1
0
mirror of https://github.com/cookiengineer/audacity synced 2025-05-07 07:12:34 +02:00
audacity/lib-src/redland/docs/storage.html
2010-01-24 09:19:39 +00:00

542 lines
18 KiB
HTML

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<html xmlns="http://www.w3.org/1999/xhtml" lang="en" xml:lang="en">
<head>
<title>Redland RDF Application Framework - Storage Modules</title>
</head>
<body>
<h1 style="text-align:center">Redland RDF Application Framework - Storage Modules</h1>
<h2>Introduction</h2>
<p>Redland includes several modules that implement the
<a href="api/storage.html">storage API</a> and provide a variety of
different features and functionality. This document gives the
details of what each implementation module provides and
the storage options used.
</p>
<p>Store types:</p>
<ul>
<li><a href="#hashes">hashes</a></li>
<li><a href="#trees">trees</a></li>
<li><a href="#file">file</a></li>
<li><a href="#mysql">mysql</a></li>
<li><a href="#memory">memory</a></li>
<li><a href="#postgresql">postgresql</a></li>
<li><a href="#sqlite">sqlite</a></li>
<li><a href="#tstore">tstore</a></li>
<li><a href="#uri">uri</a></li>
</ul>
<h2><a name="hashes">Store 'hashes'</a></h2>
<p>This module is always present (cannot be removed) and provides
indexed storage using Redland <a href="api/storage.html">hashes</a>
to store various combinations of subject, predicate and object for
faster access. Context nodes are also stored in a hash when used.
The hashes may be in-memory (always availabe) or persistent via
Sleepycat/Berkeley DB (BDB) versions 2-4. It is the most mature and
primary persistent store and suitable for large models, tested in the
2-3 million range..</p>
<p>The main option requiring setting is the
<a name="hash-type"><code>hash-type</code></a>
which must be one of the supported Redland hashes.
Hash type <code>memory</code> is always available and if BDB
has been compiled in, <code>bdb</code> is also available.
Option <code>dir</code> can be used to set the destination
directory for the BDB files when used. Boolean option
<code>new</code> can be set to force creation or truncation
of a persistent hashed store. The storage
name must be given for hash type <code>bdb</code> since
it is used for a filename.
</p>
<p>The module provides optional contexts support enabled when
boolean storage option <code>contexts</code> is set. This
can be used with any hash type.</p>
<p>Examples:</p>
<pre>
/* A new BDB hashed persistent store in the current directory */
storage=librdf_new_storage(world, "hashes", "db1",
"new='yes',hash-type='bdb',dir='.'");
/* Hashed in-memory store */
storage=librdf_new_storage(world, "hashes", NULL,
"hash-type='memory'");
/* An existing BDB hashed persistent store in dir /somewhere */
storage=librdf_new_storage(world, "hashes", "dv2",
"hash-type='bdb',dir='/somewhere'");
/* An existing BDB hashed store with contexts */
storage=librdf_new_storage(world, "hashes", "db3",
"hash-type='bdb',contexts='yes'");
</pre>
<p>In Python:</p>
<pre>
from RDF import *
...
# Create a new BDB store
storage = HashStorage("db4", options="new='yes',hash-type='bdb'")
</pre>
<p>In Perl:</p>
<pre>
use RDF::Redland;
...
# Open an existing BDB store
$storage=new RDF::Redland::Storage("hashes", "db5",
"hash-type='bdb',dir='.'");
</pre>
<p>Summary:</p>
<ul>
<li>Persistent or in-memory</li>
<li>Suitable for larger models</li>
<li>Indexed</li>
<li>Large disk usage with BDB</li>
<li>Optional contexts (with option <code>contexts</code> set)</li>
</ul>
<h2><a name="trees">Store 'trees'</a></h2>
<p>This module is always present (cannot be removed) and provides
indexed storage using several balanced trees to store statements
for fast querying. This store is not persistent, but is suitable
for large models capable of fitting in main memory.</p>
<p>By default, the store is fully indexed providing good performance
for all types of queries. Options can be used to select only specific
indices to save memory and make insertion and deletion of statements
faster. The four boolean indexing options are <code>index-spo</code>,
<code>index-sop</code>, <code>index-ops</code> and <code>index-pso</code>.
An index is fast for triple patterns where the variables are on the right
hand side of the index ordering, e.g. the spo (subject, predicate, object)
index will be fast for (s p o) (s p ?o) and (s ?p ?o) queries. The ideal
index for every type of triple pattern is:
</p>
<dl>
<dt>Any index:</dt><dd>(s p o)<br/><br/></dd>
<dt>(s p o):</dt><dd>(s p ?o), (s ?p ?o)<br/><br/></dd>
<dt>(o p s):</dt><dd>(?s p o), (?s ?p o)<br/><br/></dd>
<dt>(s o p):</dt><dd>(s ?p o)<br/><br/></dd>
<dt>(p s o):</dt><dd>(?s p ?o)<br/><br/></dd>
</dl>
<p>
With full indexing the space used is roughly equivalent to the hashes
store. Insertion and deletion with 2 indices will be roughly twice as
fast as with 4 indices, etc. In the majority of cases, full indexing
will be fine and there is no need to worry about selecting the right
index for queries.
</p>
<p>Examples:</p>
<pre>
/* A fully indexed tree store */
storage=librdf_new_storage(world, "trees", NULL, NULL);
/* A tree store with only a (s p o) index */
storage=librdf_new_storage(world, "trees", NULL, "index-spo='yes'");
/* A tree store with only (s p o) and (o p s) indices
* (fast for everything but queries with predicate variables */
storage=librdf_new_storage(world, "trees", NULL,
"index-spo='yes',index-ops='yes'");
</pre>
<p>Summary:</p>
<ul>
<li>In-memory only</li>
<li>Suitable for larger models</li>
<li>Indexed, with selectable levels of indexing</li>
<li>No contexts</li>
<li>Significantly faster than hashes for most queries</li>
<li>Slower than hashes for exact statement search (librdf_model_contains_statement)</li>
</ul>
<h2><a name="memory">Store 'memory'</a></h2>
<p>This module is always present (cannot be removed) and provides a
simple and fast in-memory store with no persistence. It is
the default store if no store name is given to the storage
constructors.</p>
<p>The memory store is not suitable for large in-memory models since
it does not do any indexing. For that, use the
<a href="#hashes">hash indexed store</a> with
<a href="#hash-type">hash-type</a> of <code>memory</code>.</p>
<p>The module provides optional contexts support enabled when
boolean storage option <code>contexts</code> is set.</p>
<p>Examples:</p>
<pre>
/* Explictly named memory storage */
storage=librdf_new_storage(world, "memory", NULL, NULL);
/* Default storage type, which is memory */
storage=librdf_new_storage(world, NULL, NULL, NULL);
/* In-memory store with contexts */
storage=librdf_new_storage(world, NULL, NULL, "contexts='yes'");
</pre>
<p>Summary:</p>
<ul>
<li>In-memory</li>
<li>Fast</li>
<li>Suitable for small models</li>
<li>No indexing</li>
<li>No persistance</li>
<li>Optional contexts (with option <code>contexts</code> set)</li>
</ul>
<h2><a name="file">Store 'file'</a></h2>
<p>This module provides an in-memory model (internally,
using a <a href="#memory">memory storage</a>) initialised from the
RDF/XML content in a file. The file is given as the storage name and
assumed to exist on opening. When a model or storage sync method
is called or the model or store is closed, a new file is created, the
old file renamed to a backup and the new file renamed to replace it.
This store was added in <a href="../RELEASE.html#rel0_9_15">Redland 0.9.15</a>
</p>
<p>There are no options for this store and contexts are not supported.</p>
<p>Example:</p>
<pre>
/* File based store from thing.rdf file */
storage=librdf_new_storage(world, "file", "thing.rdf", NULL);
</pre>
<p>Summary:</p>
<ul>
<li>In-memory</li>
<li>Suitable for small models</li>
<li>Simple local storage to content in RDF/XML</li>
<li>Not indexed</li>
<li>No contexts</li>
</ul>
<h2><a name="mysql">Store 'mysql'</a></h2>
<p>This module was written by
<a href="http://purl.org/net/morten/">Morten Frederiksen</a>
and is compiled in when MySQL 3 or 4 is available. This
store provides storage using the <a href="http://www.mysql.com/">MySQL</a>
open source database including contexts. It is a new store
added in
<a href="../RELEASE.html#rel0_9_15">Redland 0.9.15</a>
and is still under development. It has however been tested with
several million triples and deployed.</p>
<p>There are several options required with the mysql storage
in order to connect to the database. These are:</p>
<ul>
<li><code>host</code> for the database server hostname</li>
<li><code>port</code> for the database server port (defaults to the MySQL port 3306 if not given)</li>
<li><code>database</code> for the MySQL database name (not the storage name)</li>
<li><code>user</code> for the database server user name</li>
<li><code>password</code> for the database server password</li>
</ul>
<p>NOTE: Take care exposing the password as for example, program
arguments or environment variables. The
<a href="../utils/rdfproc.html">rdfproc</a>
utility can with help this by reading the password from standard
input. Inside programs, one way to prevent storing the password in a
string is to construct a Redland hash of the storage options such as
via <code>librdf_hash_from_string</code> and use
<code>librdf_new_storage_with_options</code> to create a storage.
The rdfproc utility source code demonstrates this.
</p>
<p>The storage name parameter given to the storage constructor
<code>librdf_new_storage</code> is used inside the mysql store to
allow multiple stores inside one MySQL database instance as
parameterised with the above optiosn.</p>
<p>If boolean option <code>new</code> is given, any existing MySQL
database named by the storage option <code>database</code>, say
<em>db</em> will be dropped and the appropriate new tables created.
The MySQL database <em>db</em> must already exist, such as made with
the MySQL <code>create database </code><em>db</em> command and the
appropriate priviledges set so that the user and password work.
</p>
<p>If boolean option <code>reconnect</code> is given, MySQL
reconnection will be enabled so that if the database conneciton
is dropped, MySQL will attempt to reconnect.
</p>
<p>This store always provides contexts; the boolean storage option
<code>contexts</code> is not checked.</p>
<p>Examples:</p>
<pre>
/* A new MySQL store */
storage=librdf_new_storage(world, "mysql", "db1",
"new='yes',host='localhost',database='red',user='foo','password='bar'");
/* A different, existing MySQL store db2 in the same database as above */
storage=librdf_new_storage(world, "mysql", "db2",
"host='localhost',database='red',user='foo','password='bar'");
/* An existing MySQL store on a different database server */
storage=librdf_new_storage(world, "mysql", "db3",
"host='db.example.org',database='abc',user='baz','password='blah'");
/* Opening with an options hash */
options=librdf_new_hash(world, NULL);
librdf_hash_from_string(options,
"host='db.example.org',database='abc',user='baz'");
librdf_hash_put_strings(options, "password", user_password);
storage=librdf_new_storage_with_options(world, "mysql", "db4", options);
</pre>
<p>In PHP:</p>
<pre>
# An existing store
$storage=librdf_new_storage($world, 'mysql', 'db4',
"host='127.0.0.1',database='xyz',user='foo',password='blah'");
</pre>
<p>Summary:</p>
<ul>
<li>Persistent</li>
<li>Suitable for very large models</li>
<li>Indexed but not optimised</li>
<li>Smaller disk usage than BDB</li>
<li>Possibility of free text searching</li>
<li>Contexts always provided</li>
</ul>
<h2><a name="postgresql">Store 'postgresql'</a></h2>
<p>This module was written by Shi Wenzhong based on the MySQL store
written by Morten Frederiksen and is compiled in when PostgreSQL is
available. This store provides storage using the
<a href="http://www.postgresql.org/">PostgreSQL</a>
open source database including contexts. This store was added in
This <a href="../RELEASE.html#rel1_0_3">Redland 1.0.3</a>.</p>
<p>There are several options required with the postgresql storage
in order to connect to the database. These are:</p>
<ul>
<li><code>host</code> for the database server hostname</li>
<li><code>port</code> for the database server port (defaults to the Postgresql port 3306 if not given)</li>
<li><code>database</code> for the Postgresql database name (not the storage name)</li>
<li><code>user</code> for the database server user name</li>
<li><code>password</code> for the database server password</li>
<li><code>database</code> for the Postgresql database name (not the storage name)</li>
</ul>
<p>NOTE: Before Redland 1.0.5, the parameter <code>dbname</code> had
to be used instead of <code>database</code> for the Postgresql
database name</p>
<p>NOTE: Take care exposing the password as for example, program
arguments or environment variables. The
<a href="../utils/rdfproc.html">rdfproc</a>
utility can with help this by reading the password from standard
input. Inside programs, one way to prevent storing the password in a
string is to construct a Redland hash of the storage options such as
via <code>librdf_hash_from_string</code> and use
<code>librdf_new_storage_with_options</code> to create a storage.
The rdfproc utility source code demonstrates this.
</p>
<p>The storage name parameter given to the storage constructor
<code>librdf_new_storage</code> is used inside the postgresql store to
allow multiple stores inside one PostgreSQL database instance as
parameterised with the above optiosn.</p>
<p>If boolean option <code>new</code> is given, any existing PostgreSQL
database named by the storage option <code>database</code>, say
<em>db</em> will be dropped and the appropriate new tables created.
The PostgreSQL database <em>db</em> must already exist, such as made with
the PostgreSQL <code>create database </code><em>db</em> command and the
appropriate priviledges set so that the user and password work.</p>
<p>This store always provides contexts; the boolean storage option
<code>contexts</code> is not checked.</p>
<p>Examples:</p>
<pre>
/* A new PostgreSQL store */
storage=librdf_new_storage(world, "postgresql", "db1",
"new='yes',host='localhost',database='red',user='foo','password='bar'");
/* A different, existing PostgreSQL store db2 in the same database as above */
storage=librdf_new_storage(world, "postgresql", "db2",
"host='localhost',database='red',user='foo','password='bar'");
/* An existing PostgreSQL store on a different database server */
storage=librdf_new_storage(world, "postgresql", "db3",
"host='db.example.org',database='abc',user='baz','password='blah'");
/* Opening with an options hash */
options=librdf_new_hash(world, NULL);
librdf_hash_from_string(options,
"host='db.example.org',database='abc',user='baz'");
librdf_hash_put_strings(options, "password", user_password);
storage=librdf_new_storage_with_options(world, "postgresql", "db4", options);
</pre>
<p>In PHP:</p>
<pre>
# An existing store
$storage=librdf_new_storage($world, 'postgresql', 'db4',
"host='127.0.0.1',database='xyz',user='foo',password='blah'");
</pre>
<p>Summary:</p>
<ul>
<li>Persistent</li>
<li>Suitable for very large models</li>
<li>Indexed but not optimised</li>
<li>Smaller disk usage than BDB</li>
<li>Contexts always provided</li>
</ul>
<h2><a name="sqlite">Store 'sqlite'</a></h2>
<p>This module provides storage via the
<a href="http://www.sqlite.org/">SQLite</a>
relational database when available and supports SQLite V2 and V3.
It was added in <a href="../RELEASE.html#rel1_0_0">Redland 1.0.0</a>
and is of beta quality. This store provides triples and contexts.
</p>
<p>The only option respected by this store is the
<code>new</code> one to create a new store, destroying any existing
store.
</p>
<p>Summary:</p>
<ul>
<li>Persistent</li>
<li>Suitable for small/medium models</li>
<li>Indexed but not optimized</li>
<li>Smaller disk usage than BDB</li>
<li>Contexts always provided</li>
</ul>
<h2><a name="tstore">Store 'tstore'</a></h2>
<p>This module provides storage via the
<a href="http://triplestore.aktors.org/">AKT Triplestore</a>
when available. It was added in
<a href="../RELEASE.html#rel0_9_15">Redland 0.9.15</a>
and is alpha quality - not complete or tested significantly (although
the AKT store itself is used in production). This store provides a
basic triple API but no redland contexts. The underlying RDQL
support and inference is not yet exposed in Redland but may be in
future.
</p>
<p>There are several options required with the tstore storage in
order to connect to the tstore database (which uses MySQL). These
are:</p>
<ul>
<li><code>host</code> for the database server hostname</li>
<li><code>port</code> for the database server port</li>
<li><code>database</code> for the database name (not the storage name)</li>
<li><code>user</code> for the database server user name</li>
<li><code>password</code> for the database server password</li>
</ul>
<p>NOTE: Take care exposing the password as for example, program
arguments or environment variables. The
<a href="../utils/rdfproc.html">rdfproc</a>
utility can with help this by reading the password from standard
input. Inside programs, one way to prevent storing the password in a
string is to construct a Redland hash of the storage options such as
via <code>librdf_hash_from_string</code> and use
<code>librdf_new_storage_with_options</code> to create a storage.
The rdfproc utility source code demonstrates this.
</p>
<p>Summary:</p>
<ul>
<li>Persistent</li>
<li>Suitable for very large models</li>
<li>Indexed and optimised by the AKT project</li>
<li>No Redland contexts</li>
<li>Alpha quality</li>
</ul>
<h2><a name="uri">Store 'uri'</a></h2>
<p>This module provides an in-memory model (internally,
using a <a href="#memory">memory storage</a>) initialised from the
RDF/XML content in a URI. The URI is given as the storage name and
on closing, the model is destroyed. This is a new store added in
<a href="../RELEASE.html#rel0_9_15">Redland 0.9.15</a>
and is still under development. In future it may be
extended to allow saving the store to the URI.
</p>
<p>There are no options for this store and contexts are not supported.</p>
<p>Example:</p>
<pre>
/* Read URI content into a store */
storage=librdf_new_storage(world, "uri",
"http://example.org/content.rdf", NULL);
</pre>
<p>Summary:</p>
<ul>
<li>In-memory</li>
<li>Suitable for small models</li>
<li>Easy access to web content in RDF/XML</li>
<li>Not indexed</li>
<li>No contexts</li>
</ul>
<hr />
<p>Copyright 2004-2007 <a href="http://purl.org/net/dajobe/">Dave Beckett</a>, Copyright 2004 <a href="http://www.bristol.ac.uk/">University of Bristol</a></p>
</body>
</html>