XML Schema quiz: 3 elements and at least one must be present

["Costello, Roger L." <costello@mitre.org>]

Hi Folks,

Consider this problem: you have 3 elements - A, B, C - and you need at
least one of them to be present in your XML instance document.

Here are the valid instances:

<A>...</A>
<B>...</B>
<C>...</C>

or

<A>...</A>
<B>...</B>

or

<A>...</A>
<C>...</C>

or

<B>...</B>
<C>...</C>

or

<A>...</A>

or

<B>...</B>

or

<C>...</C>

Question: will this XML Schema definition work:

<complexType name="ThreeElements">
    <choice>
        <sequence>
            <element name="A" type="string" />
            <element name="B" type="string" minOccurs="0" />
            <element name="C" type="string" minOccurs="0" />
        </sequence> 
        <sequence>
            <element name="A" type="string" minOccurs="0" />
            <element name="B" type="string" />
            <element name="C" type="string" minOccurs="0" />
        </sequence> 
        <sequence>
            <element name="A" type="string" minOccurs="0" />
            <element name="B" type="string" minOccurs="0" />
            <element name="C" type="string" />
        </sequence>
    </choice>
</complexType>

Notice that in the first choice A is required and B, C are optional.
In the second choice B is required and A, C are optional.  In the third
choice C is required and A,B are optional.

Scroll down for the answer ...














































Answer:

Let's play schema validator:  You are parsing the instance document and
encounter this element:

<A>...</A>

How shall you validate it?  In the above schema there are three
declarations for the <A> element.  The only way for you to know which
declaration to use is to "look ahead" in the instance document.  For
example, if there is no <B> element then you can eliminate the second
choice.

Requiring an XML Schema validator to "look ahead" is not allowed (it is
allowed with Relax NG).

The technical term for the above content model is: non-deterministic
content model.  XML Schema 1.0 does not allow non-deterministic content
models.

Thus, the above complexType definition is not valid.

/Roger