4-On-Disk-Tables.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#Querying On-Disk Tables"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> Objectives:\n",
    "> * Compare queries of tabular data for **on-disk** containers\n",
    "> * Compare sizes and times for those"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [1] used 0.043 MiB RAM in 0.002s, peaked 0.000 MiB above current, total RAM usage 30.840 MiB\n"
     ]
    }
   ],
   "source": [
    "from ipython_memwatcher import MemWatcher\n",
    "mw = MemWatcher()\n",
    "mw.start_watching_memory()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load movielens datasets in pandas DataFrames"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [2] used 0.023 MiB RAM in 0.004s, peaked 0.000 MiB above current, total RAM usage 30.863 MiB\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "dset = 'movielens-1m'\n",
    "fdata = os.path.join(dset, 'ratings.dat.gz')\n",
    "fitem = os.path.join(dset, 'movies.dat')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [3] used 77.191 MiB RAM in 0.727s, peaked 0.367 MiB above current, total RAM usage 108.055 MiB\n"
     ]
    }
   ],
   "source": [
    "# Import CSV files via pandas\n",
    "import pandas as pd\n",
    "# pass in column names for each CSV\n",
    "r_cols = ['user_id', 'movie_id', 'rating', 'unix_timestamp']\n",
    "ratings = pd.read_csv(fdata, sep=';', names=r_cols, compression='gzip')\n",
    "\n",
    "m_cols = ['movie_id', 'title', 'genres']\n",
    "movies = pd.read_csv(fitem, sep=';', names=m_cols,\n",
    "                     dtype={'title': object, 'genres': object})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [4] used 117.957 MiB RAM in 4.738s, peaked 122.797 MiB above current, total RAM usage 226.012 MiB\n"
     ]
    }
   ],
   "source": [
    "# Store movies and ratings in 2 separate tables in SQLite\n",
    "sqlite_norm = \"movielens-norm.sqlite\"\n",
    "if os.path.exists(sqlite_norm):\n",
    "    os.unlink(sqlite_norm)\n",
    "import sqlite3\n",
    "conn = sqlite3.connect(sqlite_norm)\n",
    "conn.text_factory = str   # Shut up problems with Unicode\n",
    "ratings.to_sql(\"ratings\", conn)\n",
    "movies.to_sql(\"movies\", conn)\n",
    "conn.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [5] used 85.109 MiB RAM in 0.151s, peaked 0.000 MiB above current, total RAM usage 311.121 MiB\n"
     ]
    }
   ],
   "source": [
    "# create one merged DataFrame\n",
    "lens = pd.merge(movies, ratings)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 1000209 entries, 0 to 1000208\n",
      "Data columns (total 6 columns):\n",
      "movie_id          1000209 non-null int64\n",
      "title             1000209 non-null object\n",
      "genres            1000209 non-null object\n",
      "user_id           1000209 non-null int64\n",
      "rating            1000209 non-null int64\n",
      "unix_timestamp    1000209 non-null int64\n",
      "dtypes: int64(4), object(2)\n",
      "memory usage: 53.4+ MB\n",
      "In [6] used 61.055 MiB RAM in 0.225s, peaked 0.000 MiB above current, total RAM usage 372.176 MiB\n"
     ]
    }
   ],
   "source": [
    "lens.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [7] used 46.141 MiB RAM in 5.776s, peaked 107.930 MiB above current, total RAM usage 418.316 MiB\n"
     ]
    }
   ],
   "source": [
    "# Store lens denormalized in 1 table in SQLite\n",
    "sqlite_denorm = \"movielens-denorm.sqlite\"\n",
    "if os.path.exists(sqlite_denorm):\n",
    "    os.unlink(sqlite_denorm)\n",
    "conn_denorm = sqlite3.connect(sqlite_denorm)\n",
    "conn_denorm.text_factory = str   # Shut up problems with Unicode\n",
    "lens.to_sql(\"lens\", conn_denorm)\n",
    "conn_denorm.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [8] used 0.000 MiB RAM in 0.002s, peaked 0.000 MiB above current, total RAM usage 418.316 MiB\n"
     ]
    }
   ],
   "source": [
    "del movies, ratings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 81.7 ms, sys: 121 µs, total: 81.8 ms\n",
      "Wall time: 82.1 ms\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "5121      75\n",
       "5164    3842\n",
       "5187    6031\n",
       "Name: user_id, dtype: int64"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [9] used 0.062 MiB RAM in 0.093s, peaked 0.000 MiB above current, total RAM usage 418.379 MiB\n"
     ]
    }
   ],
   "source": [
    "%time result = lens.query(\"(title == 'Tom and Huck (1995)') & (rating == 5)\")['user_id']\n",
    "result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [10] used 0.000 MiB RAM in 0.002s, peaked 0.000 MiB above current, total RAM usage 418.379 MiB\n"
     ]
    }
   ],
   "source": [
    "# keep the run-time for reference\n",
    "qtime_pandas_mem = mw.measurements.time_delta"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Querying on-disk data with SQLite (relational database)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [11] used 0.000 MiB RAM in 0.004s, peaked 0.000 MiB above current, total RAM usage 418.379 MiB\n"
     ]
    }
   ],
   "source": [
    "# Execute the query with the de-normalized SQLite database\n",
    "conn = sqlite3.connect(sqlite_denorm)\n",
    "c = conn.cursor()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 218 ms, sys: 19.6 ms, total: 238 ms\n",
      "Wall time: 236 ms\n",
      "In [12] used 0.000 MiB RAM in 0.241s, peaked 0.000 MiB above current, total RAM usage 418.379 MiB\n"
     ]
    }
   ],
   "source": [
    "q = \"select user_id from lens where title == 'Tom and Huck (1995)' and rating == 5\"\n",
    "%time result = [r for r in c.execute(q)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(75,), (3842,), (6031,)]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [13] used 0.000 MiB RAM in 0.008s, peaked 0.000 MiB above current, total RAM usage 418.379 MiB\n"
     ]
    }
   ],
   "source": [
    "# keep the run-time for reference\n",
    "qtime_sqlite_denorm = mw.measurements.time_delta\n",
    "# ...and print the result\n",
    "result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [14] used 0.000 MiB RAM in 0.006s, peaked 0.000 MiB above current, total RAM usage 418.379 MiB\n"
     ]
    }
   ],
   "source": [
    "# Execute the query with the normalized SQLite database\n",
    "conn = sqlite3.connect(sqlite_norm)\n",
    "c = conn.cursor()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 2.43 s, sys: 1.15 s, total: 3.57 s\n",
      "Wall time: 3.57 s\n",
      "In [15] used 0.000 MiB RAM in 3.575s, peaked 0.000 MiB above current, total RAM usage 418.379 MiB\n"
     ]
    }
   ],
   "source": [
    "q = \"\"\"\n",
    "select ratings.user_id from movies \n",
    "INNER JOIN ratings ON movies.movie_id = ratings.movie_id\n",
    "where movies.title == 'Tom and Huck (1995)' and ratings.rating == 5\n",
    "\"\"\"\n",
    "%time result = [r for r in c.execute(q)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(75,), (3842,), (6031,)]"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [16] used 0.000 MiB RAM in 0.007s, peaked 0.000 MiB above current, total RAM usage 418.379 MiB\n"
     ]
    }
   ],
   "source": [
    "# keep the run-time for reference\n",
    "qtime_sqlite_norm = mw.measurements.time_delta\n",
    "# ...and print the result\n",
    "result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So in general, we see that it is much faster to query tables in denormalized form, although they take much more storage:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-rw-r--r-- 1 faltet francesc,faltet 78M jul 22 06:24 movielens-denorm.sqlite\n",
      "-rw-r--r-- 1 faltet francesc,faltet 37M jul 22 06:22 movielens-norm.sqlite\n",
      "In [17] used 0.172 MiB RAM in 0.126s, peaked 0.000 MiB above current, total RAM usage 418.551 MiB\n"
     ]
    }
   ],
   "source": [
    "!ls -lh $sqlite_norm $sqlite_denorm"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In some way, storing tables in normalized form is a kind of compression, but that comes to the cost of using more time to process queries.  Let's enter another way to compress denormalized data tables on-disk:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## bcolz"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`bcolz` cannot only deal with data in-memory, but also on-disk using exactly the same API.  Let's try that."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=\n",
      "bcolz version:     0.9.0\n",
      "NumPy version:     1.9.2\n",
      "Blosc version:     1.4.1 ($Date:: 2014-07-08 #$)\n",
      "Blosc compressors: ['blosclz', 'lz4', 'lz4hc', 'snappy', 'zlib']\n",
      "Numexpr version:   2.4.3\n",
      "Python version:    2.7.10 |Anaconda 2.1.0 (64-bit)| (default, May 28 2015, 17:02:03) \n",
      "[GCC 4.4.7 20120313 (Red Hat 4.4.7-1)]\n",
      "Platform:          linux2-x86_64\n",
      "Byte-ordering:     little\n",
      "Detected cores:    4\n",
      "-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=\n",
      "In [18] used 5.422 MiB RAM in 0.100s, peaked 0.000 MiB above current, total RAM usage 423.973 MiB\n"
     ]
    }
   ],
   "source": [
    "import bcolz\n",
    "bcolz.print_versions()\n",
    "bcolz.defaults.cparams['cname'] = 'lz4'\n",
    "bcolz.defaults.cparams['clevel'] = 5\n",
    "# bcolz.set_nthreads(4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [19] used 0.324 MiB RAM in 0.361s, peaked 77.980 MiB above current, total RAM usage 424.297 MiB\n"
     ]
    }
   ],
   "source": [
    "# Import a pandas DataFrame in a ctable on-disk\n",
    "bcolz_dir = \"movielens-denorm.bcolz\"\n",
    "if os.path.exists(bcolz_dir):\n",
    "    import shutil\n",
    "    shutil.rmtree(bcolz_dir)\n",
    "zlens = bcolz.ctable.fromdataframe(lens, rootdir=bcolz_dir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 105 ms, sys: 44.3 ms, total: 149 ms\n",
      "Wall time: 111 ms\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([  75, 3842, 6031])"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [20] used 0.070 MiB RAM in 0.118s, peaked 0.000 MiB above current, total RAM usage 424.367 MiB\n"
     ]
    }
   ],
   "source": [
    "%time result = zlens[\"(title == b'Tom and Huck (1995)') & (rating == 5)\"]['user_id']\n",
    "result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([  75, 3842, 6031])"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [21] used 0.000 MiB RAM in 0.005s, peaked 0.000 MiB above current, total RAM usage 424.367 MiB\n"
     ]
    }
   ],
   "source": [
    "# keep the run-time for reference\n",
    "qtime_bcolz = mw.measurements.time_delta\n",
    "# ...and print the result (for reference too)\n",
    "result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 84.9 ms, sys: 29.4 ms, total: 114 ms\n",
      "Wall time: 84.6 ms\n",
      "In [22] used 0.000 MiB RAM in 0.091s, peaked 0.000 MiB above current, total RAM usage 424.367 MiB\n"
     ]
    }
   ],
   "source": [
    "# One can optimize the query more\n",
    "%time result = [r.user_id for r in zlens.where(\"(title == b'Tom and Huck (1995)') & (rating == 5)\", outcols=['user_id'])]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[75, 3842, 6031]"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [23] used 0.000 MiB RAM in 0.005s, peaked 0.000 MiB above current, total RAM usage 424.367 MiB\n"
     ]
    }
   ],
   "source": [
    "# keep the run-time for reference\n",
    "qtime_bcolz_opt = mw.measurements.time_delta\n",
    "# ...and print the result (for reference too)\n",
    "result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's see the space consumed on-disk:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6,6M\tmovielens-denorm.bcolz\n",
      "In [24] used 0.000 MiB RAM in 0.124s, peaked 0.000 MiB above current, total RAM usage 424.367 MiB\n"
     ]
    }
   ],
   "source": [
    "!du -sh $bcolz_dir"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "and let's compare it with the internal estimated size:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "ctable((1000209,), [('movie_id', '<i8'), ('title', 'S82'), ('genres', 'S47'), ('user_id', '<i8'), ('rating', '<i8'), ('unix_timestamp', '<i8')])\n",
       "  nbytes: 153.57 MB; cbytes: 7.89 MB; ratio: 19.45\n",
       "  cparams := cparams(clevel=5, shuffle=True, cname='lz4')\n",
       "  rootdir := 'movielens-denorm.bcolz'\n",
       "[(1, 'Toy Story (1995)', \"Animation|Children's|Comedy\", 1, 5, 978824268)\n",
       " (1, 'Toy Story (1995)', \"Animation|Children's|Comedy\", 6, 4, 978237008)\n",
       " (1, 'Toy Story (1995)', \"Animation|Children's|Comedy\", 8, 4, 978233496)\n",
       " ...,\n",
       " (3952, 'Contender, The (2000)', 'Drama|Thriller', 5837, 4, 1011902656)\n",
       " (3952, 'Contender, The (2000)', 'Drama|Thriller', 5927, 1, 979852537)\n",
       " (3952, 'Contender, The (2000)', 'Drama|Thriller', 5998, 4, 1001781044)]"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [25] used 0.008 MiB RAM in 0.011s, peaked 0.000 MiB above current, total RAM usage 424.375 MiB\n"
     ]
    }
   ],
   "source": [
    "zlens"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So, the internal estimated size is a bit higher than the actual size on-disk, which means that the former is in general a conservative guess.\n",
    "\n",
    "Another interesting fact is that the de-normalized version of the ctable takes less space than the normalized version in a relational database, and the query speed is still faster (than SQLite at least).\n",
    "\n",
    "But there are other ways to store compressed tables.  Let's visit PyTables and how it performs:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## PyTables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [26] used 0.766 MiB RAM in 0.558s, peaked 0.000 MiB above current, total RAM usage 425.141 MiB\n"
     ]
    }
   ],
   "source": [
    "h5denorm = \"movielens-denorm.h5\"\n",
    "if os.path.exists(h5denorm):\n",
    "    os.unlink(h5denorm)\n",
    "zlens.tohdf5(h5denorm, nodepath='/h5lens')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [27] used 0.035 MiB RAM in 0.013s, peaked 0.000 MiB above current, total RAM usage 425.176 MiB\n"
     ]
    }
   ],
   "source": [
    "import tables\n",
    "h5file = tables.open_file(h5denorm)\n",
    "h5lens = h5file.root.h5lens"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "/h5lens (Table(1000209,), shuffle, blosc:lz4(5)) ''\n",
       "  description := {\n",
       "  \"movie_id\": Int64Col(shape=(), dflt=0, pos=0),\n",
       "  \"title\": StringCol(itemsize=82, shape=(), dflt='', pos=1),\n",
       "  \"genres\": StringCol(itemsize=47, shape=(), dflt='', pos=2),\n",
       "  \"user_id\": Int64Col(shape=(), dflt=0, pos=3),\n",
       "  \"rating\": Int64Col(shape=(), dflt=0, pos=4),\n",
       "  \"unix_timestamp\": Int64Col(shape=(), dflt=0, pos=5)}\n",
       "  byteorder := 'little'\n",
       "  chunkshape := (407,)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [28] used 0.004 MiB RAM in 0.007s, peaked 0.000 MiB above current, total RAM usage 425.180 MiB\n"
     ]
    }
   ],
   "source": [
    "# Let's have a look at the HDF5 table\n",
    "h5lens"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 353 ms, sys: 15.7 ms, total: 368 ms\n",
      "Wall time: 340 ms\n",
      "In [29] used 0.020 MiB RAM in 0.346s, peaked 0.000 MiB above current, total RAM usage 425.199 MiB\n"
     ]
    }
   ],
   "source": [
    "%time result = [r['user_id'] for r in h5lens.where(\"(title == b'Tom and Huck (1995)') & (rating == 5)\")]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[75, 3842, 6031]"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [30] used 0.004 MiB RAM in 0.006s, peaked 0.000 MiB above current, total RAM usage 425.203 MiB\n"
     ]
    }
   ],
   "source": [
    "# keep the run-time for reference\n",
    "qtime_pytables = mw.measurements.time_delta\n",
    "# ...and print the result (for reference too)\n",
    "result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "So, it seems that PyTables has a similar performance than SQLite for the queries (denormalized case).  What about the size of the HDF5 file?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-rw-r--r-- 1 faltet francesc,faltet 7,1M jul 22 06:31 movielens-denorm.h5\n",
      "In [31] used 0.004 MiB RAM in 0.127s, peaked 0.000 MiB above current, total RAM usage 425.207 MiB\n"
     ]
    }
   ],
   "source": [
    "h5file.close()\n",
    "!ls -lh $h5denorm"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "7.1 MB is very close to 6.6 MB that used bcolz (which is expected because both are using LZ4 as the compressor), and much less than pandas and SQLite.\n",
    "\n",
    "Now, pandas comes with its own interface to PyTables via the HDFStore sub-package.  Let's check that out."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## HDFStore (pandas HDF5 via PyTables)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [32] used -10.055 MiB RAM in 11.207s, peaked 103.492 MiB above current, total RAM usage 415.152 MiB\n"
     ]
    }
   ],
   "source": [
    "h5pandas = \"movielens-pandas.h5\"\n",
    "if os.path.exists(h5pandas):\n",
    "    os.unlink(h5pandas)\n",
    "from pandas import HDFStore\n",
    "hdf = HDFStore(h5pandas, complevel=5, complib=\"blosc:lz4\")\n",
    "hdf.put('h5lens', lens, format='table', data_columns=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 54.3 ms, sys: 196 µs, total: 54.5 ms\n",
      "Wall time: 52 ms\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div style=\"max-height:1000px;max-width:1500px;overflow:auto;\">\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>user_id</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>5121</th>\n",
       "      <td>75</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5164</th>\n",
       "      <td>3842</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5187</th>\n",
       "      <td>6031</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      user_id\n",
       "5121       75\n",
       "5164     3842\n",
       "5187     6031"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [33] used 0.066 MiB RAM in 0.062s, peaked 0.000 MiB above current, total RAM usage 415.219 MiB\n"
     ]
    }
   ],
   "source": [
    "%time hdf.select('h5lens', where=\"(title == 'Tom and Huck (1995)') & (rating == 5)\", columns=[\"user_id\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[75, 3842, 6031]"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [34] used 0.000 MiB RAM in 0.006s, peaked 0.000 MiB above current, total RAM usage 415.219 MiB\n"
     ]
    }
   ],
   "source": [
    "# keep the run-time for reference\n",
    "qtime_pandas_hdf5 = mw.measurements.time_delta\n",
    "# ...and print the result (for reference too)\n",
    "result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As we can see, the time is very close to the time that it takes the query on a pandas DataFrame in memory.  What about the size on-disk?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [35] used 0.004 MiB RAM in 0.221s, peaked 0.000 MiB above current, total RAM usage 415.223 MiB\n"
     ]
    }
   ],
   "source": [
    "hdf.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-rw-r--r-- 1 faltet francesc,faltet 13M jul 22 06:36 movielens-pandas.h5\n",
      "In [36] used 0.102 MiB RAM in 0.127s, peaked 0.000 MiB above current, total RAM usage 415.324 MiB\n"
     ]
    }
   ],
   "source": [
    "!ls -lh $h5pandas"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ok, so 13 MB is around 80% larger than a pure PyTables file with the same info (7.1 MB).  Where the overhead comes from?  Well, it turns out that pandas makes use of the indexing capabilities of PyTables automatically, so let's make PyTables to index the columns that participate in the query."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Indexing (PyTables)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [37] used 0.000 MiB RAM in 0.026s, peaked 0.000 MiB above current, total RAM usage 415.324 MiB\n"
     ]
    }
   ],
   "source": [
    "## Copy the original PyTables table into another file\n",
    "import shutil\n",
    "h5idx = \"movielens-indexed.h5\"\n",
    "if os.path.exists(h5idx):\n",
    "    os.unlink(h5idx)\n",
    "shutil.copyfile(h5denorm, h5idx)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [38] used 0.000 MiB RAM in 0.004s, peaked 0.000 MiB above current, total RAM usage 415.324 MiB\n"
     ]
    }
   ],
   "source": [
    "# Open the new file in 'a'ppend mode\n",
    "h5i = tables.open_file(h5idx, mode=\"a\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1000209"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [39] used 0.000 MiB RAM in 4.484s, peaked 161.793 MiB above current, total RAM usage 415.324 MiB\n"
     ]
    }
   ],
   "source": [
    "# Create an index for the 'title' column\n",
    "h5lens = h5i.root.h5lens\n",
    "h5lens.cols.title.create_csindex()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1 loops, best of 1: 8.48 ms per loop\n",
      "In [40] used 0.008 MiB RAM in 0.016s, peaked 0.000 MiB above current, total RAM usage 415.332 MiB\n"
     ]
    }
   ],
   "source": [
    "# ...and redo the query...\n",
    "t = %timeit -r1 -n1 -o result = [r['user_id'] for r in h5lens.where(\"(title == b'Tom and Huck (1995)') & (rating == 5)\")]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[75, 3842, 6031]"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [41] used 0.004 MiB RAM in 0.007s, peaked 0.000 MiB above current, total RAM usage 415.336 MiB\n"
     ]
    }
   ],
   "source": [
    "# keep the run-time for reference\n",
    "qtime_pytables_index1 = t.best\n",
    "# ...and print the result (for reference too)\n",
    "result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So, by indexing one column we have got a 30x of acceleration wrt PyTables (11 vs 350 ms).  What happens if we index the 'rating' column?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1000209"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [42] used 0.008 MiB RAM in 0.918s, peaked 0.000 MiB above current, total RAM usage 415.344 MiB\n"
     ]
    }
   ],
   "source": [
    "h5lens.cols.rating.create_csindex()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1 loops, best of 1: 2.9 ms per loop\n",
      "In [43] used 0.004 MiB RAM in 0.010s, peaked 0.000 MiB above current, total RAM usage 415.348 MiB\n"
     ]
    }
   ],
   "source": [
    "t = %timeit -r1 -n1 -o result = [r['user_id'] for r in h5lens.where(\"(title == b'Tom and Huck (1995)') & (rating == 5)\")]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[75, 3842, 6031]"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [44] used 0.008 MiB RAM in 0.006s, peaked 0.000 MiB above current, total RAM usage 415.355 MiB\n"
     ]
    }
   ],
   "source": [
    "# keep the run-time for reference\n",
    "qtime_pytables_index2 = t.best\n",
    "# ...and print the result (for reference too)\n",
    "result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ok, that's another ~5x additional acceleration, and the best time that we ever reached for this query."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [45] used 0.000 MiB RAM in 0.159s, peaked 0.000 MiB above current, total RAM usage 415.355 MiB\n"
     ]
    }
   ],
   "source": [
    "h5i.close()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##Indexing (SQLite)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We are of course curious to find out how indexing in relational databases fares against PyTables' own indexing."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [58] used 0.000 MiB RAM in 0.142s, peaked 0.000 MiB above current, total RAM usage 428.637 MiB\n"
     ]
    }
   ],
   "source": [
    "sqlite_idx = \"movielens-indexed.sqlite\"\n",
    "if os.path.exists(sqlite_idx):\n",
    "    os.unlink(sqlite_idx)\n",
    "shutil.copyfile(sqlite_denorm, sqlite_idx)\n",
    "conn_idx = sqlite3.connect(sqlite_idx)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [59] used 0.000 MiB RAM in 3.471s, peaked 0.000 MiB above current, total RAM usage 428.637 MiB\n"
     ]
    }
   ],
   "source": [
    "c = conn_idx.cursor()\n",
    "c.execute(\"CREATE INDEX index_title ON lens (title)\")\n",
    "conn_idx.commit()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1 loops, best of 1: 964 µs per loop\n",
      "In [60] used 0.000 MiB RAM in 0.008s, peaked 0.000 MiB above current, total RAM usage 428.637 MiB\n"
     ]
    }
   ],
   "source": [
    "q = \"select user_id from lens where title == 'Tom and Huck (1995)' and rating == 5\"\n",
    "t = %timeit -r1 -n1 -o result = [r for r in c.execute(q)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[75, 3842, 6031]"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [61] used 0.000 MiB RAM in 0.008s, peaked 0.000 MiB above current, total RAM usage 428.637 MiB\n"
     ]
    }
   ],
   "source": [
    "# keep the run-time for reference\n",
    "qtime_sqlite_index1 = t.best\n",
    "# ...and print the result (for reference too)\n",
    "result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Wow, with ~1 ms we are getting the best figure so far, and faster than the best figure with indexed PyTables."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##Exercise\n",
    "\n",
    "What will happen if index the rating column?  We should get better speed, right?  Try it out and report the improvement that you are seeing."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###Solution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [62] used 0.000 MiB RAM in 1.813s, peaked 0.000 MiB above current, total RAM usage 428.637 MiB\n"
     ]
    }
   ],
   "source": [
    "c.execute(\"CREATE INDEX index_rating ON lens (rating)\")\n",
    "conn_idx.commit()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1 loops, best of 1: 212 ms per loop\n",
      "In [63] used 0.000 MiB RAM in 0.218s, peaked 0.000 MiB above current, total RAM usage 428.637 MiB\n"
     ]
    }
   ],
   "source": [
    "q = \"select user_id from lens where title == 'Tom and Huck (1995)' and rating == 5\"\n",
    "t = %timeit -r1 -n1 -o result = [r for r in c.execute(q)]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "Uh, we are getting **far** worse performance with 2 indexes.  Take away lesson: indexing is a tricky thing and the best way to assess the improvement in performance is by trying it out, so do not blindly index *everthing*."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Final performance comparison"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [64] used 0.000 MiB RAM in 0.005s, peaked 0.000 MiB above current, total RAM usage 428.637 MiB\n"
     ]
    }
   ],
   "source": [
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f9afe1e9990>"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA68AAAFKCAYAAADyjmnxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f9afdf0f390>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [65] used -37.348 MiB RAM in 0.414s, peaked 37.348 MiB above current, total RAM usage 391.289 MiB\n"
     ]
    }
   ],
   "source": [
    "# Query times\n",
    "labels = [\"pandas (in-memory)\", \"bcolz\", \"PyTables\", \"SQLite denorm\", \"pandas w/ PyTables (indexed)\", \"PyTables (indexed)\", \"SQLite (indexed)\"]\n",
    "df = pd.DataFrame({'time (sec)': [qtime_pandas_mem, qtime_bcolz_opt, qtime_pytables, qtime_sqlite_denorm, qtime_pandas_hdf5, qtime_pytables_index2, qtime_sqlite_index1]}, index=labels)\n",
    "pd.options.display.mpl_style = 'default'\n",
    "df.plot(kind='barh', figsize=(12,5), fontsize=16, title=\"Query times for MovieLens 1m\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Final size comparison"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "119M\tmovielens-indexed.sqlite\n",
      "78M\tmovielens-denorm.sqlite\n",
      "37M\tmovielens-norm.sqlite\n",
      "13M\tmovielens-pandas.h5\n",
      "11M\tmovielens-indexed.h5\n",
      "7,1M\tmovielens-denorm.h5\n",
      "6,6M\tmovielens-denorm.bcolz\n",
      "5,8M\tmovielens-1m\n",
      "In [66] used 0.000 MiB RAM in 0.133s, peaked 0.000 MiB above current, total RAM usage 391.289 MiB\n"
     ]
    }
   ],
   "source": [
    "!du -sh movielens* | sort -nr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f9afde8f410>"
      ]
     },
     "execution_count": 67,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA6wAAAFKCAYAAAAZudLyAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f9b06da6710>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In [67] used 0.883 MiB RAM in 0.295s, peaked 0.000 MiB above current, total RAM usage 392.172 MiB\n"
     ]
    }
   ],
   "source": [
    "# Container sizes\n",
    "df = pd.DataFrame({'size (MB)': [53, 6.6, 7.1, 78, 13, 11, 119]}, index=labels)\n",
    "pd.options.display.mpl_style = 'default'\n",
    "df.plot(kind='barh', figsize=(12,5), fontsize=16, title=\"Container sizes for MovieLens 1m\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# Final Overview"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Fortunately, there are many different solutions for storing large datasets both in-memory and on-disk.  Here it is a summary of the ones that we have seen in this tutorial:\n",
    "\n",
    "* In-core\n",
    "  * Python lists and dictionaries: Included in Python.  Very flexible.  Not efficient for large datasets.\n",
    "  * NumPy: Good for large amounts of data and with a lot of functionality.  Does not support compression.\n",
    "  * Pandas: Good for large amounts of data and with a lot of functionality.  Does not support compression.\n",
    "  * bcolz: Good for large amounts of data, but reduced functionality.  Supports compression.\n",
    "  \n",
    "* Out-of-core\n",
    "  * Pandas/HDFStore: Good for large datasets.  Supports indexing and compression.  Medium query speeds.\n",
    "  * PyTables: Good for large datasets.  Supports indexing and compression.  Good query speeds.\n",
    "  * bcolz: Good for large datasets.  Supports compression, but not indexing.  Medium query speeds.\n",
    "  * SQLite: Can be used for large datasets, but requires lots of storage.  Supports indexing.  Excellent query speeds."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As you see, every solution has strengths and weaknesses, with wild variations in resource consumptions, so a wise thing to do is to know them better and try to apply the best candidate to your scenario."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Hope that helped and **enjoy data**."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}