This page contains miscellaneous notes and explanation of midPoint schema concepts.
QNames, URIs and Other Animals
QNames are used for identification of many things in midPoint: properties (items), schema data types, resource attributes, relations, …
How to properly render input field for QName properties in GUI? If that QName is supposed to be an object type then we should render a dropdown with list of all object types (e.g. focusType element). If the field is a relation then it has quite a custom logic to present relations to choose from. If the field specified name of a new relation, we should present something that can be used to specify completely new namespace. How should GUI know what type of field to render?
And the same for the parser/validator. The parser can validate references to data types, but it has to know that this particular QName is a data type reference.
We need namespaces. But people hate them. Therefore we should hide them whenever possible.
E.g. we should not force user to set "org:manager" or "http://midpoint.evolveum.com/xml/ns/public/common/org-3#namaner" in XML or input fields. Just specifying "manager" should be enough - unless there is a naming conflict.
QNames and URIs are basically the same thing: XML Conventions
Yet, there are two data types in XSD: QName and anyURI.
The solution has several parts:
Create new data type for qname-that-is-also-uri data type. That should be used instead of existing QName and anyURI data types. And it should act as both. The values may be full URIs, local part of a QName or (legacy) QName with a namespace prefix (XML only).
In case that only a local part of a QName is specified then the namespace is automatically added. The default namespace may be defined in the schema (as special schema annotation). Or the QName will go with empty namespace and in that case parser or application (midPoint) will try to match the local part to existing namespace. The default namespace from the enclosing XML document (xmlns="…") will not be used.
For the QNames that point to a schema data types there will be another special annotation. It will specify the "root" data type that is valid as a value. E.g. it will specify "FocusType".
Why do the classes generated from midPoint schemans not have @XmlRootElement annotations? The @XmlRootElement annotation is really a tricky one. The generated Java class corresponds to XSD complex type. XSD complex type can be used with many elements under normal circumstances. If it is used by many elements, what element name would you put in @XmlRootElement annotation? You may not even know all the elements, because the data type may be used by other schemas. So, the JAXB code generator puts the XmlRootElement annotation in the code only in cases that the data type can be used only with a single element. And that’s the case when the data type definition is embedded directly into element definition. We generally avoid such definitions in midPoint schemas, as we want to have maximum reusability of data types. Therefore perhaps the only such case in midPoint schemas is the "objects" element in the common schema.