An alternative implementation for PR 3801 in IPython.

PR3801-alternative.ipynb

{
 "metadata": {
  "name": ""
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "# Implementation alternatives for [PR 3801](https://github.com/ipython/ipython/pull/3801)\n",
      "\n",
      "This is the code in the PR:"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "import re\n",
      "\n",
      "START_WITH_MAGIC_PATTERN = re.compile(\"%+\")  # find %\n",
      "\n",
      "def penalize_magics_key(word):\n",
      "    \"\"\"key for sorting that penalizes magic commands in the ordering\n",
      "\n",
      "    Normal words are left alone.\n",
      "\n",
      "    Magic commands have the initial % moved to the end, e.g.\n",
      "    %matplotlib is transformed as follows:\n",
      "\n",
      "    %matplotlib -> matplotlib%\n",
      "\n",
      "    [The choice of the final % is arbitrary.]\n",
      "\n",
      "    Since \"matplotlib\" < \"matplotlib%\" as strings, \n",
      "    \"timeit\" will appear before the magic \"%timeit\" in the ordering\n",
      "\n",
      "    For consistency, move \"%%\" to the end, so cell magics appear *after*\n",
      "    line magics with the same name.\n",
      "\n",
      "    \"\"\"\n",
      "\n",
      "    # Move all % to end of the *key* (don't change the actual text, of course)\n",
      "\n",
      "    m = START_WITH_MAGIC_PATTERN.match(word) # at the start of the string\n",
      "\n",
      "    if m != None:  # if there is a match\n",
      "        word = word[m.end():] + m.group()  # move all %'s to the end\n",
      "       \n",
      "    return word"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 19
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "This alternative uses only string methods:"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "def penalize_magics_key2(word):\n",
      "    i = word.rfind('%')\n",
      "    if i in (0,1):\n",
      "        s = i+1\n",
      "        return word[s:]+word[:s]\n",
      "    else:\n",
      "        return word"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 43
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "Sanity checks:"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "for s in ['%foo', '%%foo', '%foo%', 'foo%%']:\n",
      "    a = penalize_magics_key(s)\n",
      "    b = penalize_magics_key2(s)\n",
      "    if a!=b:\n",
      "        print 'Ori:', s, 're:', a,'str:', b"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "Ori: %foo% re: foo%% str: %foo%\n"
       ]
      }
     ],
     "prompt_number": 44
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "Note that the new version is actually better in terms of behavior, since it only transforms pure magic calls and not other strings."
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "Performance checks:"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "a  = '%%matplotlib'\n",
      "%timeit penalize_magics_key(a)\n",
      "%timeit penalize_magics_key2(a)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "1000000 loops, best of 3: 975 ns per loop\n",
        "1000000 loops, best of 3: 552 ns per loop"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n"
       ]
      }
     ],
     "prompt_number": 46
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "The string-based code is also almost twice as fast."
     ]
    }
   ],
   "metadata": {}
  }
 ]
}