AW: [xml] IDC bug?

["Kasimier Buchcik" <k buchcik 4commerce de>]

Hi, 

> -----Ursprüngliche Nachricht-----
> Von: xml-bounces gnome org [mailto:xml-bounces gnome org] Im 
> Auftrag von Frans Englich
> Gesendet: Mittwoch, 5. April 2006 17:24
> An: xml gnome org
> Betreff: [xml] IDC bug?
>
>
>
> Either I'm not using IDCs correctly(which may very well be 
> the case), or 
> there's a bug in xmllint. Consider this schema:
>
> <xs:schema xmlns:xs="http://www.w3.org/2001/XMLSchema"; 
> elementFormDefault="qualified">
>     <xs:element name="test-group">
>         <xs:complexType>
>             <xs:sequence>
>                 <xs:element name="test-case" minOccurs="0" 
> maxOccurs="unbounded">
>                     <xs:unique name="idc1">
>                         <xs:selector xpath="."/>
>                         <xs:field xpath="@name"/>
>                     </xs:unique>
>                 </xs:element>
>                 <xs:element ref="test-group" minOccurs="0" 
> maxOccurs="unbounded" />
>             </xs:sequence>
>             <xs:attribute name="name" type="xs:string" 
> use="required" />
>         </xs:complexType>
>         <xs:unique name="idc2">
>             <xs:selector xpath="."/>
>             <xs:field xpath="@name"/>
>         </xs:unique>
>     </xs:element>
> </xs:schema>
>
>
> The following two documents validates.
>
>
> <test-group name="group1">
>         <test-group name="group1"/>
> </test-group>
>
> <test-group name="group1">
>         <test-case name="test-case1"/>
>         <test-case name="test-case1"/>
> </test-group>
>
>
> But as I see it, they shouldn't. Or should they?
>
> This was produced with libxml2 version 20623.

The two documents you provided are valid (tested with
Saxon, Xerces and XSV).
I think a good way of understanding what's going on with
IDC key/unique is to imagine that a local list of values
(selected by the <xs:selector> and <xs:field>s) is built
on the element-node for which an IDC key/unique is declared;
after the validation is finished with such an element-node,
the local list disappears.

So for the <test-case>s we get:
<test-case name="test-case1"/> --> list with 1 entry: "test-case1"
<test-case name="test-case1"/> --> list with 1 entry: "test-case1"

Both lists are seperate from each other, so there's no error.

I think what you want, in order to disallow duplicate names for
the test-cases, is to declare the xs:unique on the test-group and
select all names of all test-case children:

<xs:unique name="idc2">
  <xs:selector xpath="./test-case"/>
  <xs:field xpath="@name"/>
</xs:unique>

This will generate a list on the test-group:

<test-group name="group1">       --> list with 2 entries: 
  <test-case name="test-case1"/>     "test-case1"           
  <test-case name="test-case1"/>     "test-case1"  
</test-group>

And this will - with some processors - produce an error.
"With some processors", since there's a gray area in the spec:
Libxml2 won't report an error, since there's an other issue within
the schema. The element "test-case" doesn't specify a type, so
the type defaults to xs:anyType; an element of this type can
have any content and any attributes (lax attribute wildcard).
The spec is silent about the case when IDCs resolve to
non-declared elements/attributes.

In xmlschemas.c, xmlSchemaVAttributesComplex(), it reads:
"Only "assessed" attribute information items will be visible to
 IDCs. I.e. not "lax" (without declaration) and "skip" wild attributes."

XSV 2.10-1 won't report an error as well.

On the other side Saxon 8.7 and Xerces-J 2.5.1 *do* report an error.

I think we should go the Saxon/Xerces way here. It would be great
if you could bugzilla this.

Possible workarounds:
1) declare the "name" attribute locally on "test-case"
2) declare a global attribute: 
  <xs:attribute name="name" type="xs:string"/>
  The lax attribute wildcard of the xs:anyType will then pick
  this global attribute declaration for the "name" attribute
  of "test-case" and the IDCs will resolve to that attribute.

Your schema could look like the following. It additionally
has a "tests" element, acting as a container for test-groups
and test-cases, plus defining IDCs. Due to "//" in
the selector ".//test-group" it will assure unique names
for all test-groups in the document.
   
Schema "idc.xsd":

<xs:schema xmlns:xs="http://www.w3.org/2001/XMLSchema";>

  <xs:element name="test-case">
    <xs:complexType>
      <xs:attribute name="name" type="xs:string" use="required"/>
    </xs:complexType>
  </xs:element>

  <xs:element name="test-group">
    <xs:complexType>
      <xs:sequence>
        <xs:element ref="test-case" minOccurs="0" maxOccurs="unbounded"/>
        <xs:element ref="test-group" minOccurs="0" maxOccurs="unbounded"/>
      </xs:sequence>
      <xs:attribute name="name" type="xs:string" use="required"/>
    </xs:complexType>   
  </xs:element>

  <xs:element name="tests">
    <xs:complexType>
      <xs:sequence>        
          <xs:element ref="test-case" minOccurs="0" maxOccurs="unbounded"/>
        <xs:element ref="test-group" minOccurs="0" maxOccurs="unbounded"/>
      </xs:sequence>
    </xs:complexType>

    <xs:unique name="unique-test-group">
      <xs:selector xpath=".//test-group"/>
      <xs:field xpath="@name"/>
    </xs:unique>

    <xs:unique name="unique-test-case">
      <xs:selector xpath=".//test-case"/>
      <xs:field xpath="@name"/>
    </xs:unique>

  </xs:element>

</xs:schema>

Instance "idc.xml":

<tests
        xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance";
        xsi:noNamespaceSchemaLocation="idc.xsd">
        
        <test-case name="test-case1"/>
        <test-case name="test-case1"/>
</tests>

> xmllint --noout --schema idc.xsd idc.xml
idc.xml:7: element test-case: Schemas validity error :
Element 'test-case': Duplicate key-sequence ['test-case1'] in
unique identity-constraint 'unique-test-case'.
idc.xml fails to validate

Regards,

Kasimier