Trike - Training Wheels for Triples

Trike.ipynb

{
 "metadata": {
  "kernelspec": {
   "codemirror_mode": {
    "name": "python",
    "version": 2
   },
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "name": "",
  "signature": "sha256:7545436b58570935c2a0fca05c72334cb3a3d1e371e62592e5da64b335ca1303"
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "# Trike: training wheels for semantic queries\n",
      "## An abuse of most of python's `__` functions for pithy semantic queries\n",
      "- Be shorter than\n",
      "  - Raw SPARQL\n",
      "  - SurfRDF\n",
      "- Fewer strings\n",
      "- Better handling of Literals (remains to be seen)"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from __future__ import unicode_literals\n",
      "\n",
      "import sys\n",
      "import os\n",
      "\n",
      "from rdflib.plugins.stores.sparqlstore import SPARQLStore\n",
      "from rdflib import Graph, ConjunctiveGraph, Namespace\n",
      "from rdflib.namespace import (\n",
      "    RDF as rdf,\n",
      "    RDFS as rdfs,\n",
      "    FOAF as foaf\n",
      ")\n",
      "\n",
      "from jinja2 import Template\n",
      "\n",
      "from rdflib.term import Literal, URIRef"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 3
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "Use data from [LinkedMBD](http://linkedmdb.org)."
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "store = SPARQLStore(\"http://data.linkedmdb.org/sparql\", context_aware=False)\n",
      "graph = Graph(store)\n",
      "\n",
      "movie = Namespace(\"http://data.linkedmdb.org/resource/movie/\")"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 4
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "The classes."
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "class URITerm(object):\n",
      "    def __init__(self, uri):\n",
      "        self.uri = uri\n",
      "    \n",
      "    def __str__(self):\n",
      "        return \"<%s>\" % self.uri\n",
      "    \n",
      "\n",
      "class Variable(object):\n",
      "    def __init__(self, name):\n",
      "        self.name = name\n",
      "\n",
      "    def __str__(self):\n",
      "        return \"?%s\" % self.name \n",
      "    \n",
      "    \n",
      "class LiteralBinding(object):\n",
      "    _next_id = 0\n",
      "    \n",
      "    def __init__(self, literal, *args, **kwargs):\n",
      "        self.id = LiteralBinding.next_id()\n",
      "        self.literal = Literal(literal, *args, **kwargs)\n",
      "\n",
      "    @classmethod\n",
      "    def next_id(cls):\n",
      "        cls._next_id += 1\n",
      "        return cls._next_id\n",
      "    \n",
      "    @property\n",
      "    def name(self):\n",
      "        return \"literal_%s\" % self.id\n",
      "        \n",
      "    def __str__(self):\n",
      "        #return \"?%s\" % self.name\n",
      "        return '\"%s\"' % self.literal\n",
      "\n",
      "    \n",
      "class Query(object):\n",
      "    def __init__(self, graph):\n",
      "        self._graph = graph\n",
      "        self._wheres = set()\n",
      "        self._variables = {}\n",
      "        self._literal_bindings = []\n",
      "\n",
      "        self._template = Template(\"\"\"\n",
      "            SELECT {% for var in variables %}{{ var }} {% endfor %}\n",
      "            WHERE {\n",
      "                {% for subj, pred, obj in wheres %}\n",
      "                    {{subj}} {{pred}} {{obj}} .\n",
      "                {% endfor %}}\n",
      "            \"\"\")\n",
      "    \n",
      "    def _query(self, bindings=None):\n",
      "        # What is the query string?\n",
      "        context = dict(\n",
      "            variables=bindings or self._variables.values(),\n",
      "            wheres=sorted(self._wheres)\n",
      "        )\n",
      "        return self._template.render(**context)\n",
      "    \n",
      "    def _term(self, term):\n",
      "        # What should I do with this term?\n",
      "        if isinstance(term, URIRef):\n",
      "            return URITerm(term)\n",
      "        elif isinstance(term, Variable):\n",
      "            return term\n",
      "        else:\n",
      "            term = LiteralBinding(Literal(term))\n",
      "            self._literal_bindings.append(term)\n",
      "            return term\n",
      "    \n",
      "    # MAGIC STUFF!\n",
      "    \n",
      "    def __getattr__(self, name):\n",
      "        return self._variables.setdefault(name, Variable(name))\n",
      "    \n",
      "    def __getitem__(self, sliced):\n",
      "        if isinstance(sliced, slice):\n",
      "            where = tuple([\n",
      "                self._term(term)\n",
      "                for term in (sliced.start, sliced.stop, sliced.step)\n",
      "            ])\n",
      "            self._wheres.add(where)\n",
      "        else:\n",
      "            return URIRef(sliced)\n",
      "        \n",
      "    def __call__(self, *args):\n",
      "        query = self._query(args)\n",
      "        \n",
      "        return self._graph.query(query)\n",
      "    \n",
      "    def __str__(self):\n",
      "        return self._query()\n"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 25
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "Finally, we're ready to make a query! How about \n",
      "> Who are all those actors that have worked with both Ridley Scott and George Lucas?"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "q = Query(graph)\n",
      "\n",
      "q[q.actor : movie.actor_name : q.actor_name]\n",
      "q[q.actor : movie.performance : q.george_movie_performance]\n",
      "q[q.actor : movie.performance : q.ridley_movie_performance]\n",
      "\n",
      "q[q.george : movie.director_name : \"George Lucas\"]\n",
      "q[q.ridley : movie.director_name : \"Ridley Scott\"]\n",
      "\n",
      "q[q.george_movie : movie.performance : q.george_movie_performance]\n",
      "q[q.george_movie : movie.director : q.george]\n",
      "q[q.george_movie : rdfs.label : q.george_movie_name]\n",
      "\n",
      "q[q.ridley_movie : movie.director : q.ridley]\n",
      "q[q.ridley_movie : movie.performance : q.ridley_movie_performance]\n",
      "q[q.ridley_movie : rdfs.label : q.ridley_movie_name]"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 41
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "In the end, we are just writing SPARQL (until something better comes along)."
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "print str(q)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "            SELECT ?actor_name ?george_movie ?george_movie_name ?actor ?ridley_movie_performance ?george_movie_performance ?ridley_movie_name ?ridley_movie ?ridley ?george \n",
        "            WHERE {\n",
        "                \n",
        "                    ?george <http://data.linkedmdb.org/resource/movie/director_name> \"George Lucas\" .\n",
        "                \n",
        "                    ?actor <http://data.linkedmdb.org/resource/movie/actor_name> ?actor_name .\n",
        "                \n",
        "                    ?actor <http://data.linkedmdb.org/resource/movie/performance> ?ridley_movie_performance .\n",
        "                \n",
        "                    ?actor <http://data.linkedmdb.org/resource/movie/performance> ?george_movie_performance .\n",
        "                \n",
        "                    ?ridley_movie <http://data.linkedmdb.org/resource/movie/performance> ?ridley_movie_performance .\n",
        "                \n",
        "                    ?ridley_movie <http://data.linkedmdb.org/resource/movie/director> ?ridley .\n",
        "                \n",
        "                    ?ridley_movie <http://www.w3.org/2000/01/rdf-schema#label> ?ridley_movie_name .\n",
        "                \n",
        "                    ?george_movie <http://www.w3.org/2000/01/rdf-schema#label> ?george_movie_name .\n",
        "                \n",
        "                    ?george_movie <http://data.linkedmdb.org/resource/movie/performance> ?george_movie_performance .\n",
        "                \n",
        "                    ?george_movie <http://data.linkedmdb.org/resource/movie/director> ?george .\n",
        "                \n",
        "                    ?ridley <http://data.linkedmdb.org/resource/movie/director_name> \"Ridley Scott\" .\n",
        "                }\n",
        "            \n"
       ]
      }
     ],
     "prompt_number": 51
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "Sweet, sweet results."
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "results = q(q.actor_name, q.george_movie_name, q.ridley_movie_name)\n",
      "for actor, movie1, movie2 in sorted(results):\n",
      "    print \"%s was in '%s' and '%s'\" % (actor, movie1, movie2)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "Ewan McGregor was in 'Star Wars Episode I: The Phantom Menace' and 'Black Hawk Down'\n",
        "Ewan McGregor was in 'Star Wars Episode II: Attack of the Clones' and 'Black Hawk Down'\n",
        "Ewan McGregor was in 'Star Wars Episode III: Revenge of the Sith' and 'Black Hawk Down'\n",
        "Harrison Ford was in 'American Graffiti' and 'Blade Runner'\n",
        "Harrison Ford was in 'Star Wars Episode IV: A New Hope' and 'Blade Runner'\n",
        "Harrison Ford was in 'The Star Wars Holiday Special' and 'Blade Runner'\n",
        "Liam Neeson was in 'Star Wars Episode I: The Phantom Menace' and 'Kingdom of Heaven'\n"
       ]
      }
     ],
     "prompt_number": 52
    }
   ],
   "metadata": {}
  }
 ]
}