tinysort 0.1

The Word Count Example

Currently the only MapReduce implementation is in-memory and serial, but an implementation with parallelized map and reduce phases will be added.

from collections import Counter
import json
import re
import sys

from tinymr.memory import MemMapReduce


class WordCount(MemMapReduce):

    """
    The go-to MapReduce example.

    Don't worry, a better example is on its way:
    https://github.com/geowurster/tinymr/issues/11
    """

    # Counting word occurrence does not benefit from sorting post-map or
    # post-reduce and our `final_reducer()` doesn't care about key order
    # so we can disable sorting for a speed boost.
    sort_map = False
    sort_reduce = False
    sort_final_reduce = False

    def __init__(self):

        """
        Stash a regex to strip off punctuation so we can grab it later.
        """

        self.pattern = re.compile('[\W_]+')

    def mapper(self, line):

        """
        Take a line of text from the input file and figure out how many
        times each word appears.

        An alternative, simpler implementation would be:

            def mapper(self, item):
                for word in item.split():
                    word = self.pattern.sub('', word)
                    if word:
                        yield word.lower(), 1

        This simpler example is more straightforward, but holds more data
        in-memory.  The implementation below does more work per line but
        potentially has a smaller memory footprint.  Like anything
        MapReduce the implementation benefits greatly from knowing a lot
        about the input data.
        """

        # Normalize all words to lowercase
        line = line.lower().strip()

        # Strip off punctuation
        line = [self.pattern.sub('', word) for word in line]

        # Filter out empty strings
        line = filter(lambda x: x != '', line)

        # Return an iterable of `(word, count)` pairs
        return Counter(line).items()

    def reducer(self, key, values):

        """
        Just sum the number of times each word appears in the entire
        dataset.

        At this phase `key` is a word and `values` is an iterable producing
        all of the values for that word from the map phase.  Something like:

            key = 'the'
            values = (1, 1, 2, 2, 1)

        The word `the` appeared once on 3 lines and twice on two lines for
        a total of `7` occurrences, so we yield:

            ('the', 7)
        """

        yield key, sum(values)

    def output(self, pairs):

        """
        Normally this phase is where the final dataset is written to disk,
        but since we're operating in-memory we just want to re-structure as
        a dictionary.

        `pairs` is an iterator producing `(key, iter(values))` tuples from
        the reduce phase, and since we know that we only produced one key
        from each reduce we want to extract it for easier access later.
        """

        return {k: tuple(v)[0] for k, v in pairs}


wc = WordCount()
with open('LICENSE.txt') as f:
    out = wc(f)
    print(json.dumps(out, indent=4, sort_keys=True))

Truncated output:

{
    "a": 1,
    "above": 2,
    "advised": 1,
    "all": 1,
    "and": 8,
    "andor": 1
}

$ head -10 LICENSE.txt

New BSD License

Copyright (c) 2015, Kevin D. Wurster
All rights reserved.

Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:

* Redistributions of source code must retain the above copyright notice, this
  list of conditions and the following disclaimer.

# Input line
line = 'Copyright (c) 2015, Kevin D. Wurster'

# Sanitized words
words = ['Copyright', 'c', '2015', 'Kevin', 'D', 'Wurster']

# Return tuples with word as the first element and count as the second
pairs = [('Copyright', 1), ('c', 1), ('2015', 1), ('Kevin', 1), ('D', 1), ('Wurster', 1)]

{
    '2015': [(1,)]
    'above': [(1,)]
    'all': [(1,)]
    'and': [(1,), (1,), (1,)]
    'are': [(1,), (1,)]
    'binary': [(1,)]
    'bsd': [(1,)]
    'c': [(1,)]
    'code': [(1,)]
}

# The ``reducer()`` receives a key and an iterator of values
key = 'the'
values = (1, 1, 1)
def reducer(key, values):
    yield key, sum(values)

{
    '2015': [(1,)]
    'above': [(1,)]
    'all': [(1,)]
    'and': [(3,)]
    'are': [(2,)]
    'binary': [(1,)]
    'bsd': [(1,)]
    'c': [(1,)]
    'code': [(1,)]
}

values = [
    ('the', (3,)),
    ('in', (1,),
]

return {k: tuple(v)[0] for k, v in values}

Developing

$ git clone https://github.com/geowurster/tinymr.git
$ cd tinymr
$ pip install -e .\[dev\]
$ py.test tests --cov tinymr --cov-report term-missing

License

See LICENSE.txt

Changelog

See CHANGES.md