\documentclass[11pt]{article}
\usepackage{times}
\usepackage{pl}
\usepackage{plpage}
\usepackage{alltt}
\usepackage{html}
\usepackage{verbatim}
\sloppy
\makeindex

\onefile
\htmloutput{.}			% Output directory
\htmlmainfile{semweb}		% Main document file
\bodycolor{white}		% Page colour

\renewcommand{\runningtitle}{SWI-Prolog Semantic Web Library (version 3)}

\newcommand{\elem}[1]{{\tt\string<#1\string>}}

\begin{document}

\title{SWI-Prolog Semantic Web Library 3.0}
\author{Jan Wielemaker \\
	University of Amsterdam/VU University Amsterdam \\
	The Netherlands \\
	E-mail: \email{J.Wielemaker@vu.nl}}

\maketitle

\begin{abstract}

This document describes the SWI-Prolog \jargon{semweb} package. The core
of this package is an efficient main-memory based RDF store that is
tightly connected to Prolog. Additional libraries provide reading and
writing RDF/XML and Turtle data, caching loaded RDF documents and
persistent storage. This package is the core of a ready-to-run platform
for developing Semantic Web applications named
\href{http://cliopatria.swi-prolog.org}{ClioPatria}, which is distributed
separately. The SWI-Prolog RDF store is among the most memory efficient
main-memory stores for
RDF\footnote{\url{http://cliopatria.swi-prolog.org/help/source/doc/home/vnc/prolog/src/ClioPatria/web/help/memusage.txt}}

Version~3 of the RDF library enhances concurrent use of the library by
allowing for lock-free reading and writing using short-held locks.  It
provides Prolog compatible \jargon{logical update view} on the triple
store and isolation using \jargon{transactions} and \jargon{snapshots}.
This version of the library provides near real-time modification and
querying of RDF graphs, making it particularly interesting for handling
streaming RDF and graph manipulation tasks.
\end{abstract}

\vfill
\pagebreak
\tableofcontents

\newpage


\section{Introduction}
\label{sec:semweb-intro}

The core of the SWI-Prolog package \file{semweb} is an efficient
main-memory RDF store written in C that is tightly integrated with
Prolog. It provides a fully logical predicate rdf/3 to query the RDF
store efficiently by using multiple (currently 9) indexes. In addition,
SWI-Prolog provides libraries for reading and writing XML/RDF and Turtle
and a library that provides persistency using a combination of efficient
binary snapshots and journals.

Below, we describe a few usage scenarios that guides the current design
of this Prolog-based RDF store.


\paragraph{Application prototyping platform}

Bundled with \href{http://cliopatria.swi-prolog.org}{ClioPatria}, the
store is an efficient platform for prototyping a wide range of semantic
web applications. Prolog, connected to the main-memory based store is a
productive platform for writing application logic that can be made
available through the SPARQL endpoint of ClioPatria, using an
application specific API (typically based on JSON or XML) or as an HTML
based end-user application. Prolog is more versatile than SPARQL, allows
composing of the logic from small building blocks and does not suffer
from the \jargon{Object-relational impedance mismatch}.


\paragraph{Data integration}

The SWI-Prolog store is optimized for entailment on the
\const{rdfs:subPropertyOf} relation. The \const{rdfs:subPropertyOf}
relation is crucial for integrating data from multiple sources while
preserving the original richness of the sources because integration can
be achieved by defining the native properties as sub-properties of
properties from a unifying schema such as Dublin Core.


\paragraph{Dynamic data}

This RDF store is one of the few stores that is primarily based on
\jargon{backward reasoning}. The big advantage of backward reasoning is
that it can much easier deal with changes to the database because it
does not have to take care of propagating the consequences. Backward
reasoning reduces storage requirements. The price is more reasoning
during querying. In many scenarios the extra reasoning using a main
memory will outperform the fetching the precomputed results from
external storage.


\paragraph{Prototyping reasoning systems}

Reasoning systems, not necessarily limited to entailment reasoning, can
be prototyped efficiently on the Prolog based store. This includes
`what-if' reasoning, which is supported by \jargon{snapshot} and
\jargon{transaction} isolation. These features, together with the
concurrent loading capabilities, make the platform well equiped to
collect relevant data from large external stores for intensive
reasoning. Finally, the
\href{http://www.swi-prolog.org/pldoc/package/tipc.html}{TIPC} package
can be used to create networks of cooperating RDF based agents.


\paragraph{Streaming RDF}

Transactions, snapshots, concurrent modifications and the database
monitoring facilities (see rdf_monitor/2) make the platform well suited
for prototyping systems that deal with streaming RDF data.


\section{Scalability}
\label{sec:semweb-scalability}

Depending on the OS and further application restrictions, the SWI-Prolog
RDF stores scales to about 15~million triples on 32-bit hardware. On
64-bit hardware, the scalability is limited by the amount of physical
memory, allowing for approximately 4~million triples per gigabyte. The
other limiting factor for practical use is the time required to load
data and/or restore the database from the persistent file backup.
Performance depends highly on hardware, concurrent performance and
whether or not the data is spread over multiple (named) graphs that can
be loaded in parallel. Restoring over 20~million triples per minute is
feasible on medium hardware (Intel i7/{2600} running Ubuntu 12.10).


		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
		 %	     RDF API		%
		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\section{Two RDF APIs}
\label{sec:semweb-rdfapi}

The current `semweb' package provides two sets of interface predicates.
The original set is described in \secref{semweb-rdf-db}. The new API is
described in \secref{semweb-rdf11}. The original API was designed when
RDF was not yet standardised and did not yet support data types and
language indicators. The new API is designed from the RDF 1.1
specification, introducing consistent naming and access to literals
using the \jargon{value space}. The new API is currently defined on top
of the old API, so both APIs can be mixed in a single application.


		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
		 %	      RDF11		%
		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\subsection{library(semweb/rdf11): The RDF database}
\label{sec:semweb-rdf11}

\InputIfFileExists{rdf11}{}{}
\InputIfFileExists{rdf11containers}{}{}


		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
		 %	      RDF_DB		%
		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


\subsection{library(semweb/rdf_db): The RDF database}
\label{sec:semweb-rdf-db}

\InputIfFileExists{rdfdb}{}{}


\subsection{Monitoring the database}		\label{sec:rdfmonitor}

The predicate rdf_monitor/2 allows registrations of call-backs with the
RDF store. These call-backs are typically used to keep other databases
in sync with the RDF store. For example,
\pllib{library(semweb/rdf_persistency)} monitors the RDF store for
maintaining a persistent copy in a set of files and
\pllib{library(semweb/rdf_litindex)} uses added and deleted literal
values to maintain a fulltext index of literals.

\begin{description}
    \predicate{rdf_monitor}{2}{:Goal, +Mask}
\arg{Goal} is called for modifications of the database. It is called
with a single argument that describes the modification. Defined
events are:

\begin{description}
    \termitem{assert}{+S, +P, +O, +DB}
A triple has been asserted.
    \termitem{retract}{+S, +P, +O, +DB}
A triple has been deleted.
    \termitem{update}{+S, +P, +O, +DB, +Action}
A triple has been updated.
    \termitem{new_literal}{+Literal}
A new literal has been created.  \arg{Literal} is the argument of
\term{literal}{Arg} of the triple's object.  This event is introduced
in version 2.5.0 of this library.
    \termitem{old_literal}{+Literal}
The literal \arg{Literal} is no longer used by any triple.
    \termitem{transaction}{+BeginOrEnd, +Id}
Mark begin or end of the \emph{commit} of a transaction started by
rdf_transaction/2. \arg{BeginOrEnd} is \term{begin}{Nesting} or
\term{end}{Nesting}. \arg{Nesting} expresses the nesting level of
transactions, starting at `0' for a toplevel transaction. \arg{Id} is
the second argument of rdf_transaction/2. The following transaction Ids
are pre-defined by the library:

    \begin{description}
	\termitem{parse}{Id}
A file is loaded using rdf_load/2.  \arg{Id} is one of \term{file}{Path}
or \term{stream}{Stream}.
	\termitem{unload}{DB}
All triples with source \arg{DB} are being unloaded using rdf_unload/1.
	\termitem{reset}{}
Issued by rdf_reset_db/0.
    \end{description}

    \termitem{load}{+BeginOrEnd, +Spec}
Mark begin or end of rdf_load_db/1 or load through rdf_load/2 from
a cached file. \arg{Spec} is currently defined as \term{file}{Path}.
    \termitem{rehash}{+BeginOrEnd}
Marks begin/end of a re-hash due to required re-indexing or garbage
collection.
\end{description}

\arg{Mask} is a list of events this monitor is interested in. Default
(empty list) is to report all events. Otherwise each element is of the
form +Event or -Event to include or exclude monitoring for certain
events. The event-names are the functor names of the events described
above. The special name \const{all} refers to all events and
\term{assert}{load} to assert events originating from rdf_load_db/1. As
loading triples using rdf_load_db/1 is very fast, monitoring this at the
triple level may seriously harm performance.

This predicate is intended to maintain derived data, such as a journal,
information for \emph{undo}, additional indexing in literals, etc. There
is no way to remove registered monitors. If this is required one should
register a monitor that maintains a dynamic list of subscribers like the
XPCE broadcast library. A second subscription of the same hook predicate
only re-assignes the mask.

The monitor hooks are called in the order of registration and in the
same thread that issued the database manipulation. To process all
changes in one thread they should be send to a thread message queue. For
all updating events, the monitor is called while the calling thread has
a write lock on the RDF store. This implies that these events are
processed strickly synchronous, even if modifications originate from
multiple threads. In particular, the \const{transaction} \emph{begin},
\ldots{} \emph{updates} \ldots{} \emph{end} sequence is never
interleaved with other events. Same for \const{load} and \const{parse}.
\end{description}


\subsection{Issues with rdf_db}				\label{sec:rdfissues}

This RDF low-level module has been created after two year experimenting
with a plain Prolog based module and a brief evaluation of a second
generation pure Prolog implementation. The aim was to be able to handle
upto about 5 million triples on standard (notebook) hardware and deal
efficiently with \const{subPropertyOf} which was identified as a crucial
feature of RDFS to realise fusion of different data-sets.

The following issues are identified and not solved in suitable manner.

\begin{description}
    \item [\const{subPropertyOf} of \const{subPropertyOf}] is not
supported.

    \item [Equivalence]
Similar to \const{subPropertyOf}, it is likely to be profitable to
handle resource identity efficient.  The current system has no support
for it.
\end{description}


		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
		 %	     PLUGIN		%
		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\section{Plugin modules for rdf_db}
\label{sec:plugin}

The \pllib{rdf_db} module provides several hooks for extending its
functionality. Database updates can be monitored and acted upon through
the features described in \secref{rdfmonitor}. The predicate rdf_load/2
can be hooked to deal with different formats such as \jargon{rdfturtle},
different input sources (e.g.\ http) and different strategies for
caching results.

\subsection{Hooks into the RDF library}
\label{sec:semweb-hooks}

The hooks below are used to add new RDF file formats and sources from
which to load data to the library. They are used by the modules
described below and distributed with the package.  Please examine the
source-code if you want to add new formats or locations.

\begin{description}
    \item[\pllib{library(semweb/turtle)}]
Load files in the Turtle format.  See \secref{turtle}.
    \item[\pllib{library(semweb/rdf_zlib_plugin)}]
Load \program{gzip} compressed files transparently. See
\secref{semweb-zlib}.
    \item[\pllib{library(semweb/rdf_http_plugin)}]
Load RDF documents from HTTP servers.  See \secref{http}.
    \item[\pllib{library(http/http_ssl_plugin)}]
May be combined with \pllib{library(semweb/rdf_http_plugin)} to load
RDF from HTTPS servers.
    \item[\pllib{library(semweb/rdf_persistency)}]
Provide persistent backup of the triple store.
    \item[\pllib{library(semweb/rdf_cache)}]
Provide caching RDF sources using fast load/safe files to speedup
restarting an application.
\end{description}

\begin{description}
    \predicate{rdf_db:rdf_open_hook}{3}{+Input, -Stream, -Format}
Open an input. \arg{Input} is one of \term{file}{+Name},
\term{stream}{+Stream} or \term{url}{Protocol, URL}.  If this hook
succeeds, the RDF will be read from Stream using rdf_load_stream/3.
Otherwise the default open functionality for file and stream are
used.

    \predicate{rdf_db:rdf_load_stream}{3}{+Format, +Stream, +Options}
Actually load the RDF from \arg{Stream} into the RDF database.
\arg{Format} describes the format and is produced either by
rdf_input_info/3 or rdf_file_type/2.

    \predicate{rdf_db:rdf_input_info}{3}{+Input, -Modified, -Format}
Gather information on \arg{Input}. \arg{Modified} is the last
modification time of the source as a POSIX time-stamp (see time_file/2).
\arg{Format} is the RDF format of the file.  See rdf_file_type/2 for
details. It is allowed to leave the output variables unbound. Ultimately
the default modified time is `0' and the format is assumed to be
\const{xml}.

    \predicate{rdf_db:rdf_file_type}{2}{?Extension, ?Format}
True if \arg{Format} is the default RDF file format for files
with the given extension.  \arg{Extension} is lowercase and
without a '.'.  E.g.\ \const{owl}.  \arg{Format} is either a
built-in format (\const{xml} or \const{triples}) or a format
understood by the rdf_load_stream/3 hook.

    \predicate{rdf_db:url_protocol}{1}{?Protocol}
True if \arg{Protocol} is a URL protocol recognised by rdf_load/2.
\end{description}


\subsection{library(semweb/rdf_zlib_plugin): Reading compressed RDF}
\label{sec:semweb-zlib}

\index{gz, format}\index{gzip}\index{compressed data}%
This module uses the \pllib{zlib} library to load compressed files
on the fly.  The extension of the file must be \fileext{gz}.  The
file format is deduced by the extension after stripping the \fileext{gz}
extension.  E.g.\ \exam{rdf_load('file.rdf.gz')}.


\subsection{library(semweb/rdf_http_plugin): Reading RDF from a HTTP server}
\label{sec:http}

\index{xhtml}%
This module allows for \exam{rdf_load('http://...')}. It exploits the
library \pllib{http/http_open.pl}. The format of the URL is determined
from the mime-type returned by the server if this is one of
\const{text/rdf+xml}, \const{application/x-turtle} or
\const{application/turtle}. As RDF mime-types are not yet widely
supported, the plugin uses the extension of the URL if the claimed
mime-type is not one of the above.  In addition, it recognises
\const{text/html} and \const{application/xhtml+xml}, scanning
the XML content for embedded RDF.

\InputIfFileExists{rdfcache.tex}{}{}



		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
		 %	     LITINDEX		%
		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\subsection{library(semweb/rdf_litindex): Indexing words in literals}
\label{sec:rdflitindex}

The library \pllib{semweb/rdf_litindex.pl} exploits the primitives
of \secref{rdflitmap} and the NLP package to provide indexing on words
inside literal constants.  It also allows for fuzzy matching using
stemming and `sounds-like' based on the \jargon{double metaphone}
algorithm of the NLP package.

\begin{description}
    \predicate{rdf_find_literals}{2}{+Spec, -ListOfLiterals}
Find literals (without type or language specification) that satisfy
\arg{Spec}. The required indices are created as needed and kept
up-to-date using hooks registered with rdf_monitor/2.  Numerical
indexing is currently limited to integers in the range $\pm 2^30$
($\pm 2^62 on 64-bit platforms$).  \arg{Spec} is defined as:

    \begin{description}
	\termitem{and}{Spec1, Spec2}
Intersection of both specifications.

	\termitem{or}{Spec1, Spec2}
Union of both specifications.

	\termitem{not}{Spec}
Negation of \arg{Spec}.  After translation of the full specification to
\jargon{Disjunctive Normal Form} (DNF), negations are only allowed
inside a conjunction with at least one positive literal.

	\termitem{case}{Word}
Matches all literals containing the word \arg{Word}, doing the match
case insensitive and after removing diacritics.

	\termitem{stem}{Like}
Matches all literals containing at least one word that has the same stem
as \arg{Like} using the Porter stem algorithm.  See NLP package for
details.

	\termitem{sounds}{Like}
Matches all literals containing at least one word that `sounds like'
\arg{Like} using the double metaphone algorithm.  See NLP package for
details.

	\termitem{prefix}{Prefix}
Matches all literals containing at least one word that starts with
Prefix, discarding diacritics and case.

	\termitem{between}{Low, High}
Matches all literals containing an integer token in the range
\arg{Low}..\arg{High}, including the boundaries.

	\termitem{ge}{Low}
Matches all literals containing an integer token with value
\arg{Low} or higher.

	\termitem{le}{High}
Matches all literals containing an integer token with value
\arg{High} or lower.

	\termitem{Token}{}
Matches all literals containing the given token.  See tokenize_atom/2
of the NLP package for details.
    \end{description}

    \predicate{rdf_token_expansions}{2}{+Spec, -Expansions}
Uses the same database as rdf_find_literals/2 to find possible
expansions of \arg{Spec}, i.e.\ which words `sound like', `have prefix',
etc. \arg{Spec} is a compound expression as in rdf_find_literals/2.
\arg{Expansions} is unified to a list of terms \term{sounds}{Like,
Words}, \term{stem}{Like, Words} or \term{prefix}{Prefix, Words}. On
compound expressions, only combinations that provide literals are
returned.  Below is an example after loading the ULAN%
	\footnote{Unified List of Artist Names from the Getty
		  Foundation.}
database and showing all words that sounds like `rembrandt' and
appear together in a literal with the word `Rijn'.  Finding this
result from the 228,710 literals contained in ULAN requires 0.54
milliseconds (AMD 1600+).

\begin{code}
?- rdf_token_expansions(and('Rijn', sounds(rembrandt)), L).

L = [sounds(rembrandt, ['Rambrandt', 'Reimbrant', 'Rembradt',
			'Rembrand', 'Rembrandt', 'Rembrandtsz',
			'Rembrant', 'Rembrants', 'Rijmbrand'])]
\end{code}

Here is another example, illustrating handling of diacritics:

\begin{quote}\begin{alltt}
?- rdf_token_expansions(case(cafe), L).

L = [case(cafe, [cafe, caf\'e])]
\end{alltt}\end{quote}

    \predicate{rdf_tokenize_literal}{2}{+Literal, -Tokens}
Tokenize a literal, returning a list of atoms and integers in the range
$-1073741824 \ldots 1073741823$. As tokenization is in general domain
and task-dependent this predicate first calls the hook
\term{rdf_litindex:tokenization}{Literal, -Tokens}. On failure it
calls tokenize_atom/2 from the NLP package and deletes the following:
atoms of length 1, floats, integers that are out of range and the
english words \const{and}, \const{an}, \const{or}, \const{of},
\const{on}, \const{in}, \const{this} and \const{the}.  Deletion first
calls the hook \term{rdf_litindex:exclude_from_index}{token, X}.  This
hook is called as follows:

\begin{code}
no_index_token(X) :-
	exclude_from_index(token, X), !.
no_index_token(X) :-
	...
\end{code}
\end{description}

\subsubsection{Literal maps: Creating additional indices on literals}
\label{sec:rdflitmap}

`Literal maps' provide a relation between literal values, intended to
create additional indexes on literals. The current implementation can
only deal with integers and atoms (string literals). A literal map
maintains an ordered set of \jargon{keys}. The ordering uses the same
rules as described in \secref{rdflitindex}. Each key is associated with
an ordered set of \jargon{values}. Literal map objects can be shared
between threads, using a locking strategy that allows for multiple
concurrent readers.

Typically, this module is used together with rdf_monitor/2 on the
channals \const{new_literal} and \const{old_literal} to maintain an
index of words that appear in a literal. Further abstraction using
Porter stemming or Metaphone can be used to create additional search
indices. These can map either directly to the literal values, or
indirectly to the plain word-map. The SWI-Prolog NLP package provides
complimentary building blocks, such as a tokenizer, Porter stem and
Double Metaphone.


\begin{description}
    \predicate{rdf_new_literal_map}{1}{-Map}
Create a new literal map, returning an opaque handle.

    \predicate{rdf_destroy_literal_map}{1}{+Map}
Destroy a literal map. After this call, further use of the \arg{Map}
handle  is illegal.  Additional synchronisation is needed if maps that
are shared between threads are destroyed to guarantee the handle is
no longer used.  In some scenarios rdf_reset_literal_map/1
provides a safe alternative.

    \predicate{rdf_reset_literal_map}{1}{+Map}
Delete all content from the literal map.

    \predicate{rdf_insert_literal_map}{3}{+Map, +Key, +Value}
Add a relation between \arg{Key} and \arg{Value} to the map.  If
this relation already exists no action is performed.

    \predicate{rdf_insert_literal_map}{4}{+Map, +Key, +Value, -KeyCount}
As rdf_insert_literal_map/3.  In addition, if \arg{Key} is a new key in
\arg{Map}, unify \arg{KeyCount} with the number of keys in \arg{Map}.
This serves two purposes.  Derived maps, such as the stem and metaphone
maps need to know about new keys and it avoids additional foreign calls
for doing the progress in \file{rdf_litindex.pl}.

    \predicate{rdf_delete_literal_map}{2}{+Map, +Key}
Delete \arg{Key} and all associated values from the map.  Succeeds
always.

    \predicate{rdf_delete_literal_map}{2}{+Map, +Key, +Value}
Delete the association between \arg{Key} and \arg{Value} from the map.
Succeeds always.

    \predicate[det]{rdf_find_literal_map}{3}{+Map, +KeyList, -ValueList}
Unify \arg{ValueList} with an ordered set of values associated to
all keys from \arg{KeyList}.  Each key in \arg{KeyList} is either an
atom, an integer or a term \term{not}{Key}.  If not-terms are provided,
there must be at least one positive keywords.  The negations are tested
after establishing the positive matches.

    \predicate{rdf_keys_in_literal_map}{3}{+Map, +Spec, -Answer}
Realises various queries on the key-set:

    \begin{description}
	\termitem{all}{}
Unify \arg{Answer} with an ordered list of all keys.

	\termitem{key}{+Key}
Succeeds if \arg{Key} is a key in the map and unify \arg{Answer}
with the number of values associated with the key.  This provides
a fast test of existence without fetching the possibly large associated
value set as with rdf_find_literal_map/3.

	\termitem{prefix}{+Prefix}
Unify \arg{Answer} with an ordered set of all keys that have the given
prefix. \arg{Prefix} must be an atom. This call is intended for
auto-completion in user interfaces.

	\termitem{ge}{+Min}
Unify \arg{Answer} with all keys that are larger or equal to the
integer \arg{Min}.

	\termitem{le}{+Max}
Unify \arg{Answer} with all keys that are smaller or equal to the
integer \arg{Max}.


	\termitem{between}{+Min, +Max}
Unify \arg{Answer} with all keys between \arg{Min} and \arg{Max}
(including).
    \end{description}

    \predicate{rdf_statistics_literal_map}{2}{+Map, +Key(-Arg...)}
Query some statistics of the map.  Provides keys are:
    \begin{description}
	\termitem{size}{-Keys, -Relations}
Unify \arg{Keys} with the total key-count of the index and
\arg{Relation} with the total \arg{Key}-\arg{Value} count.
    \end{description}
\end{description}


		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
		 %	    PERSISTENCY		%
		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\subsection{library(semweb/rdf_persistency): Providing persistent storage}
\label{sec:rdfpersistency}

\index{Persistent store}%
The \pllib{semweb/rdf_persistency} provides reliable persistent storage
for the RDF data. The store uses a directory with files for each source
(see rdf_source/1) present in the database. Each source is represented
by two files, one in binary format (see rdf_save_db/2) representing the
base state and one represented as Prolog terms representing the changes
made since the base state. The latter is called the \jargon{journal}.


\begin{description}
    \predicate{rdf_attach_db}{2}{+Directory, +Options}
Attach \arg{Directory} as the persistent database. If \arg{Directory}
does not exist it is created. Otherwise all sources defined in the
directory are loaded into the RDF database. Loading a source means
loading the base state (if any) and replaying the journal (if any). The
current implementation does not synchronise triples that are in the
store before attaching a database. They are not removed from the
database, nor added to the presistent store. Different merging options
may be supported through the \arg{Options} argument later.  Currently
defined options are:

    \begin{description}
	\termitem{concurrency}{+PosInt}
Number of threads used to reload databased and journals from the
files in \arg{Directory}.  Default is the number of physical CPUs
determined by the Prolog flag \const{cpu_count} or 1 (one) on
systems where this number is unknown.  See also concurrent/3.

	\termitem{max_open_journals}{+PosInt}
The library maintains a	pool of open journal files.  This option
specifies the size of this pool.  The default is 10.  Raising the
option can make sense if many writes occur on many different named
graphs. The value can be lowered for scenarios where write operations
are very infrequent.

	\termitem{silent}{Boolean}
If \const{true}, supress loading messages from rdf_attach_db/2.

	\termitem{log_nested_transactions}{Boolean}
If \const{true}, nested \emph{log} transactions are added to the journal
information. By default (\const{false}), no log-term is added for nested
transactions.
    \end{description}

The database is locked against concurrent access using a file
\file{lock} in \arg{Directory}. An attempt to attach to a locked
database raises a \const{permission_error} exception.  The error
context contains a term \term{rdf_locked}{Args}, where args is
a list containing \term{time}{Stamp} and \term{pid}{PID}.  The
error can be caught by the application.  Otherwise it prints:

\begin{code}
ERROR: No permission to lock rdf_db `/home/jan/src/pl/packages/semweb/DB'
ERROR: locked at Wed Jun 27 15:37:35 2007 by process id 1748
\end{code}

    \predicate{rdf_detach_db}{0}{}
Detaches the persistent store.  No triples are removed from the RDF
triple store.

    \predicate{rdf_current_db}{1}{-Directory}
Unify \arg{Directory} with the current database directory.  Fails if no
persistent database is attached.

    \predicate{rdf_persistency}{2}{+DB, +Bool}
Change presistency of named database (4th argument of rdf/4). By default
all databases are presistent. Using \const{false}, the journal and
snapshot for the database are deleted and further changes to triples
associated with \arg{DB} are not recorded. If \arg{Bool} is \const{true}
a snapshot is created for the current state and further modifications
are monitored.  Switching persistency does not affect the triples in the
in-memory RDF database.

    \predicate{rdf_flush_journals}{1}{+Options}
Flush dirty journals.  With the option \term{min_size}{KB} only journals
larger than \arg{KB} Kbytes are merged with the base state.  Flushing a
journal takes the following steps, ensuring a stable state can be
recovered at any moment.
    \begin{enumerate}
        \item	Save the current database in a new file using the
		extension \fileext{new}.
	\item	On success, delete the journal
	\item	On success, atomically move the \fileext{new} file
		over the base state.
    \end{enumerate}

Note that journals are \emph{not} merged automatically for two reasons.
First of all, some applications may decide never to merge as the journal
contains a complete \jargon{changelog} of the database.  Second, merging
large databases can be slow and the application may wish to schedule
such actions at quiet times or scheduled maintenance periods.
\end{description}

\subsubsection{Enriching the journals}
\label{sec:enrich}

The above predicates suffice for most applications. The predicates in
this section provide access to the journal files and the base state
files and are intented to provide additional services, such as reasoning
about the journals, loaded files, etc.%
    \footnote{A library \pllib{rdf_history} is under development
	      exploiting these features supporting wiki style editing
	      of RDF.}

Using \term{rdf_transaction}{Goal, log(Message)}, we can add additional
records to enrich the journal of affected databases with \arg{Term} and
some additional bookkeeping information. Such a transaction adds a term
\term{begin}{Id, Nest, Time, Message} before the change operations on
each affected database and \term{end}{Id, Nest, Affected} after the
change operations. Here is an example call and content of the journal
file \file{mydb.jrn}.  A full explanation of the terms that appear in
the journal is in the description of rdf_journal_file/2.

\begin{code}
?- rdf_transaction(rdf_assert(s,p,o,mydb), log(by(jan))).
\end{code}

\begin{code}
start([time(1183540570)]).
begin(1, 0, 1183540570.36, by(jan)).
assert(s, p, o).
end(1, 0, []).
end([time(1183540578)]).
\end{code}

Using \term{rdf_transaction}{Goal, log(Message, DB)}, where \arg{DB} is
an atom denoting a (possibly empty) named graph, the system guarantees
that a non-empty transaction will leave a possibly empty transaction
record in DB. This feature assumes named graphs are named after the user
making the changes. If a user action does not affect the user's graph,
such as deleting a triple from another graph, we still find record of
all actions performed by some user in the journal of that user.

\begin{description}
    \predicate{rdf_journal_file}{2}{?DB, ?JournalFile} True if
\arg{File} is the absolute file name of an existing named graph
\arg{DB}. A journal file contains a sequence of Prolog terms of the
following format.%
	\footnote{Future versions of this library may use an XML
		  based language neutral format.}

\begin{description}
    \termitem{start}{Attributes}
Journal has been opened.  Currently \arg{Attributes} contains a
term \term{time}{Stamp}.

    \termitem{end}{Attributes}
Journal was closed.  Currently \arg{Attributes} contains a
term \term{time}{Stamp}.

    \termitem{assert}{Subject, Predicate, Object}
A triple \{Subject, Predicate, Object\} was added to the database.

    \termitem{assert}{Subject, Predicate, Object, Line}
A triple \{Subject, Predicate, Object\} was added to the database
with given \arg{Line} context.

    \termitem{retract}{Subject, Predicate, Object}
A triple  \{Subject, Predicate, Object\} was deleted from the database.
Note that an rdf_retractall/3 call can retract multiple triples.  Each
of them have a record in the journal.  This allows for `undo'.

    \termitem{retract}{Subject, Predicate, Object, Line}
Same as above, for a triple with associated line info.

    \termitem{update}{Subject, Predicate, Object, Action}
See rdf_update/4.

    \termitem{begin}{Id, Nest, Time, Message}
Added before the changes in each database affected by a transaction
with transaction identifier \term{log}{Message}.  \arg{Id} is an
integer counting the logged transactions to this database.  Numbers
are increasing and designed for binary search within the journal file.
\arg{Nest} is the nesting level, where `0' is a toplevel transaction.
\arg{Time} is a time-stamp, currently using float notation with two
fractional digits.  \arg{Message} is the term provided by the user
as argument of the \term{log}{Message} transaction.

    \termitem{end}{Id, Nest, Others}
Added after the changes in each database affected by a transaction with
transaction identifier \term{log}{Message}. \arg{Id} and \arg{Nest}
match the begin-term.  \arg{Others} gives a list of other databases
affected by this transaction and the \arg{Id} of these records.  The
terms in this list have the format \arg{DB}:\arg{Id}.
\end{description}

    \predicate{rdf_db_to_file}{2}{?DB, ?FileBase}
Convert between \arg{DB} (see rdf_source/1) and file base-file used
for storing information on this database.  The full file is located
in the directory described by rdf_current_db/1 and has the extension
\fileext{trp} for the base state and \fileext{jrn} for the journal.
\end{description}


		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
		 %	      TURTLE		%
		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\input{turtle.tex}
\input{rdfntriples.tex}
\InputIfFileExists{rdfa.tex}{}{}


		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
		 %	       RDFS		%
		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\section{library(semweb/rdfs): RDFS related queries}
\label{sec:rdfs}

\index{RDF-Schema}%
The \pllib{semweb/rdfs} library adds interpretation of the triple store
in terms of concepts from RDF-Schema (RDFS). There are two ways to
provide support for more high level languages in RDF. One is to view
such languages as a set of \jargon{entailment rules}. In this model the
rdfs library would provide a predicate \predref{rdfs}{3} providing the
same functionality as rdf/3 on union of the raw graph and triples that
can be derived by applying the RDFS entailment rules.

Alternatively, RDFS provides a view on the RDF store in terms of
individuals, classes, properties, etc., and we can provide predicates
that query the database with this view in mind.  This is the approach
taken in the \pllib{semweb/rdfs.p}l library, providing calls like
\term{rdfs_individual_of}{?Resource, ?Class}.%
	\footnote{The SeRQL language is based on querying the deductive
		  closure of the triple set. The SWI-Prolog SeRQL
		  library provides \jargon{entailment modules} that
		  take the approach outlined above.}


\subsection{Hierarchy and class-individual relations}
\label{sec:semweb-rdfs-classes}

The predicates in this section explore the \const{rdfs:subPropertyOf},
\const{rdfs:subClassOf} and \const{rdf:type} relations.  Note that the
most fundamental of these, \const{rdfs:subPropertyOf}, is also used
by rdf_has/[3,4].

\begin{description}
    \predicate{rdfs_subproperty_of}{2}{?SubProperty, ?Property}
True if \arg{SubProperty} is equal to \arg{Property} or \arg{Property}
can be reached from \arg{SubProperty} following the
\const{rdfs:subPropertyOf} relation. It can be used to test as well as
generate sub-properties or super-properties. Note that the commonly used
semantics of this predicate is wired into rdf_has/[3,4].%
	\bug{The current implementation cannot deal with
	     cycles}.%
	\bug{The current implementation cannot deal with predicates
	     that are an \const{rdfs:subPropertyOf} of
	     \const{rdfs:subPropertyOf}, such as
	     \const{owl:samePropertyAs}.}

    \predicate{rdfs_subclass_of}{2}{?SubClass, ?Class}
True if \arg{SubClass} is equal to \arg{Class} or \arg{Class}
can be reached from \arg{SubClass} following the
\const{rdfs:subClassOf} relation. It can be used to test as
well as generate sub-classes or super-classes.%
	\bug{The current implementation cannot deal with
	     cycles}.

    \predicate{rdfs_class_property}{2}{+Class, ?Property}
True if the domain of \arg{Property} includes \arg{Class}.  Used to
generate all properties that apply to a class.

    \predicate{rdfs_individual_of}{2}{?Resource, ?Class}
True if \arg{Resource} is an indivisual of \arg{Class}.  This implies
\arg{Resource} has an \const{rdf:type} property that refers to
\arg{Class} or a sub-class thereof.  Can be used to test, generate
classes \arg{Resource} belongs to or generate individuals described
by \arg{Class}.
\end{description}

\subsection{Collections and Containers}
\label{sec:semweb-rdfs-containers}

\index{parseType,Collection}%
\index{Collection,parseType}%
The RDF construct \const{rdf:parseType}=\const{Collection} constructs
a list using the \const{rdf:first} and \const{rdf:next} relations.

\begin{description}
    \predicate{rdfs_member}{2}{?Resource, +Set}
Test or generate the members of \arg{Set}.  \arg{Set} is either an
individual of \const{rdf:List} or \const{rdfs:Container}.

    \predicate{rdfs_list_to_prolog_list}{2}{+Set, -List}
Convert \arg{Set}, which must be an individual of \const{rdf:List} into
a Prolog list of objects.

    \predicate{rdfs_assert_list}{2}{+List, -Resource}
Equivalent to rdfs_assert_list/3 using \arg{DB} = \const{user}.

    \predicate{rdfs_assert_list}{3}{+List, -Resource, +DB}
If \arg{List} is a list of resources, create an RDF list \arg{Resource}
that reflects these resources.  \arg{Resource} and the sublist resources
are generated with rdf_bnode/1.  The new triples are associated with the
database \arg{DB}.
\end{description}


		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
		 %	     LIBRARY		%
		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\InputIfFileExists{rdflib.tex}{}{}

		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
		 %	  PLDOC LIBRARIES	%
		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\InputIfFileExists{sparqlclient.tex}{}{}
\InputIfFileExists{rdfcompare.tex}{}{}
\InputIfFileExists{rdfportray.tex}{}{}


\section{Related packages}
\label{sec:semweb-related-packages}

\index{ClioPatria}\index{SPARQL}%
The core infrastructure for storing and querying RDF is provided by this
package, which is distributed as a core package with SWI-Prolog.
\href{http://cliopatria.swi-prolog.org}{ClioPatria} provides a
comprehensive server infrastructure on top of the \textit{semweb} and
\textit{http} packages. ClioPatria provides a SPARQL~1.1 endpoint,
linked open data (LOD) support, user management, a web interface and an
extension infrastructure for programming (semantic) web applications.

\index{Thea}\index{OWL2}%
\href{http://www.semanticweb.gr/TheaOWLLib/}{Thea} provides access to OWL
ontologies at the level of the abstract syntax. Can interact with
external DL reasoner using DIG.



		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
		 %	    VERSION 3		%
		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\section{Version 3 release notes}
\label{sec:semweb-version3}

RDF-DB version 3 is a major redesign of the SWI-Prolog RDF
infrastructure. Nevertheles, version~3 is almost perfectly upward
compatible with version~2. Below are some issues to take into
consideration when upgrading.

Version~2 did not allow for modifications while read operations were in
progress, for example due to an open choice point. As a consequence,
operations that both queried and modified the database had to be wrapped
in a transaction or the modifications had to be buffered as Prolog data
structures. In both cases, the RDF store was not modified during the
query phase. In version~3, modifications are allowed while read
operations are in progress and follow the Prolog \textbf{logical update
view} semantics. This is different from using a transaction in
version~2, where the view for all read operations was frozen at the
start of the transaction. In version~3, every read operation sees the
store frozen at the moment that the operation was started.

We illustrate the difference by writing a forwards entailment rule that
adds a \jargon{sibling} relation. In \textbf{version~2}, we could
perform this operation using one of the following:

\begin{code}
add_siblings_1 :-
	findall(S-O,
		( rdf(S, f:parent, P),
		  rdf(O, f:parent, P),
		  S \== O
		),
		Pairs),
	forall(member(S-O, Pairs), rdf_assert(S,f:sibling,O)).

add_siblings_2 :-
	rdf_transaction(
	    forall(( rdf(S, f:parent, P),
		     rdf(O, f:parent, P),
		     S \== O
		   ),
		   rdf_assert(S, f:sibling, O))).
\end{code}

In \textbf{version~3}, we can write this in the natural Prolog style
below. In itself, this may not seem a big advantage because wrapping
such operations in a transaction is often a good style anyway. The story
changes with more complicated constrol structures that combine
iterations with steps that depend on triples asserted in previous steps.
Such scenarios can be programmed naturally in the current version.

\begin{code}
add_siblings_3 :-
	forall(( rdf(S, f:parent, P),
		 rdf(O, f:parent, P),
		 S \== O
	       ),
	       rdf_assert(S, f:sibling, O)).
\end{code}

In version~3, code that combines queries with modification has the same
semantics whether executed inside or outside a transaction. This
property makes reusing such predicates predictable.


\begin{description}
    \item [rdf_statistics/2]
Various statistics have been renamed or changed:
    \begin{shortlist}
        \item \const{sources} is renamed into \const{graphs}
        \item \const{triples_by_file} is renamed into
	      \const{triples_by_graph}
	\item \const{gc} has additional arguments
	\item \const{core} is removed.
    \end{shortlist}

    \item [rdf_generation/1]
Generations inside a transaction are represented as
\arg{BaseGeneration}+\arg{TransactionGeneration}, where
\arg{BaseGeneration} is the global generation where the transaction
started and \arg{TransactionGeneration} expresses the generation within
the transaction. Counting generation has changed as well. In particular,
comitting a transaction steps the generation only by one.

    \item [rdf_current_ns/1, rdf_register_ns/2, rdf_register_ns/3]
These predicates are renamed into rdf_current_prefix/1,
rdf_register_prefix/2, rdf_register_prefix/3.  The old predicates
are still available as deprecated predicates.

    \item [rdf_unload/1] now only accepts a source location and
deletes the associated graph using rdf_unload_graph/1.
\end{description}


\section*{Acknowledgements}

This research was supported by the following projects: MIA and
MultimediaN project (www.multimedian.nl) funded through the BSIK
programme of the Dutch Government, the FP-6 project HOPS of the European
Commission, the COMBINE project supported by the ONR Global NICOP grant
N62909-11-1-7060 and the Dutch national program COMMIT.



		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
		 %	      FOOTER		%
		 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\printindex

\end{document}