There are many good introductions to Clojure; for instance see Clojure Doc or Learn Clojure. However, most of these are aimed at learning Clojure as a programming language; at its simplest, Tawny-OWL can be used as a textual method for writing OWL based ontologies, and understanding how Clojure works as a programming language is not necessary. This, therefore, is a high-level overview describing the basic tools that are necessary for Tawny-OWL use.
Clojure is a modern JVM-based Lisp implementation. Clojure, and therefore, Tawny-OWL files are just text and can, therefore, be read with any text-editor; although using one which can cope with alternative line-endings (i.e. not notepad!) will make life easier.
Building an ontology is, in general, a complex business; in general, it works better with a good editor or IDE. There are many of these that you can use; I am a die-hard Emacs fan, but Eclipse, IntelliJ and Netbeans are all good alternatives. Clojure Doc has reasonable coverage.
For those using Emacs, if you are not familiar with lisp I would advice ignoring the advice of many tutorials to install paredit. It really is good for Lisp, but Tawny-OWL doesn't really nest very much, so it's not that big an advantage, and can be disconcerting in first use.
Clojure is highly-dynamic, so "build" may be the wrong term. There is no compile-test cycle with Clojure. However, Tawny-OWL uses "Leiningen" to perform two tasks. First, it resolves dependencies; any ontology built with Tawny-OWL will have a large number of these, and they can change over time, so not maintaining this by hand is a boon. Secondly, Lein provides an easy way to launch the REPL -- read-eval-print-loop.
Leiningen can be installed from a package manager, although the most release (Lein 2) is quite recent and many repositories have the older Lein 1. In this case, you need to install from the main lein website. Installation is relatively straight-forward on both Windows and Linux. Tawny-OWL will work with Lein 1, but the test case fail.
Leiningen uses a project file to define the build. The file is called
project.clj
and it, itself, written in Clojure. A reasonable example for
Tawny-OWL can be see in the
Tawny Pizza ontology.
(defproject pizza "0.1-SNAPSHOT"
:description "The Pizza Ontology in tawny-owl"
:dependencies [[org.clojure/clojure "1.4.0"]
[uk.org.russet/tawny-owl "0.9"]
]
:main pizza.core
)
The project definition is itself written in Clojure; in most cases Tawny
projects will require just Clojure and Tawny as dependencies (the version
numbers of these will change!). The description, and project name will, of
course be unique to your project. Leiningen will create a basic project.clj
by running the lein new
command.
Clojure uses namespaces to avoid name clashes. Although less rigid than Java,
most often a clojure namespace maps to a single file. Leiningen expects files
at a specific location. All sources files are held in, unsurprisingly, the
src
directory. Directories below this create a namespace hierarchy. So, the
main source for the pizza ontology is found at src/pizza
.
The first thing that you need to do with a Tawny ontology is evaluate a file; that is, have Clojure interpret all the statements, and turn them into an something that you can interact with. Unfortunately, the different environments that you might choose do this differently. However, the are plenty of tutorials; for instance I use emacs.
This is the main thing that you may want to do with Tawny, and it is covered in its own documentation.
You don't need a main function for an ontology, but it can be useful. For instance, tawny-pizza includes a simple main, that writes the ontology in a couple of different formats. The practical upshot of this is that if you generate an single jar file for a project, running this file by, for instance, clicking on it will generate OWL files.
Previously, we saw a project definition including the statement :main pizza.core
. This describes the namespace that the main method is expected to
appear in. In this file (src/pizza/core.clj
), we find a namespace
declaration.
(ns pizza.core
(:use [tawny.owl]
[pizza.pizza])
(:gen-class)
)
and a simple main method. In practice, a main method isn't that useful, but it will be run whenever the project jar is clicked on. In this case, the method simply saves the ontology out in a couple of formats.
(defn -main [& args]
(with-ontology pizzaontology
(save-ontology "pizza.rdf" :rdf)
(save-ontology "pizza.owl" :owl)))
With Tawny, there are three ways to publish an ontology. First, an ontology
developed in Tawny can be written out as OWL, in any of many formats, using
the save-ontology
function, which is covered in more detail in the
getting-started documentation. Anyone using an OWL
ontology will be able to consume this form, including another Tawny user.
Secondly, it is possible to release the Tawny-OWL files themselves as source.
This will be useful to other Tawny-OWL users, but is only really necessary if
they wish to modify the ontology; it is best accomplished through a version
control system. Finally, the Tawny-OWL files can be packaged as a jar file and
published to a maven repository. This is the best way to share an ontology
with other Tawny-OWL users who wish to use, but not change your ontology. The
most straight-forward to do this is through the
Clojars repository; there is a good
tutorial available for
this.
Can you improve this documentation? These fine people already did:
Phillip Lord & BastienEdit on GitHub
cljdoc is a website building & hosting documentation for Clojure/Script libraries
× close