XML Schema Design: restrict the occurrence of an element according to its position in the document

Question

Hello I came across a problem regarding to restrict the occurrence of an element according to its position in the document. Actually "position" might not be a proper term but I couldn't think of a better way to summarise the problem. Anyway let me explain a bit about the business logic. The schema will be designed for a play script. In a play, one actor can play multiple roles. A play contains many scenes, in a scene, there are several occurrences of stage-direction(Enter/Exit), between two stage-directions, actors give speeches. A sample xml instance would look like this:

<PLAY>
    <CAST>
        <ROLE>
            <ACTOR id="1">Alex</ACTOR>
            <PERSONA>Superman</PERSONA>
        </ROLE>
        <ROLE>
            <ACTOR id="1">Alex</ACTOR>
            <PERSONA>Batman</PERSONA>
        </ROLE>
        <ROLE>
            <ACTOR id="2">John</ACTOR>
            <PERSONA>Hulk</PERSONA>
        </ROLE>
    </CAST>
    <TITLE>Lego Movie</TITLE>
    <SCENE>
        <TITLE>SCENE I</TITLE>
        <STAGEDIR>Enter Superman, Hulk</STAGEDIR>
        <SPEECH>
            <SPEAKER actorID="1">Superman</SPEAKER>
            <LINE>Hahaha</LINE>
        </SPEECH>
        <SPEECH>
            <SPEAKER actorID="2">Hulk</SPEAKER>
            <LINE>Hahaha</LINE>
        </SPEECH>
        <STAGEDIR>Exit Superman, Enter Batman</STAGEDIR>
        <SPEECH>
            <SPEAKER actorID="1">Batman</SPEAKER>
            <LINE>Yo</LINE>
        </SPEECH>
        <SPEECH>
            <SPEAKER actorID="2">Hulk</SPEAKER>
            <LINE>Yo</LINE>
        </SPEECH>
    </SCENE>
</PLAY>

The restriction required here is that if an actor is playing multiple roles, it is not allowed for two persona to be talking to each other on the stage at the same time (if an actor is playing multiple roles, an actor's persona cannot appear on the stage at the same time). For example Alex is playing both Superman and Batman, they cannot appear on the stage at the same time since they are played by a same person. Currently the schema looks like:

<xs:complexType name="PLAYTYPE">
    <xs:sequence>
        <xs:element ref="CAST"/>
        <xs:element ref="TITLE"/>
        <xs:element ref="SCENE"/>
    </xs:sequence>
</xs:complexType>

<xs:complexType name="CASTTYPE">
    <xs:sequence>
        <xs:element ref="ROLE" maxOccurs="unbounded"/>
    </xs:sequence>
</xs:complexType>

<xs:complexType name="ROLETYPE">
    <xs:sequence>
        <xs:element ref="ACTOR"/>
        <xs:element ref="PERSONA"/>
    </xs:sequence>
</xs:complexType>

<xs:complexType name="SCENETYPE">
    <xs:sequence>
        <xs:element ref="TITLE"/>
        <xs:element ref="STAGEDIR"/>
        <xs:element ref="SPEECH" maxOccurs="unbounded"/>
    </xs:sequence>
</xs:complexType>

<xs:complexType name="ACTORTYPE">
    <xs:sequence>
        <xs:element ref="NAME"/>
    </xs:sequence>
    <xs:attributeGroup ref="attlist.ACTOR"/>
</xs:complexType>

<xs:complexType name="STAGEDIRTYPE" mixed="true">
    <xs:sequence>
        <xs:element minOccurs="1" maxOccurs="unbounded" ref="PERSONA"/>
    </xs:sequence>
    <xs:attributeGroup ref="attlist.SPEAKER"/>
</xs:complexType>

<xs:complexType name="SPEECHTYPE">
    <xs:sequence>
        <xs:element ref="SPEAKER"/>
        <xs:element ref="LINE"/>
    </xs:sequence>
    <xs:attributeGroup ref="attlist.ACTOR"/>
</xs:complexType>

<xs:complexType name="SPEAKERTYPE" mixed="true">
    <xs:attributeGroup ref="attlist.SPEAKER"/>
</xs:complexType>

<xs:element name="PLAY" type="PLAYTYPE"/>
<xs:element name="CAST" type="CASTTYPE"/>
<xs:element name="ROLE" type="ROLETYPE"/>
<xs:element name="ACTOR" type="ACTORTYPE"/>
<xs:element name="SCENE" type="SCENETYPE"/>
<xs:element name="STAGEDIR" type="STAGEDIRTYPE"/>
<xs:element name="SPEAKER" type="SPEAKERTYPE"/>
<xs:element name="TITLE" type="xs:string"/>
<xs:element name="PERSONA" type="xs:string"/>
<xs:element name="NAME" type="xs:string"/>
<xs:element name="LINE" type="xs:string"/>

<xs:attributeGroup name="attlist.ACTOR">
    <xs:attribute name="id" use="required"/>
</xs:attributeGroup>

<xs:attributeGroup name="attlist.SPEAKER">
    <xs:attribute name="actorID" use="required"/>
</xs:attributeGroup>

So how to achieve this kind of restriction in the schema?

Answer 1

To enforce the rule that for every one <SCENE> any <SPEAKER> elements with the same actorID attributes must contain the same text, you can declare your SCENETYPE with a XPath 2.0 assertion, as shown below:

<xs:complexType name="SCENETYPE">
    <xs:sequence>
        <xs:element ref="TITLE"/>
        <xs:element ref="STAGEDIR"/>
        <xs:element ref="SPEECH" maxOccurs="unbounded"/>
    </xs:sequence>
    <xs:assert test="every $speakerId in SPEECH/SPEAKER/@actorID
                       satisfies
                        (every $speaker in SPEECH/SPEAKER[@actorID=$speakerId]
                          satisfies 
                            (SPEECH/SPEAKER[@actorID=$speakerId])[1] = $speaker)"/>
</xs:complexType>

This will work in XSD 1.1 , which supports <xs:assert> . You might also be able to do something like this using Schematron extensions for XSD 1.0.

With this restriction, this block will validate, since there is a consistent mapping: 1 = Superman , 2 = Hulk :

<SCENE>
    ...
    <SPEECH id="1">
        <SPEAKER actorID="1">Superman</SPEAKER>
        <LINE>Hahaha</LINE>
    </SPEECH>
    <SPEECH id="2">
        <SPEAKER actorID="2">Hulk</SPEAKER>
        <LINE>Hahaha</LINE>
    </SPEECH>
    <SPEECH id="3">
        <SPEAKER actorID="1">Superman</SPEAKER>
        <LINE>Yo</LINE>
    </SPEECH>
    <SPEECH id="4">
        <SPEAKER actorID="2">Hulk</SPEAKER>
        <LINE>Yo</LINE>
    </SPEECH>
</SCENE>

But this will fail validation, since 1 contains Superman and also Batman :

<SCENE>
    <TITLE>SCENE I</TITLE>
    <STAGEDIR actorID="0"><PERSONA>Superman</PERSONA><PERSONA>Hulk</PERSONA><PERSONA>Batman</PERSONA></STAGEDIR>
    <SPEECH id="1">
        <SPEAKER actorID="1">Superman</SPEAKER>
        <LINE>Hahaha</LINE>
    </SPEECH>
    <SPEECH id="2">
        <SPEAKER actorID="2">Hulk</SPEAKER>
        <LINE>Hahaha</LINE>
    </SPEECH>
    <SPEECH id="3">
        <SPEAKER actorID="1">Batman</SPEAKER>
        <LINE>Yo</LINE>
    </SPEECH>
    <SPEECH id="4">
        <SPEAKER actorID="2">Hulk</SPEAKER>
        <LINE>Yo</LINE>
    </SPEECH>
</SCENE>

If you are still designing this schema, a better option would be to check for this consistency in your <STAGEDIR> element, since it would be possible to have the same actor being two characters in the same scene if it exits and returns in a different moment. Since a <PERSONA> can be played by different actors, you might want to have some way to associate a currentActorId for example, and when you list the <PERSONA> elements in the <STAGEDIR> you can check for consistency there, instead of checking the entire scene.