easyLOD exposes arbitrary content as Linked Open Data. This data could come from a relational database, from a set of static files, or from an external API available over the web. Each source of data has its own plugin.
easyLOD's goal is to make it as simple to publish Linked Data while incorporating as many Linked Open Data best practices as possible. For more information on publishing Linked Data, consult Tom Heath and Christian Bizer's Linked Data: Evolving the Web into a Global Data Space (http://linkeddatabook.com/editions/1.0/).
- No dependencies other than PHP 5.3 or higher.
- Written in Slim, a PHP micro-framework (included in the easyLOD distribution).
- Performs content negotiation to allow Linked Data browsers to request RDF-encoded data, human-readable web browsers to request an ordinary web page about the resource.
- Plugins for data sources are easy to write (more info is provided below).
- Gets data from sources in realtime, not by writing out representations to files.
- Resource URIs use 'namespaces' to identify data sources, so when data source back end changes, URIs don't need to (more info is provided below).
Everything on the Web of Data must have a unique URI. HTTP URIs have a server name and a path, and within a given server name, the unique parts of the URI are expressed in its path. If an organization assigns unique identifiers to the things it describes, these identifers can be used as the unique parts of URIs.
easyLOD imposes a specific pattern for URIs, namely, the string 'resource', a string (the 'namespace'), then a colon (':'), then another string (the 'ID'). The combination of namespace and ID must be unique. For example, a URI managed by easyLOD looks like this:
http://myorg.edu/pathtoeasylod/resource/namespace3:foo
The ID (in the case of the example above, 'foo') can be any string of characters that are valid in URIs. easyLOD doesn't assing the IDs -- that's up to you. The namespace (in this case, 'namespace3') maps to a data source plugin (see below for how to create explicit mappings from namespaces to plugins). The unique combination of namespace and ID tell easyLOD which data source plugin to use, and then which item managed by that plugin the request is for.
easyLOD uses plugins to retrieve data, which it then wraps in RDF/XML to send to the Linked Data browser. If the browser making the request is not a Linked Data browser (i.e., probably a human using Chrome, Firefox, etc.), the data is wrapped in HTML and sent to the browser. Plugins can also redirect users to external websites that describe the item corresponding to the identifier.
Five plugins are provided with easyLOD:
- a plugin that retrieves Dublin Core metadata from CONTENTdm (http://contentdm.org/), which provides a web-services API
- a plugin that gets FOAF data (http://xmlns.com/foaf/spec/) from a small CSV file (included in the plugin directory)
- a plugin that retrieves Dublin Core metadata describing books from a MySQL database (SQL file is included in the plugin directory)
- a plugin that serves static RDF files transformed from MODS descriptions for items in an Islandora (http://islandora.ca/) repository
- a plugin that scrapes information from author pages on Simon Fraser University's Summit Institutional Respository (http://summit.sfu.ca) using the PHP Simple HTML DOM Parser (http://simplehtmldom.sourceforge.net/), and exposes that information as RDF containing FOAF, BIO (http://vocab.org/bio/0.1/.html), OWL (http://www.w3.org/TR/owl-ref/), and Dublin Core properties.
These are intended to illustrate how information can be retieved from different data sources.
There are two ways to tell easyLOD which namespaces should map to which data source plugins: 1) direct mappings and 2) configured mappings. In a direct mapping, the URI's namespace is identical to the plugin name; in a configured mapping, an entry in a file named 'plugins.php' indicates which namespaces invoke which plugins. Both approaches can work at the same time, although if there is a conflict (that is, there is both a configured mapping for a namespace and there is a plugin file with the same name), the configured mapping wins.
If the namespaces in your resource URIs map one-to-one to your data source plugins, you do not need a plugins.php file.
If you want to use the same plugin for URIs with different namespaces, or you want to use a namespace that is cleaner, shorter, or descriptive of your data than a plugin filename would allow, you need to use a plugins.php file. In other words, configured mappings allow more flexibility in choosing your namespaces. easyLOD ships with an example plugins.php file that illustrates configured mappings. The file contains an associative array $plugins that has members representing namespaces; each of these is also an associative array that defines which data source plugin it to be used and also an optional 'dataSourceConfg' associative array that overrides the plugin's config settings as defined in its dataSourceConfig() function:
$plugins = array(
// Configuration for URIs using the 'foo' namespace.
'foo' => array(
'plugin' => 'foaf',
'dataSourceConfig' => array(
'input_file' => 'data_sources/foaf/people_foo.csv',
),
),
// Configuration for URIs using the 'bar' namespace.
'bar' => array(
'plugin' => 'foaf',
'dataSourceConfig' => array(
'input_file' => 'data_sources/foaf/people_bar.csv',
),
),
);
All of a plugin's settings must be included in the entries in plugins.php. Each of the top-level entries in the plugins.php file (in this example, 'foo' and 'bar') are independent, and the order in which they appear in the file doesn't matter. $plugins is just an ordinary PHP array.
Unzip the distribution, put it somewhere in your web server's document root, and then use the following command to test the FOAF plugin (this command assumes you are issuing it on the same server where you installed easyLOD):
curl -L -H 'Accept: application/rdf+xml' http://localhost/easyLOD/resource/foaf:random@modnar44.com
You should see the following:
<?xml version="1.0" encoding="UTF-8"?>
<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:foaf="http://xmlns.com/foaf/0.1/">
<rdf:Description rdf:about="http://localhost/easyLOD/resource/foaf:random@modnar44.com">
<foaf:mbox>random@modnar44.com</foaf:mbox>
<foaf:surname>Jalfrezi</foaf:surname>
<foaf:givenName>Large</foaf:givenName>
<foaf:page>http://jalfrezitogo.edu</foaf:page>
</rdf:Description>
</rdf:RDF>
Now open a graphical web browser and go to http://localhost/easyLOD/resource/foaf:random@modnar44.com. You should see a simple web page with this content:
Large Jalfrezi
mbox: random@modnar44.com
surname: Jalfrezi
givenName: Large
page: http://jalfrezitogo.edu
The two different responses to the same URI illustrate the Linked Data concept of content negotiation: browsers that request RDF, which is what Linked Data browsers do, receive your data in RDF, while ordinary web browsers receive your data in a format that is more readable by humans.
A separate tool, viewLOD (https://github.com/mjordan/viewLOD) is available for emulating both human and Linkded Data browsers using a simple web form.
easyLOD does not provide a mechanism for querying the Linked Data it exposes. Many Linked Data providers supply a SPARQL endpoint to allow queries. One strategy for providing a SPARQL endpoint for easyLOD would be to:
- Configure easyLOD to expose data using a plugin.
- Write a script to iterate through all of the identifiers in your data and send corresponding queries to their easyLOD resource URLs.
- Have the script write out the RDF/XML representations of your resources as static files.
- Install and configure the Mulgara RDF Database (http://www.mulgara.org/) or some other RDF triple store that provides its own SPARQL endpoint.
- Load your data as RDF triples.
- Expose the databases's SPARQL endpoint for Linked Data applications to query.