orange-variable.ipynb

{
 "metadata": {
  "name": "",
  "signature": "sha256:cfd825ef318eb3d896cf218060d459df23b04a94c05ea842ebee858bae869d2b"
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "Let's read a small CSV file into Orange:"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from __future__ import print_function\n",
      "import Orange\n",
      "\n",
      "FILE_A = 'tiny.attach.tab'\n",
      "table_a = Orange.data.Table(FILE_A)\n",
      "\n",
      "print(table_a.domain)\n",
      "for inst in table_a:\n",
      "    print(inst)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[id, foo, CLASS]\n",
        "['a1', 'x', 'False']\n",
        "['a2', 'y', 'True']\n",
        "['a3', 'z', 'False']\n"
       ]
      }
     ],
     "prompt_number": 1
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "How nice. We can also do things like ask for the range of possible values associated with a variable"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "class_values = table_a.domain['CLASS']\n",
      "list(class_values)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 2,
       "text": [
        "[<orange.Value 'CLASS'='False'>, <orange.Value 'CLASS'='True'>]"
       ]
      }
     ],
     "prompt_number": 2
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "Splendid! So in the A table domain, the 'CLASS' variable is associated with the values 'True' and 'False'.\n",
      "\n",
      "So shall we look at a second table?"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "FILE_B = 'tiny.relate.tab'\n",
      "table_b = Orange.data.Table(FILE_B)\n",
      "\n",
      "print(table_b.domain)\n",
      "for inst in table_b:\n",
      "    print(inst)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[id, bar, CLASS]\n",
        "['a1', 'b', 'Narration']\n",
        "['a2', 'd', 'Elaboration']\n",
        "['a3', 'e', 'Background']\n"
       ]
      }
     ],
     "prompt_number": 3
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "So far so good, now what do you think should be the values associated with variable 'CLASS' in table B?"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "list(table_b.domain['CLASS'])"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 4,
       "text": [
        "[<orange.Value 'CLASS'='False'>,\n",
        " <orange.Value 'CLASS'='True'>,\n",
        " <orange.Value 'CLASS'='Background'>,\n",
        " <orange.Value 'CLASS'='Elaboration'>,\n",
        " <orange.Value 'CLASS'='Narration'>]"
       ]
      }
     ],
     "prompt_number": 4
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "Wait, what? Where did 'False' and 'True' come from?\n",
      "Hang on, let's check back in table A"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "list(table_a.domain['CLASS'])"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 5,
       "text": [
        "[<orange.Value 'CLASS'='False'>,\n",
        " <orange.Value 'CLASS'='True'>,\n",
        " <orange.Value 'CLASS'='Background'>,\n",
        " <orange.Value 'CLASS'='Elaboration'>,\n",
        " <orange.Value 'CLASS'='Narration'>]"
       ]
      }
     ],
     "prompt_number": 5
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "Huh. Did you know that variable descriptors in Orange were global and of course mutable?  I didn't but [looking more closely at the manual](http://docs.orange.biolab.si/reference/rst/Orange.feature.descriptor.html#Orange.feature.Descriptor), it says quite plainly:\n",
      "\n",
      "> Orange considers two variables (e.g. in two different data tables) the same if they have the same descriptor. It is allowed - but not recommended - to have different descriptors with the same name.\n",
      "\n",
      "This matters particularly if we are use two different tables that have a local 'CLASS' feature, particularly if we're doing things like looking up the probability for 'True' by its presumed index `1` (`\u0ca0_\u0ca0`)"
     ]
    }
   ],
   "metadata": {}
  }
 ]
}

tiny.attach.tab

CLASS	id	foo
d	d	d
class		
False	a1	x
True	a2	y
False	a3	z

tiny.relate.tab

CLASS	id	bar
d	d	d
class		
Narration	a1	b
Elaboration	a2	d
Background	a3	e