The org.apache.forrest.core.document.DocumentFactory class is responsible for providing the type of document when a reader has retrieved it from the source URL. Clearly, there is an unlimited number of potential document types that Forrest can process. It is therefore necessary to be able to configure the DocumentFactory without touching the Java code. This document describes how this configuration is performed.
Using the mime tyoe works well where the mime-type uniquely identifies the document type, however, this is not always the case. For example, an XML document is always identified by the mime-type "application/xml". We therefore need a way of identifying the specific document type in such cases.
For XML documents we are often provided with a reference to a DTD that defines the XML document type. We can use this information to define the document type within Forrest. For example:
The above definition says that when an XML document is discovered we should look to see if there is a reference to a DTD with the public ID of "-//APACHE//DTD Documentation V2.0//EN". If there is such a reference in a doctype definition then we have identified the XML document type with the ID "org.apache.forrest.xdoc2".
Using the Doctype definition is fine if the document is well formed, but it is not a requirement for an XML document to be well formed. We therefore need a way of identifying a document type when the XML source does not contain a DTD.
Unfortunately, there is no guarenteed method for identifying such an XML document. The best we can do is look at the struture of the document and infer the type from what we find. For example:
This definition simply looks at the root element of the document and, if a match is found then we have identified the document type. Clearly, this could result in false matches where two document types have the same root element. We therefore need a way of further refining the document structure definition. For example:
This definition expects to see a root element called "root" with an "id" and a "url" attibute.
A different refinement would be:
This definition expects a root element with the name "root" and two child elements named "child1" and "child2". These elements need not be in this order, but they must both be present. The "child2" element must have an id with the name "colour".