SYNOPSIS
use RDF::Trine::Namespace qw(rdf rdfs);
my $one_triple = "<test/classA> <${rdfs}domain> <test/ClassB> .";
my $reasoner = RDF::TrineX::RuleEngine::Jena->new;
my $model_inferred = $reasoner->apply_rules(
input => \ $one_triple,
rules => 'rdfs-fb',
purge_schemas => ':all',
);
print $model_inferred->size;
# 7
my $serializer = RDF::Trine::Serializer->new('turtle' , namespaces => { rdf => $rdf, rdfs => $rdfs });
print $serializer->serialize_model_to_string( $model_inferred );
# <test/ClassB> rdfs:subClassOf rdfs:Resource, <test/ClassB> ;
# a rdfs:Class .
# <test/classA> rdfs:domain <test/ClassB> ;
# a rdf:Property, rdfs:Resource .
DESCRIPTION
This module is a convenience wrapper around a call to Jena's `
jena.RuleMap ' command line rule-engine interface. It transparently
handles serialization and creation of temporary files, but it relies on
a working Java installation and knowledge of the location of the Jena
framework.
Finding Jena
When building this module, the Jena framework can be downloaded or a
path to an existing Jena installation can be specified. This path is
stored in a shared file. If you can't or don't want to specify it at
build time, you can set the JENAROOT environment variable to the
location of the extracted Jena download. Finally you can pass the path
to it at runtime to the constructor.
RDF::Trine vs. Jena Format names
Trine | Jena
---------+----------------------------
ntriples | N-TRIPLE
turtle | TURTLE
rdxml | RDF/XML, RDF/XML-ABBREV
n3 | N3-PP, N3-PLAIN, N3-TRIPLE
ATTRIBUTES
JENAROOT
A Path::Class::Dir object of the Jena directory.
JENA_VERSION
The Version of Jena used, determined from the
`jena-X.X.X-sources.jar' file.
JENA_SOURCES_JAR
Archive::Zip object for the `jena-X.X.X-sources.jar' file. Contains
the predefined rulesets.
JENA_CLASSPATH
Array reference holding the paths to all the `<jar'> files required
for Jena to run.
METHODS
new
Returns a new RDF::TrineX::RuleEngine::Jena object. Before
The optional `JENAROOT' argument holds the path to the extracted Jena
source. If not set, `JENAROOT' is determined as described in JENAROOT.
apply_rules
Applies a set of Jena rules to RDF input and adds the inferred
statements to the output model.
`input => $input_data'
required
`$input_data' is serialized, written to a temporary file and fed to
exec_jena_rulemap as the `filename_input' argument. Currently, the
following data types are handled:
* RDF::Trine::Model.
my $model = RDF::Trine::Model->temporary_model;
RDF::Trine::Parser->new('turtle')->parse_file_into_model('my_file.ttl');
$reasoner->apply_rules(
input => $model,
rules => ...,
);
* String: Treated as the path to a file containing a serialized RDF
graph.
$reasoner->apply_rules(
input => 'my_file.nt',
rules => ...,
);
* Scalar reference: Treated as a reference to a serialized RDF
graph.
my $input_ttl = <'EOF';
@prefix rdfs:http://www.w3.org/2000/01/rdf-schema# .
<Tiny> rdfs:subClassOf <Small> .
EOF
$reasoner-apply_rules(
input => \ $input_ttl,
input_format => 'TURTLE',
rules => ...,
);
`rules => $rules_data'
required
`$rules_data' can be any of the following:
* String matching one of the available_rulesets: The appropriate
rules file is loaded from JENA_SOURCES_JAR.
$reasoner->apply_rules(
input => ...,
rules => 'rdfs',
);
* Scalar reference: The dereferenced value is treated as a string of
rules.
my $rules = "[dummy: (?a ?b ?c) -> (?a rdfs:label "This is stupid") ]";
$reasoner->apply_rules(
input => ...,
rules => \ $rules,
);
* Any other string: Treat `$rules_data' as a filename and load rules
from there.
$reasoner->apply_rules(
input => ...,
rules => '/path/to/my/ruleset.rules',
);
`output => ($model|":fh"|":filename"|":string"|$string)'
If specified, inferred statements are written to this model,
otherwise a temporary model is created. If you set output to the
same value as input, inferred statements are added to the original
model.
* `$model': The statements are added to this RDF::Trine::Model.
Setting this to the same model as in `input' will cause all
rule-based statement removals to be ignored since there currently is
no way of tracking which statements *were* by applying the rules.
* `":fh"': If this special string (case-insensitive) is supplied, a
readable filehandle to the raw output of jena.RuleMap is returned.
`purge_schemas' is ignored.
my $fh = $reasoner->apply_rules(
input => ...,
rules => ...,
output => ':FH',
);
while (<$fh>) {
my ($s, $p, $o ) = $_ =~ m/^\s*<([^>]+>\s+<([^>]+>\s+<([^>]+>\s*.$/;
}
* `":filename"': If this special string (case-insensitive) is
supplied, the filename of the temporary file containing the raw
output of jena.rulemap is returned . `purge_schemas' is ignored.
use File::Slurp;
my $fname = $reasoner->apply_rules(
input => ...,
rules => ...,
output => ':filename',
);
my $contents = read_file $fname;
* `":string"': If this special string (case-insensitive) is
supplied, the complete raw output of jena.RuleMap is returned.
`purge_schemas' is ignored.
my $serialized = $reasoner->apply_rules(
input => ...,
rules => ...,
output => ':sTRing',
);
* `$string': Any other string is treated as a filename to write the
raw output of jena.RuleMap to. `purge_schemas' is ignored.
my $serialized = $reasoner->apply_rules(
input => 'data.nt',
rules => ...,
output => 'data_inferred.nt',
);
`purge_schemas => (\@list_of_schemanames|":all")'
Jena's rule engine adds lots and lots of schema statements about
rdf, rdfs, owl, xsd plus some internals. You can tell
RDF::TrineX::RuleEngine::Jena to purge those statements by supplying
an array ref of schema names to purge_schemas.
Specifying `:all' removes all schema statements,
RDF::TrineX::RuleEngine::Jena knows about.
$reasoner->apply_rules(
input => ...,
rules => ...,
purge_schemas => ':all',
);
is equivalent to
$reasoner->apply_rules(
input => ...,
rules => ...,
purge_schemas => [qw( rdf rdfs daml xsd owl jena )],
);
exec_jena_rulemap
Sets and resets CLASSPATH and runs `java jena.RuleMap ...' using a
system call. This is all this function does, capturing STDIN and STDERR
and parsing/serializing happens in apply_rules.
Arguments:
filename_rules
Filename of the `.rules' file
filename_input
File name of the file containing the assertions.
input format
The format of the input file, in Jena notation (i.e. 'N-TRIPLE',
'TURTLE', 'RDF/XML'...)
output_format
Format of the result printed to STDOUT, again in Jena notation.
additions_only
When this flag is set, Jena will only return deduced and schema
statements, as opposed to the original model with added and removed
statements when the flag is not set.
_model_difference
Given two models A and B, remove all statements from A that are also in
B.
_remove_tautologies
Remove all statements of the form `X owl:equivalentProperty X'.
available_rulesets
Lists the available predefined rulesets shipped with Jena that aren't
broken. Currently, these are:
* daml-micro
* owl-fb
* owl-fb-micro
* owl-fb-mini
* rdfs
* rdfs-b
* rdfs-b-tuned
* rdfs-fb
* rdfs-fb-lp-expt
* rdfs-fb-tgc
* rdfs-fb-tgc-noresource
* rdfs-noresource
get_ruleset_filename
Get the filename of a predefined ruleset within JENA_SOURCES_JAR.
AUTHOR
Konstantin Baierer <kba@cpan.org>
SEE ALSO
http://answers.semanticweb.com/questions/1453/reasoning-and-sparql-throu
gh-arq-command-line