<page title="...">
  ...
  <rdf pred="..." obj="..."/>
  ...
</page>

Subject is the document url, pred is the IRI of the predicate (possibly of the form prefix:term). obj is the IRI of the object.

<page title="...">
  ...
  <rdf pred="..." href="..."/>
  ...
</page>

Subject and predicate are as in the first example. href is of the form path[#id] to indicate an document of the site, and eventually a node id in this document. This is reduced to the complete IRI of the document or block referenced by document#id.

<page title="...">
   ...
   <p id="id1">
     <rdf pred="..." .../>
   </p>
</page>

Here the subject is the IRI formed by the document url and the id of the <p> node, for example http://my.example.com/foo/page1.html#id1 if my document has path foo/page1 and base url of the site is http://my.example.com.

To add the triples, the <rdf pred="pref:term" ...>...</rdf> node is rewritten to

<rdf:Description rdf:about="<document-url>[#id]">
  <pref:term ...>
    ...
  </pref:term>
</rdf:Description>

and this XML tree is parsed using OCaml-RDF RDF/XML parser to add the triples to the document graph. If a href or obj attribute was used, the corresponding IRI is associated to the `rdf:resource` attribute in the rewritten node:

<rdf:Description rdf:about="<document-url>[#id]">
  <pref:term rdf:resource="..." ...>
    ...
  </pref:term>
</rdf:Description>

and the whole <rdf:Description> node is parsed by the RDF/XML parser.

It is possible to insert triple with a subject not being the current document or a block of the current document, using the subject attribute:

<rdf subject="http://foo.bar" pred="..." .../>

It is sometimes useful to force the subject to be the current document url even if the <rdf> node is under a XML node with an id:

<p id="...">
  <rdf subject="&lt;doc-url/&gt;" pred="..." .../>
</p>

Graphs to load at startup are given in the sources parameter of the .stog/config-rdf file. For each graph, we indicate its name (an IRI), and options. Options describe where to load the graph from.

Here is an example of configuration file describing two graphs to load, one from a file, and one from a MySQL database:

{
  namespaces: "pairs (uri, name) specifying namespaces",
  namespaces: [("http://super-namespace.net/", "super")],

  "graph-file": "name of main graph output file",
  "graph-file": "graph.rdf",

  sources: "pairs (uri, options) specifying graphs to load and associate to uris.
            Options is a list of pair (name, value). To load a load, use
            [\"file\", \"myfile.ttl\"].
            Other options can be given to access graphs from a database
            (see OCaml-RDF's Rdf_graph documentation)",
  sources:
  [
   ("http://foo.net/", [ ("file", "foo_net.ttl") ]),
   ("http://foo.bar.net/",
    [("storage", "mysql2"), ("database", "genet_example"), ("user", "genet"),
     ("host", "localhost"), ("password", ""), ("port", "3306")])
  ]
}

The options associated to each IRI are the options passed to the Rdf.Graph.open_graph function in OCaml-RDF.

By default, files are expected to be in Turtle format. Setting the "format" option to "rdf/xml" indicates that the file is in RDF/XML format.

A graph can be loaded when the XML node <rdf-load> is encountered. This rule is registered in level 202.

The general form of using this rule is:

<rdf-load att1="..." att2="..."/>

Each attribute defines a pair (name, value) used as an option. The complete option list is used as the option lists specified in the .stog/config-rdf file (see Loading graphs at startup).

The following code loads a graph named http://foo.net/ from a Turtle file foo_net.ttl in the same directory as the source file of the current document:

<rdf-load file="./foo_net.ttl" name="http://foo.net/"/>

If the filename is not explicitely relative to the current directory (i.e. if it does not start with "." or "..") then is it used as is.

The following code loads a graph named http://foo.bar.net/ from a MySQL database, using the given options:

<rdf-load name="http://foo.bar.net/" storage="mysql2"
  database="genet_example" user="genet" host="localhost" password="" port="3306"/>

If no file attribute is given, and the <rdf-load> node is not empty, then the PCDATA content of the node is used as Turtle code describing the triples to load in the graph. For example:

<rdf-load name="http://foo.bar.org/"><![CDATA[
@prefix toto: <http://toto.net/>
<http://foo.bar.org/toto> a toto:file .
]]></rdf-load>