Source code for owlrl.OWLRLExtras

# -*- coding: utf-8 -*-
#
"""

Extension to OWL 2 RL, ie, some additional rules added to the system from OWL 2 Full. It is implemented through
the :class:`.OWLRL_Extension` class, whose reference has to be passed to the relevant semantic class (i.e., either the OWL 2 RL
or the combined closure class) as an 'extension'.

The added rules and features are:

    - self restriction
    - owl:rational datatype
    - datatype restrictions via facets
 
In more details, the rules that are added:

    1. self restriction 1: :code:`?z owl:hasSelf ?x. ?x owl:onProperty ?p. ?y rdf:type ?z. => ?y ?p ?y.`
    2. self restriction 2: :code:`?z owl:hasSelf ?x. ?x owl:onProperty ?p. ?y ?p ?y. => ?y rdf:type ?z.`

**Requires**: `RDFLib`_, 4.0.0 and higher.

.. _RDFLib: https://github.com/RDFLib/rdflib

**License**: This software is available for use under the `W3C Software License`_.

.. _W3C Software License: http://www.w3.org/Consortium/Legal/2002/copyright-software-20021231

**Organization**: `World Wide Web Consortium`_

.. _World Wide Web Consortium: http://www.w3.org

**Author**: `Ivan Herman`_

.. _Ivan Herman: http://www.w3.org/People/Ivan/

"""

__author__ = "Ivan Herman"
__contact__ = "Ivan Herman, ivan@w3.org"
__license__ = "W3C® SOFTWARE NOTICE AND LICENSE, http://www.w3.org/Consortium/Legal/2002/copyright-software-20021231"

import rdflib
from rdflib.namespace import RDF, RDFS, OWL, XSD

from fractions import Fraction as Rational

from .DatatypeHandling import AltXSDToPYTHON


# noinspection PyPep8Naming
from .CombinedClosure import RDFS_OWLRL_Semantics
from .OWLRL import OWLRL_Annotation_properties

from .XsdDatatypes import OWL_RL_Datatypes, OWL_Datatype_Subsumptions

from .RestrictedDatatype import extract_faceted_datatypes

#######################################################################################################################
# Rational datatype


# noinspection PyPep8Naming
def _strToRational(v):
    """Converting a string to a rational.

    According to the OWL spec: numerator must be an integer, denominator a positive integer (ie, xsd['integer'] type),
    and the denominator should not have a '+' sign.

    @param v: the literal string defined as boolean
    @return corresponding Rational value
    @rtype: Rational
    @raise ValueError: invalid rational string literal
    """
    try:
        r = v.split("/")
        if len(r) == 2:
            n_str = r[0]
            d_str = r[1]
        else:
            n_str = r[0]
            d_str = "1"
        if d_str.strip()[0] == "+":
            raise ValueError("Invalid Rational literal value %s" % v)
        else:
            return Rational(
                AltXSDToPYTHON[XSD.integer](n_str),
                AltXSDToPYTHON[XSD.positiveInteger](d_str),
            )
    except:
        raise ValueError("Invalid Rational literal value %s" % v)


#######################################################################################################################


# noinspection PyPep8Naming,PyBroadException
[docs]class OWLRL_Extension(RDFS_OWLRL_Semantics): """ Additional rules to OWL 2 RL. The initialization method also adds the :code:`owl:rational` datatype to the set of allowed datatypes with the :py:func:`._strToRational` function as a conversion between the literal form and a Rational. The :code:`xsd:decimal` datatype is also set to be a subclass of :code:`owl:rational`. Furthermore, the restricted datatypes are extracted from the graph using a separate method in a different module (:py:func:`.RestrictedDatatype.extract_faceted_datatypes`), and all those datatypes are also added to the set of allowed datatypes. In the case of the restricted datatypes and extra subsumption relationship is set up between the restricted and the base datatypes. :cvar extra_axioms: Additional axioms that have to be added to the deductive closure (in case the axiomatic triples are required). :var restricted_datatypes: list of the datatype restriction from :class:`.RestrictedDatatype`. :type restricted_datatypes: list of L{restricted datatype<RestrictedDatatype.RestrictedDatatype>} instances """ extra_axioms = [ (OWL.hasSelf, RDF.type, RDF.Property), (OWL.hasSelf, RDFS.domain, RDF.Property), ] def __init__(self, graph, axioms, daxioms, rdfs=False): """ @param graph: the RDF graph to be extended @type graph: rdflib.Graph @param axioms: whether (non-datatype) axiomatic triples should be added or not @type axioms: Boolean @param daxioms: whether datatype axiomatic triples should be added or not @type daxioms: Boolean @param rdfs: whether RDFS extension is done @type rdfs: boolean """ RDFS_OWLRL_Semantics.__init__(self, graph, axioms, daxioms, rdfs) self.rdfs = rdfs self.add_new_datatype( OWL.rational, _strToRational, OWL_RL_Datatypes, subsumption_dict=OWL_Datatype_Subsumptions, subsumption_key=XSD.decimal, subsumption_list=[OWL.rational], ) self.restricted_datatypes = extract_faceted_datatypes(self, graph) for dt in self.restricted_datatypes: self.add_new_datatype( dt.datatype, dt.toPython, OWL_RL_Datatypes, subsumption_dict=OWL_Datatype_Subsumptions, subsumption_key=dt.datatype, subsumption_list=[dt.base_type], ) # noinspection PyShadowingNames def _subsume_restricted_datatypes(self): """ A one-time-rule: all the literals are checked whether they are (a) of type restricted by a faceted (restricted) datatype and (b) whether the corresponding value abides to the restrictions. If true, then the literal gets an extra tag as being of type of the restricted datatype, too. """ literals = self._literals() for rt in self.restricted_datatypes: # This is a recorded restriction. The base type is: base_type = rt.base_type # Look through all the literals for lt in literals: # check if the type of that literal matches. Note that this also takes # into account the subsumption datatypes, that have been taken # care of by the 'regular' OWL RL process if (lt, RDF.type, base_type) in self.graph: try: # the conversion or the check may go wrong and raise an exception; then simply move on if rt.checkValue(lt.toPython()): # yep, this is also of type 'rt' self.store_triple((lt, RDF.type, rt.datatype)) except: continue
[docs] def restriction_typing_check(self, v, t): """ Helping method to check whether a type statement is in line with a possible restriction. This method is invoked by rule "cls-avf" before setting a type on an allValuesFrom restriction. The method is a placeholder at this level. It is typically implemented by subclasses for extra checks, e.g., for datatype facet checks. :param v: the resource that is to be 'typed'. :param t: the targeted type (i.e., Class). :return: Boolean. :rtype: bool """ # Look through the restricted datatypes to see if 't' corresponds to one of those... # There are a bunch of possible exceptions here with datatypes, but they can all # be ignored... try: for rt in self.restricted_datatypes: if rt.datatype == t: # bingo if v in self.literal_proxies.bnode_to_lit: return rt.checkValue( self.literal_proxies.bnode_to_lit[v].lit.toPython() ) else: return True # if we got here, no restriction applies return True except: return True
[docs] def one_time_rules(self): """ This method is invoked only once at the beginning, and prior of, the forward chaining process. At present, only the L{subsumption} of restricted datatypes<_subsume_restricted_datatypes>} is performed. """ RDFS_OWLRL_Semantics.one_time_rules(self) # it is important to flush the triples at this point, because # the handling of the restriction datatypes rely on the datatype # subsumption triples added by the superclass self.flush_stored_triples() self._subsume_restricted_datatypes()
[docs] def add_axioms(self): """ Add the :class:`.OWLRL_Extension.extra_axioms`, related to the self restrictions. """ RDFS_OWLRL_Semantics.add_axioms(self) for t in self.extra_axioms: self.graph.add(t)
[docs] def rules(self, t, cycle_num): """ Go through the additional rules implemented by this module. :param t: A triple (in the form of a tuple). :type t: tuple :param cycle_num: Which cycle are we in, starting with 1. This value is forwarded to all local rules; it is also used locally to collect the bnodes in the graph. :type cycle_num: int """ RDFS_OWLRL_Semantics.rules(self, t, cycle_num) z, q, x = t if q == OWL.hasSelf: for p in self.graph.objects(z, OWL.onProperty): for y in self.graph.subjects(RDF.type, z): self.store_triple((y, p, y)) for y1, y2 in self.graph.subject_objects(p): if y1 == y2: self.store_triple((y1, RDF.type, z))
# noinspection PyPep8Naming
[docs]class OWLRL_Extension_Trimming(OWLRL_Extension): """ This Class adds only one feature to :class:`.OWLRL_Extension`: to initialize with a trimming flag set to :code:`True` by default. This is pretty experimental and probably contentious: this class *removes* a number of triples from the Graph at the very end of the processing steps. These triples are either the by-products of the deductive closure calculation or are axiom like triples that are added following the rules of OWL 2 RL. While these triples *are necessary* for the correct inference of really 'useful' triples, they may not be of interest for the application for the end result. The triples that are removed are of the form (following a SPARQL-like notation): - :code:`?x owl:sameAs ?x`, :code:`?x rdfs:subClassOf ?x`, :code:`?x rdfs:subPropertyOf ?x`, :code:`?x owl:equivalentClass ?x` type triples. - :code:`?x rdfs:subClassOf rdf:Resource`, :code:`?x rdfs:subClassOf owl:Thing`, :code:`?x rdf:type rdf:Resource`, :code:`owl:Nothing rdfs:subClassOf ?x` type triples. - For a datatype that does *not* appear explicitly in a type assignments (ie, in a :code:`?x rdf:type dt`) the corresponding :code:`dt rdf:type owl:Datatype` and :code:`dt rdf:type owl:DataRange` triples, as well as the disjointness statements with other datatypes. - annotation property axioms. - a number of axiomatic triples on :code:`owl:Thing`, :code:`owl:Nothing` and :code:`rdf:Resource` (eg, :code:`owl:Nothing rdf:type owl:Class`, :code:`owl:Thing owl:equivalentClass rdf:Resource`, etc). Trimming is the only feature of this class, done in the :py:meth:`.post_process` step. If users extend :class:`OWLRL_Extension`, this class can be safely mixed in via multiple inheritance. :param graph: The RDF graph to be extended. :type graph: :class:`rdflib.Graph` :param axioms: Whether (non-datatype) axiomatic triples should be added or not. :type axioms: bool :param daxioms: Whether datatype axiomatic triples should be added or not. :type daxioms: bool :param rdfs: Whether RDFS extension is done. :type rdfs: bool """ def __init__(self, graph, axioms, daxioms, rdfs=False): """ @param graph: the RDF graph to be extended @type graph: rdflib.Graph @param axioms: whether (non-datatype) axiomatic triples should be added or not @type axioms: Boolean @param daxioms: whether datatype axiomatic triples should be added or not @type daxioms: Boolean @param rdfs: whether RDFS extension is done @type rdfs: boolean """ OWLRL_Extension.__init__(self, graph, axioms, daxioms, rdfs=False)
[docs] def post_process(self): """ Do some post-processing step performing the trimming of the graph. See the :class:`.OWLRL_Extension_Trimming` class for further details. """ OWLRL_Extension.post_process(self) self.flush_stored_triples() to_be_removed = set() for t in self.graph: s, p, o = t if s == o: if ( p == OWL.sameAs or p == OWL.equivalentClass or p == RDFS.subClassOf or p == RDFS.subPropertyOf ): to_be_removed.add(t) if ( (p == RDFS.subClassOf and (o == OWL.Thing or o == RDFS.Resource)) or (p == RDF.type and o == RDFS.Resource) or (s == OWL.Nothing and p == RDFS.subClassOf) ): to_be_removed.add(t) for dt in OWL_RL_Datatypes: # see if this datatype appears explicitly in the graph as the type of a symbol if len([s for s in self.graph.subjects(RDF.type, dt)]) == 0: to_be_removed.add((dt, RDF.type, RDFS.Datatype)) to_be_removed.add((dt, RDF.type, OWL.DataRange)) for t in self.graph.triples((dt, OWL.disjointWith, None)): to_be_removed.add(t) for t in self.graph.triples((None, OWL.disjointWith, dt)): to_be_removed.add(t) for an in OWLRL_Annotation_properties: self.graph.remove((an, RDF.type, OWL.AnnotationProperty)) to_be_removed.add((OWL.Nothing, RDF.type, OWL.Class)) to_be_removed.add((OWL.Nothing, RDF.type, RDFS.Class)) to_be_removed.add((OWL.Thing, RDF.type, OWL.Class)) to_be_removed.add((OWL.Thing, RDF.type, RDFS.Class)) to_be_removed.add((OWL.Thing, OWL.equivalentClass, RDFS.Resource)) to_be_removed.add((RDFS.Resource, OWL.equivalentClass, OWL.Thing)) to_be_removed.add((OWL.Class, OWL.equivalentClass, RDFS.Class)) to_be_removed.add((OWL.Class, RDFS.subClassOf, RDFS.Class)) to_be_removed.add((RDFS.Class, OWL.equivalentClass, OWL.Class)) to_be_removed.add((RDFS.Class, RDFS.subClassOf, OWL.Class)) to_be_removed.add((RDFS.Datatype, RDFS.subClassOf, OWL.DataRange)) to_be_removed.add((RDFS.Datatype, OWL.equivalentClass, OWL.DataRange)) to_be_removed.add((OWL.DataRange, RDFS.subClassOf, OWL.Datatype)) to_be_removed.add((OWL.DataRange, OWL.equivalentClass, OWL.Datatype)) for t in to_be_removed: self.graph.remove(t)