Skip to main content

An RDFa parser wth a simple dictionary-like interface.

Project description


Nathan R. Yergler <>

Creative Commons


rdfadict and its dependencies may be installed using easy_install (recommended)

$ easy_install rdfadict

or by using the standard distutils

$ python install

If you are installing from source, you will also need the following packages:

easy_install will satisfy depedencies for you if necessary.


rdfadict parses RDFa metadata encoded in HTML or XHTML documents. It can parse a block of text (as a string), or a URL. For example, given the following block of sample text:

>>> rdfa_sample = """
... <div xmlns:dc=""
...      xmlns:xsd="">
... <h1 property="dc:title">Vacation in the South of France</h1>
... <h2>created
... by <span property="dc:creator">Mark Birbeck</span>
... on <span property="dc:date" type="xsd:date"
...          content="2006-01-02">
...   January 2nd, 2006
...    </span>
... </h2>
... </div>"""

Triples can be extracted using rdfadict:

>>> import rdfadict
>>> base_uri = ""
>>> parser = rdfadict.RdfaParser()
>>> triples = parser.parse_string(rdfa_sample, base_uri)

We define the variable base_uri to let the parser know what URI assertions without subjects apply to.

Based on our example text, we expect to get three triples back – title, creator and date. Triple are indexed as a dictionary, first by subject, then by predicate, finally retuning a list of objects. For example, a list of all subjects is retrieved using:

>>> triples.keys()

If assertions were made about resources other than the default, those URIs would appear in this list. We can verify how many predicates were found for this subject by accessing the next level of the dictionary:

>>> len(triples[''].keys())

Finally, we can retrieve the value for the title by fully dereferencing the dictionary:

>>> triples[''][
...     '']
['Vacation in the South of France']

Note that the objects are stored as a list by the default triple sink.

Multiple Assertions

Because the property attribute always denotes triple with a literal string as its object and rel and rev denote triples with URIs as their objects, it is possible to make multiple assertions with a single HTML tag.

For example:

>>> multi_rdfa = """
... <div xmlns:foaf=""
...      xmlns:dc="">
...   This photo was taken by <a about="photo1.jpg" property="dc:title"
...   content="Portrait of Mark" rel="dc:creator"
...   rev="foaf:img"
...   href="">Mark Birbeck</a>.
... </div>
... """

In this statement we are making three assertions: two involving URI objects (specified by rel and rev), and one involving the property.

>>> import rdfadict
>>> parser = rdfadict.RdfaParser()
>>> multi_base_uri = ""
>>> triples = parser.parse_string(multi_rdfa, multi_base_uri)

We expect the triples generated to have two subjects: the photo URI (for the rel and property assertions) and the href URI (for the rev assertion).

>>> len(triples.keys()) == 2
>>> '' in triples.keys()
>>> '' in triples.keys()

Finally, we verify that the assertions made about each subject are correct:

>>> len(triples[''].keys()) == 2
>>> triples[''] \
...          ['']
>>> triples[''] \
...          ['']
['Portrait of Mark']
>>> triples['']
{'': ['']}

Resolving Statements

When resolving statements, the REL, REV, CLASS and PROPERTY attributes expect a CURIE, while the HREF property expects a URI. When resolving CURIEs, un-namespaced values which are not HTML reserved words (such as license) are ignored to prevent “triple bloat”.

Given an example:

>>> rdfa_sample2 = """
... <div xmlns:dc=""
...      xmlns:xsd="">
... <link rel="alternate" href="/foo/bar" />
... <h1 property="dc:title">Vacation in the South of France</h1>
... <h2>created
... by <span property="dc:creator">Mark Birbeck</span>
... on <span property="dc:date" type="xsd:date"
...          content="2006-01-02">
...   January 2nd, 2006
...    </span>
... </h2>
... <img src="/myphoto.jpg" class="photo" />
... (<a href="" rel="license"
...    about="/myphoto.jpg">CC License</a>)
... </div>"""

We can extract RDFa triples from it:

>>> parser = rdfadict.RdfaParser()
>>> base_uri2 = ""
>>> triples = parser.parse_string(rdfa_sample2, base_uri2)

This block of RDFa includes a license statement about another document, the photo:

>>> len(triples[""])
>>> triples[""].keys()
>>> triples[""] \
...    ['']

There are two things to note with respect to this example. First, the relative URI for the photo is resolved with respect to the base_uri value. Second, the “class” attribute is not processed, because it’s value is not in a declared namespace:

>>> 'photo' in [ n.lower() for n in
...      triples[''].keys() ]

Similar to this case is the link tag in the example HTML. Based on the subject resolution rules for link and meta tags, no subject can be resolved for this assertion. However, this does not throw an exception because the value of the rel attribute is not namespaced.

Consider an alternative, contrived example:

>>> link_sample = """
... <div xmlns:dc=""
...      xmlns:xsd=""
...      about="">
... <link rel="dc:creator" href="" />
... </div>"""

Based on the subject resolution rules for link tags, we expect to see one assertion: that represents the creator of This can be tested; note we supply a different base_uri to ensure the subject is being properly resolved.

>>> parser = rdfadict.RdfaParser()
>>> link_base_uri = ''
>>> triples = parser.parse_string(link_sample, link_base_uri)
>>> triples.keys()
>>> len(triples[''])
>>> triples['']['']

Note that this HTML makes no assertions about the source document:

>>> link_base_uri in triples.keys()

If the HTML sample is modified slightly, and the about attribute is omitted, rdfadict is resolves the subject to the explicit base URI.

>>> link_sample = """
... <div xmlns:dc=""
...      xmlns:xsd="" >
... <link rel="dc:creator" href="" />
... </div>"""
>>> parser = rdfadict.RdfaParser()
>>> link_base_uri = ''
>>> triples = parser.parse_string(link_sample, link_base_uri)
>>> link_base_uri in triples.keys()

If a namespace is unable to be resolved, the assertion is ignored.

>>> ns_sample = """
... <a href="" rel="foo:bar">Content</a>
... """
>>> parser = rdfadict.RdfaParser()
>>> triples = parser.parse_string(ns_sample, '')
>>> triples

See the RDFa Primer for more RDFa examples.

Parsing Files

rdfadict can parse from three sources: URLs, file-like objects, or strings. The examples thus far have parsed strings using the parse_string method. A file-like object can also be used:

>>> from StringIO import StringIO
>>> file_sample = """
... <html>
...  <body>
...    <a href=""
...       rel="license">the license</a>
...  </body>
... </html>
... """
>>> parser = rdfadict.RdfaParser()
>>> result = parser.parse_file(StringIO(file_sample),
...                            "")
>>> result.keys()
>>> result['']
{'': ['']}

Parsing By URL

rdfadict can parse a document retrievable by URI. Behind the scenes it uses urllib2 to open the document.

>>> parser = rdfadict.RdfaParser()
>>> result = \
... parser.parse_url('')
>>> print result['']\
... [''][0]
表示 2.1 日本

Note that parse_file is not recommended for use with urllib2 handler objects. In the event that pyRdfa encounters a non-XHTML source, it re-opens the URL to begin processing with a more tolerant parser. When parse_file is used to initiate parsing, it is unable to re-open the URL correctly.

Triple Sinks

rdfadict uses a simple interface (the triple sink) to pass RDF triples extracted back to some storage mechanism. A class which acts as a triple sink only needs to define a single method, triple. For example:

class StdOutTripleSink(object):
    """A triple sink which prints out the triples as they are received."""

    def triple(self, subject, predicate, object):
        """Process the given triple."""

        print subject, predicate, object

The default triple sink models the triples as a nested dictionary, as described above. Also included with the package is a list triple sink, which stores the triples as a list of 3-tuples. To use a different sink, pass an instance in as the sink parameter to either parse method. For example:

>>> parser = rdfadict.RdfaParser()
>>> list_sink = rdfadict.sink.SimpleTripleSink()
>>> result = parser.parse_string(rdfa_sample, base_uri, sink=list_sink)
>>> result is list_sink
>>> len(list_sink)

Note that the parse method returns the sink used. Since the sink we’re using is really just a list, the interpreter prints the contents upon return.

Limitations and Known Issues

rdfadict currently does not implement the following areas properly; numbers in parenthesis refer to the section number in the RDFa Syntax Document.

  • xml:base is not respected (2.3)

  • Typing is not implemented; this includes implicit XMLLiteral typing as well as explicit types specified by the datatype attribute (5.1)

  • Blank nodes are not guaranteed to work per the syntax document (5.2); if you try to use them, you will probably be disappointed.

  • Reification is not implemented (5.3).

Change History

0.7.3 (2011-01-27)

  • Corrected minor bug; the Sets modules was imported in all cases, resulting in a DeprecationWarning. (2010-03-17)

  • Added missing, which prevented from working when run from an sdist (as opposed to a checkout).

0.7.1 (2009-07-20)

  • Updated DictSetTripleSink to coerce triple() parameters to unicode instead of str.

0.7 (2009-06-02)

  • DictTripleSink uses encode instead of str, making it friendlier to Unicode.

  • Eliminated custom pyRDFa wrapper.

  • Added a test for handling Unicode triples.

0.6 (2008-10-14)

  • Added DictSetTripleSink

0.5.2 (2008-08-14)

  • Corrected bug with parse_url; non-XHTML sources will now be parsed correctly.

0.5.1 (2008-08-13)

  • Added parse_file method for parsing data from a file-like object.

  • parseurl and parsestring are now aliased to parse_url and parse_string respectively.

0.5 (2008-08-12)

  • rdfadict now acts as a wrapper for pyRdfa for full compliance with the candidate recommendation.

  • The cc namespace is no longer special cased with a default value.

  • Removed tidy extra and uTidylib dependency; parsing is now handled by pyRdfa which uses html5lib for handling more broken HTML.

  • Doctests are now in README.txt in the rdfadict package.

  • The default XHTML namespace is now instead of

0.4.2 (2007-06-05)

  • Corrected dependency link for uTidylib.

0.4.1 (2007-03-21)

  • Use uTidylib instead of os.system for wrapping tidy.

0.4.0 (2007-03-20)

  • Provide rudimentary fallback to Tidy when we encounter HTML which is not well-formed XML.

0.3.3 (2007-03-14)

  • Removed special case for cc:license; instead, cc namespace simply has a default value of

0.3.2 (2007-03-12)

  • Ignore assertions which have unresolvable namespace prefixes.

  • Special case handling for cc:license.

0.3.1 (2007-03-09)

  • Fixed bug in subject resolution exception handling.

0.3 (2007-03-08)

  • Fixed resolution of URIs v. CURIEs

  • Drop assertions with non-namespaced CURIEs as the predicate (per updated spec)

  • Updated test suite to comply with updated RDFa specification

  • Corrected subject resolution behavior for <link> and <meta> elements

  • Implemented entry point and extractor interface for compatibility with the ccrdf.rdfextract library.

  • Fixed parsing of rev assertions, which was formerly completely broken.

0.2 (2006-11-21)

  • Directly subclass list and dict for our sample triple sinks

  • Additional package metadata for PyPI

  • Additional documentation of sink interface and tests for the SimpleTripleSink

0.1 (2006-11-20)

  • Initial public release


Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

rdfadict-0.7.4.tar.gz (13.3 kB view hashes)

Uploaded source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page