Alberto Attilio Reggiori


RDFStore::Parser::SiRPAC - This module implements a streaming RDF Parser as a direct implementation of XML::Parser::Expat(3)


        use RDFStore::Parser::SiRPAC;
        use RDFStore::NodeFactory;
        my $p=new RDFStore::Parser::SiRPAC(
                ErrorContext => 2,
                Handlers        => {
                        Init    => sub { print "INIT\n"; },
                        Final   => sub { print "FINAL\n"; },
                        Assert  => sub { print "STATEMENT - @_\n"; }
                NodeFactory     => new RDFStore::NodeFactory() );

        $p->parse(*STDIN); #parse stream but with *blocking* Expat (see below example for n-blocking parsing using XML::Parse::ExpatNB)

        use RDFStore::Parser::SiRPAC;
        use RDFStore::NodeFactory;
        my $pstore=new RDFStore::Parser::SiRPAC(
                ErrorContext    => 2,
                Style           => 'RDFStore::Parser::Styles::RDFStore::Model',
                NodeFactory     => new RDFStore::NodeFactory(),
                style_options   =>      {
                                        persistent      =>      1,
                                        seevalues       =>      1,
                                        store_options         =>      { Name => '/tmp/test' }
        my $rdfstore_model = $pstore->parsefile('');

        #using the expat no-blocking feature (generally for large XML streams) - see XML::Parse::Expat(3)
        my $rdfstore_stream_model = $pstore->parsestream(*STDIN);


This module implements a Resource Description Framework (RDF) streaming parser completely in Perl using the XML::Parser::Expat(3) module. The actual RDF parsing happens using an instance of XML::Parser::Expat with Namespaces option enabled and start/stop and char handlers set. The RDF specific code is based on the modified version of SiRPAC of Sergey Melnik in Java; a lot of changes and adaptations have been done to actually run it under Perl. Expat options may be provided when the RDFStore::Parser::SiRPAC object is created. These options are then passed on to the Expat object on each parse call.

Exactly like XML::Parser(3) the behavior of the parser is controlled either by the Style entry elsewhere in this document and/or the Handlers entry elsewhere in this document options, or by the setHandlers entry elsewhere in this document method. These all provide mechanisms for RDFStore::Parser::SiRPAC to set the handlers needed by Expat. If neither Style nor Handlers are specified, then parsing just checks the RDF document syntax against the W3C RDF Raccomandation . When underlying handlers get called, they receive as their first parameter the Expat object, not the Parser object.

To see some examples about how to use it look at the sections below and in the samples and utils directory coming with this software distribution.

E.g. With RDFStore::Parser::SiRPAC you can easily write an script to do something like this:

        fetch -o - -q | \
                gunzip - | \
                sed -f dmoz.content.sed | - 



This is a class method, the constructor for RDFStore::Parser::SiRPAC. Options are passed as keyword value pairs. Recognized options are:

  • NodeFactory

    This option is mandatory to run the RDFStore::Parser::SiRPAC parser correctly and must contain a reference to an object of type RDFStore::NodeFactory(3). Such a reference is used during the RDF parsing to create resources, literal and statements to be passed to the registered handlers. A sample implementation is RDFStore::NodeFactory that is provided with the RDFStore package.

  • Source

    This option can be specified by the user to set a base URI to use for the generation of resource URIs during parsing. If this option is omitted the parser will try to generate a prefix for generated resources using the input filename or URL actually containing the input RDF. In a near future such an option could be obsoleted by use of XMLBase W3C raccomandation.

  • GenidNumber

    Seed the counter for bNodes with the given value

  • bCreateBags

    Flag to generate a Bag for each Description element

  • Style

    This option provides an easy way to set a given style of parser. There is one sample Sylte module provided with the RDFStore::Parser::SiRPAC distribution called RDFStore::Parser::Styles::RDFStore::Model. Such a module uses the RDFStore::Model(3) to implement a simple RDF storage. Custom styles can be provided by giving a full package name containing at least one '::'. This package should then have subs defined for each handler it wishes to have installed. See "WRITE YOUR OWN PARSER" below for a discussion on how to build one.

  • Handlers

    When provided, this option should be an anonymous hash containing as keys the type of handler and as values a sub reference to handle that type of event. All the handlers get passed as their 1st parameter the instance of Expat that is parsing the document. Further details on handlers can be found in "HANDLERS". Any handler set here overrides the corresponding handler set with the Style option.

  • ErrorContext

    This is an XML::Parser option. When this option is defined, errors are reported in context. The value should be the number of lines to show on either side of the line in which the error occurred.

All the other XML::Parser and XML::Parser::Expat options should work freely with RDFStore::Parser::SiRPAC see XML::Parser(3) and XML::Parser::Expat(3).

setHandlers(TYPE, HANDLER [, TYPE, HANDLER [...]])

This method registers handlers for various parser events. It overrides any previous handlers registered through the Style or Handler options or through earlier calls to setHandlers. By providing a false or undefined value as the handler, the existing handler can be unset.

This method returns a list of type, handler pairs corresponding to the input. The handlers returned are the ones that were in effect prior to the call.

See a description of the handler types in "HANDLERS".

parse(SOURCE, URIBASE [, OPT => OPT_VALUE [...]])

The SOURCE parameter should either be a string containing the whole RDF document, or it should be an open IO::Handle. The URIBASE can be specified by the user to set a base URI to use for the generation of resource URIs during parsing. If this option is omitted the parser will try to generate a prefix for generated resources using either the Source option of the constructor, the input filename or URL actually containing the input RDF. In a near future such an option could be obsoleted by use of XMLBase W3C raccomandation. Constructor options to XML::Parser::Expat given as keyword-value pairs may follow the URIBASE parameter. These override, for this call, any options or attributes passed through from the RDFStore::Parser::SiRPAC instance.

A die call is thrown if a parse error occurs. Otherwise it will return 1 or whatever is returned from the Final handler, if one is installed. In other words, what parse may return depends on the style.

e.g. the RDFStore::Parser::SiRPAC::RDFStore Style module returns an instance of RDFStore::Model

parsestring(STRING, URIBASE [, OPT => OPT_VALUE [...]])

This is just an alias for parse for backwards compatibility.

parsefile(URL_OR_FILE [, OPT => OPT_VALUE [...]])

Open URL_OR_FILE for reading, then call parse with the open handle. If URL_OR_FILE is a full qualified URL this module uses Socket(3) to actually fetch the content. The URIBASE parse() parameter is set to URL_OR_FILE.


Return the current (latest) bNodes counter generated/managed by the parser


As Expat, SiRPAC is an event based parser. As the parser recognizes parts of the RDF document then any handlers registered for that type of an event are called with suitable parameters. All handlers receive an instance of XML::Parser::Expat as their first argument. See "METHODS" in XML::Parser::Expat for a discussion of the methods that can be called on this object.

Init (Expat)

This is called just before the parsing of the document starts.

Final (Expat)

This is called just after parsing has finished, but only if no errors occurred during the parse. Parse returns what this returns.

Assert (Expat, Statement)

This event is generated when a new RDF statement has been generated by the parseer.start tag is recognized. Statement is of type RDFStore::Statement(3) as generated by the RDFStore::NodeFactory(3) passed as argument to the RDFStore::Parser::SiRPAC constructor.

Start_XML_Literal (Expat, Element [, Attr, Val [,...]])

This event is generated when an XML start tag is recognized within an RDF property with parseType="Literal". Element is the name of the XML element type that is opened with the start tag. The Attr & Val pairs are generated for each attribute in the start tag.

This handler should return a string containing either the original XML chunck or one f its transformations, perhaps using XSLT.

Stop_XML_Literal (Expat, Element)

This event is generated when an XML end tag is recognized within an RDF property with parseType="Literal". Note that an XML empty tag (<foo/>) generates both a Start_XML_Literal and an Stop_XML_Literal event.

Char_XML_Literal (Expat, String)

This event is generated when non-markup is recognized within an RDF property with parseType="Literal". The non-markup sequence of characters is in String. A single non-markup sequence of encoding of the string in the original document, this is given to the handler in UTF-8.

This handler should return the processed text as a string.

manage_bNodes (Expat, Factory, SystembNode)

This event is triggered when a new anonymous resource (bNode) needs to be generated by the system e.g. 'genidrdfstoreS302e313439323736363337373935353039P22533T106968250320N21' or a rdf:nodeID attribute is found into the input RDF/XML. When is not rdf:nodeID, by default the system is trying ot use 'GenidNumber' base as passed to the parser constructor to count sequentially the bNode identifiers and internally create an anonymous RDF resource node. Otherwise the counter will start from zero (0). The system will then concatenate such a sequential counter/number to a another unique string built by a hex-encoded random number between 0-1 (i.e. unpack("H*", rand()) ), the current system Process ID (PID) and system timestamp. For example, if the 'GenidNumber' is set to '45', rand()=0.149276637795509 and PID=2233 and system timestamp=1069672550, the parser will generate 'genidrdfstoreS302e313439323736363337373935353039P2233T1069672550N45' as identifier. The user should note that such a unique ID (a part the 'genidrdfstore' prefix) will allow to post-process such an identifier to get out the PID, the timestamp and the counter parts by using the 'S', 'P', 'T' and 'N' chars separators. In addition, the prefix of such an identifier should be unique across different parser runs even on the same file/source i.e. random seed, PID and timestamp uniquely identify the parse run (process). If a rdf:nodeID is encountered the system is simply copying the bNode identifier through e.g. rdf:nodeID="blue" will remain unchanged as 'blue' and so on.

NOTE: The 'manage_bNodes' could be called several times for the same 'SystembNode'

If this handler is undefined the system behaves by default as outline above - otherwise the end-application can interect with this process.

By using this handler the end-application can control how to generate identfiers for anonymous resources OR how to re-write specific bNodes as normal URI qualified resources. The end-application will get triggered 'manage_bNodes' events for each new (system generated) bNode or when a given rdf:nodeID attribute is found into the input RDF/XML source. The system wide generated 'SystembNode' identifier is also passed to the handler code. As already pointed out, multiple events could result for the same 'SystembNode'.

The 'SystembNode' parameter will either contain the bare bone system generated identifier like 'genidrdfstoreS302e313439323736363337373935353039P2233T1069672550N45' or the rdf:nodeID like 'blue' - it is recommended to the end-application to keep the 'genidrdfstore' prefix for sequential generated identifiers. This will allow in the future to immediately distinguish bNodes generated by the RDF/XML parser from rdf:nodeID or application specific ones.

By using the 'manage_bNodes' event, for example, the application could keep track of system unique (and/or sequential) identifiers for bNodes internally or re-write a given bNode.

The end-application must use the 'Factory' parameter (which should correspond to the 'NodeFactory' parameter passed to the parser constructor) to return to the parser (caller) the corresponding RDFStore::RDFNode(3) to use in place of the specific generated event.

For example, the handler could keep a kind of look-up table of system generated bNodes or input source rdf:nodeID to application specific URIs. In which case the end-application would rewrite input anonymous resource to valid world-wide unique resources.

Here are three examples - the first is simply passing/delegating to the parser the generation of an anonymous resource:

 sub manage_bNodes {
        return $_[1]->createAnonymousResource( $_[2] ); #does really nothing...pass through

The second example rewrite system wide generated bNodes to application specific bNodes:

 sub manage_bNodes {
        my ($expatm, $factory, $systemid) = @_;

        $systemid =~ s/^genidrdfstore/genidmyapplication/;

        return $factory->createAnonymousResource( $systemid );

The last example re-write parser generated and rdf:nodeID bNodes to an application specific URI list:

 my %app_uri_map = (
        'genidrdfstoreS302e313439323736363337373935353039P2233T1069672550N45' => '',
        'alberto' => '',
        'zac' => '',
        'dirkx' => ''

 sub manage_bNodes {
        my ($expatm, $factory, $systemid) = @_;

        return $factory->createResource( $app_uri_map{$systemid} );

This handler must return a valid RDFStore::Resource(3) node.


Write an extension module for you needs it is as easy as write one for XML::Parser :) Have a look at and

You can either make you Perl script a parser self by embedding the needed function hooks or write a custom Style module for RDFStore::Parser::SiRPAC.

*.pl scripts

        use RDFStore::Parser::SiRPAC;
        use RDFStore::NodeFactory;
        my $p=new RDFStore::Parser::SiRPAC(
                Handlers        => {
                        Init    => sub { print "INIT\n"; },
                        Final   => sub { print "FINAL\n"; },
                        Assert  => sub { print "STATEMENT - @_\n"; }
                NodeFactory     => new RDFStore::NodeFactory() );

or something like:

        use RDFStore::Parser::SiRPAC;
        use RDFStore::NodeFactory;
        my $p=new RDFStore::Parser::SiRPAC( NodeFactory     => new RDFStore::NodeFactory() );
        $p->setHandlers(        Init    => sub { print "INIT\n"; },
                                Final   => sub { print "FINAL\n"; },
                                Assert  => sub { print join(",",@_),"\n"; }     );

Style modules

A more sophisticated solution is to write a complete Perl5 Sytle module for RDFStore::Parser::SiRPAC that can be easily reused in your code. E.g. a perl script could use this piece of code:

        use RDFStore::Parser::SiRPAC;
        use RDFStore::Parser::SiRPAC::MyStyle;
        use RDFStore::NodeFactory;

        my $p=new RDFStore::Parser::SiRPAC(     Style => 'RDFStore::Parser::SiRPAC::MyStyle',
                                        NodeFactory     => new RDFStore::NodeFactory() );

The Style module self could stored into a file like like this:

        package RDFStore::Parser::SiRPAC::MyStyle;

        sub Init { print "INIT\n"; };
        sub Final { print "FINAL\n"; };
        sub Assert {
                print "ASSERT: ",
                                $_[1]->object()->toString(), "\n";
        sub Start_XML_Literal { print "STARTAG: ",$_[1],"\n"; };
        sub Stop_XML_Literal { print "ENDTAG: ",$_[1],"\n"; };
        sub Char_XML_Literal { print "UTF8 chrs: ",$_[1],"\n"; };


For a more complete and useful example see RDFStore::Parser::SiRPAC::RDFStore(3).


This module implements most of the W3C RDF Raccomandation as its Java counterpart SiRPAC from the Stanford University Database Group by Sergey Melnik (see This version is conformant to the latest RDF API Draft on 2000-11-13. It does not support yet:

        * aboutEach


 RDFStore::Parser::SiRPAC(3), DBMS(3) and XML::Parser(3) XML::Parser::Expat(3)

 RDFStore::Model(3) RDFStore::NodeFactory(3)

 RDF Model and Syntax Specification -

 RDF Schema Specification 1.0 -

 Benchmarking XML Parsers by Clark Cooper -

 See also

 RDF::Parser(3) from


        Alberto Reggiori <>

        Sergey Melnik <> is the original author of the streaming version of SiRPAC in Java
        Clark Cooper is the author of the XML::Parser(3) module together with Larry Wall