Did you know ... Search Documentation:
Pack prolog_rdf -- README.md

Prolog-based RDF library

This library provides advanced support for working with RDF in Prolog.


Install SWI-Prolog.


Install this library:

swipl -g 'pack_install(prolog_rdf)' -t halt


Libraries can be loaded in the following way:

:- [library(rdf_term)].

RDF-specific Prolog types

This library uses the following extended Prolog types in the documentation headers of predicates:

rdf_bnodeAn atom that starts with `_:`.
rdf_graphEither a term of type iri or the atom default.
rdf_iriAn atom that can be decomposed with uri_components/2 from library(uri).
rdf_literalA compound term of the form literal(type(iri,atom)) or literal(lang(atom,atom)).
rdf_nameAn RDF name (IRI or literal).
rdf_quadA compound term of the form rdf(rdf_nonliteral,iri,rdf_term,rdf_graph).
rdf_termAn RDF term (blank node, IRI, or literal).
rdf_tripleA compound term of the form rdf(rdf_nonliteral,iri,rdf_term).
rdf_tupleA term of type rdf_quad or rdf_triple.


This section enumerates the various modules that are included in this library.


This module contains data cleaning predicates that were previously part of LOD Laundromat. They can be used to clean RDF tuples that are streamed from an RDF source. See module [[semweb/rdf_deref]] for creating streams over RDF sources.

In order to use this module, library prolog_uriparser must be installed.

Blank node cleaning

The parsers in the Semantic Web standard library emit blank node labels that contain characters that are not allowed in standards-compliant output formats (e.g., forward slashes). This is unfortunate, since writing the data into standard-compliant formats requires maintaining a state that ensures that Prolog internal blank node labels are consistently emitted by the same standard-compliant external blank node label. See this Github issue for context.

Besides the above considerations, blank nodes form a scalability issue in general. Since blank node labels are only guaranteed to be unique within the context of an RDF document, combining data from multiple documents requires a check of all blank node labels in the to be combined documents. Furthermore, all blank node labels that appear in more than one RDF document must be consistently renamed prior to combining the data.

Since Pro-RDF focusses on scalability, it cannot rely on maintaining an internal state that consistently maps internal Prolog blank node labels to external standards-compliant blank node labels. For the same reasons, it also cannot rely on full document inspection and blank node relabeling approaches. For these reasons, the data cleaning prediates in library(rdf_clean) replace blank nodes with well-known IRIs, in line with the RDF 1.1 standard. This means that every data cleaning predicate must bind a valid well-known IRI to the BNodePrefix argument. It also means that Prolog internal blank node labels are hashed using the MD5 algorithm to provide the local names for the generated well-known IRIs. The latter ensures consistent relabeling without maintaining an internal state.

Graph cleaning

The parsers from the Semantic Web standard library denote the default graph with atom user. This is translated to atom default. For named graphs, this library checks whether they are well-formed IRIs.

IRI cleaning

IRI cleaning is the most difficult part of syntactic RDF data cleaning. To date, the IRI grammar (RFC 3987) has not yet been implemented. Since this grammar was published over a decade ago, we must anticipate a future in which the main syntactic component of the Semantic Web cannot be validated.

While there are implementations of the URI grammar (RFC 3986), the one provided by the SWI-Prolog standard library (library(uri)) is incorrect.

Because of the above two reasons we currently only check the following:

  • Whether an IRI can be decomposed into scheme, authority, path, query, and fragment components using the Prolog standard library grammar (uri_components/2).
  • Whether the scheme, authority, and path components are non-empty.
  • Whether the scheme components conforms to the IRI grammar.

Literal cleaning

For language-tagged strings, cleaning involves downcasing the language tag. While there are implementations of the language tag grammar (RFC 5646), we are not yet using these.

Simple literals, i.e., literals with neither language tag not datatype IRI, are translated to typed literals with datatype IRI xsd:string.

For typed literals, cleaning involves:

  • Cleaning the datatype IRI (see [[IRI cleaning]]).
  • Making sure the datatype IRI is not rdf:langString.
  • Cleaning the lexical form according to the datatype IRI. Lexical form cleaning is the most involved step, since there are many different datatype IRIs. Since it is impractical to implement lexical form cleaning for all datatype IRIs, we focus on those that are most widely used. For this we use rdf_literal_value/3, which is part of library library(semweb/rdf_term).


This module provides the following predicates.

##### rdf_clean_quad(+Site:uri, +Dirty:rdf_quad, -Clean:rdf_quad)

Cleans quadruple compound terms.

##### rdf_clean_triple(+Site:uri, +Dirty:rdf_triple, -Clean:rdf_triple)

Cleans triple compound terms.

##### rdf_clean_tuple(+Site:uri, +Dirty:rdf_tuple, -Clean:rdf_tuple)

Cleans quadruple and/or triple compound terms.


This module implements RDF dereferencing, i.e., the act of obtaining interpreted RDF statements based on a given RDF document, stream, or HTTP(S) URI.


This library provides the following predicates.

##### rdf_deref_file/[2,3]

Calls RDF dereferencing on local RDF documents. Uses heuristics in order to determine the RDF serialization of the file.

##### rdf_deref_stream/[3,4]

Performs RDF dereferencing on an input stream containing one of the standardized RDF serialization formats.

##### rdf_dered_uri/[2,3]

Performs RDF dereferencing on a URI, typically an HTTP(S) URI. Uses heuristics in order to determine the RDF serialization of the reply body.


This library provides primitives for generating GraphViz DOT exports of RDF terms and tuples.

This module requires library prolog_graphviz to be installed.


This module writes RDF data in a simple and standards-compliant serialization format. It contains the following predicates:


This module peeks at the beginning of a file, stream, or string in order to heuristically guesstimate the RDF serialization formats (if any) containing in that input:


This module provides support for the standardized RDF serialization format Media Types:


Guesses the RDF serialization format based on the file name extension alone.


Enumerates all standardized RDF Media Types.


Succeeds if the former argument is an RDF Media Type that syntactically encompasses the latter argument (e.g., TriG > Turtle > N-Triples, N-Quads > N-Triples).


Gives a standard file name extension for RDF serializations that are not RDFa (which is part of HTML or XHTML content).


Succeeds for RDFa Media Types.


This module provides extended support for working with RDF prefix declarations:

##### rdf_prefix/[1,2]

Enumerates the currently declared RDF prefix declarations.




Succeeds for (alias,local-name) pairs and full IRIs.




Provide the corresponding popular Prolog predicates, but apply RDF prefix notation expansion on their arguments.

RDF prefix expansion must be specifically declared for arguments in predicates. In the SWI-Prolog standard libraries, such declarations have only been added for predicates in the Semantic Web libraries, but not for predicates in other standard libraries. For example, the following will not check whether P is bound to either of the four RDFS properties, because the prefix notation is not expanded:

memberchk(P, [rdfs:domain,rdfs:range,rdfs:subClassOf,rdfs:subPropertyOf]),

With the SWI-Prolog standard library, the above call must be spelled out using rdf_equal/2 in the following way:

(   rdf_equal(P, rdfs:domain)
->  true
;   rdf_equal(P, rdfs:range)
->  true
;   rdf_equal(P, rdfs:subClassOf)
->  true
;   rdf_equal(P, rdfs:subPropertyOf)
->  true

When library(rdf_prefix) is loaded, the above can be written as follows:

rdf_prefix_memberchk(P, [rdfs:domain,rdfs:range,rdfs:subClassOf,rdfs:subPropertyOf]),






This module provides DCG rules for printing RDF terms and tuples.


This module provides advanced support for composing, decomposing, parsing, and generating RDF terms.