attribute-list declarations (<!ATTLIST ...>) of XML are captured by tag elements.

<tag name="element-type-name">
  ...list of attribute declarations ...
</tag>

tag can have attribute elements as subordinates.

<tag name="element-type-name">
  <attribute ... />
  <attribute ... />
</tag>

attribute declares an attribute. An example of attribute is shown below:

<attribute name="age" required="true" type="integer"/>

The value of the name attribute is the name of the declared attribute. In this example, it is age.

If the value of the required attribute is true, the attribute being declared is mandatory. If required is not specified, it is optional. Since required is specified in this example, the age attribute is mandatory.

The type attribute specifies a datatype name. If type is not specified, a datatype string (which allows any string) is assumed.

Consider an example of tag which contains an attribute element only.

<tag name="bar">
  <attribute name="age" required="true" type="integer"/>
</tag>

The following start tag is permitted by this tag.

<bar age="39">

The following two start tags are not permitted. In the first example, the age attribute is omitted. In the second example, the value of age is not an integer.

<bar>

<bar age="bu huo">
<!-- "bu huo" means forty years in Chinese.  In Japan, 
     it is pronounced as "FUWAKU". -->

In DTD, you do not have to write an attribute-list declaration if an element type does not have any attributes. In RELAX, you must write an empty tag element even if there are no attributes. For example, if an element type bar does not have any attributes, you have to write a tag element as below:

<tag name="bar"/>

6. Summary

If you have finished reading this STEP, you can immediately start to use RELAX. If you do not need further features, you do not have to read other STEPs. Enjoy and RELAX!

mura034@attglobal.net

STEP 1: Migration from XML DTD (without parameter entities)

1. An example module

2. The module element

3. The interface element

3.1 The export element