Massive renaming!

32-double-map-reduce/README.md

@@ -0,0 +1,22 @@
+Style #31
+==============================
+
+Very similar to style #30, but with an additional twist
+
+Constraints:
+
+- Input data is divided in chunks, similar to what an inverse multiplexer does to input signals
+
+- A map function applies a given worker function to each chunk of data, potentially in parallel
+
+- The results of the many worker functions are reshuffled in a way
+  that allows for the reduce step to be also parallelized
+
+- The reshuffled chunks of data are given as input to a second map
+  function that takes a reducible function as input
+
+Possible names:
+
+- Map-reduce 
+- Hadoop style
+- Double inverse multiplexer 

32-double-map-reduce/tf-32.py

@@ -0,0 +1,97 @@
+#!/usr/bin/env python
+import sys, re, operator, string
+from functools import reduce
+
+try:
+    xrange          # Python 2
+except NameError:
+    xrange = range  # Python 3
+
+
+#
+# Functions for map reduce
+#
+def partition(data_str, nlines):
+    """
+    Partitions the input data_str (a big string)
+    into chunks of nlines.
+    """
+    lines = data_str.split('\n')
+    for i in xrange(0, len(lines), nlines):
+        yield '\n'.join(lines[i:i+nlines])
+
+def split_words(data_str):
+    """
+    Takes a string, returns a list of pairs (word, 1),
+    one for each word in the input, so
+    [(w1, 1), (w2, 1), ..., (wn, 1)]
+    """
+    def _scan(str_data):
+        pattern = re.compile('[\W_]+')
+        return pattern.sub(' ', str_data).lower().split()
+
+    def _remove_stop_words(word_list):
+        with open('../stop_words.txt') as f:
+            stop_words = f.read().split(',')
+        stop_words.extend(list(string.ascii_lowercase))
+        return [w for w in word_list if not w in stop_words]
+
+    # The actual work of the mapper
+    result = []
+    words = _remove_stop_words(_scan(data_str))
+    for w in words:
+        result.append((w, 1))
+    return result
+
+def regroup(pairs_list):
+    """
+    Takes a list of lists of pairs of the form
+    [[(w1, 1), (w2, 1), ..., (wn, 1)],
+     [(w1, 1), (w2, 1), ..., (wn, 1)],
+     ...]
+    and returns a dictionary mapping each unique word to the
+    corresponding list of pairs, so
+    { w1 : [(w1, 1), (w1, 1)...],
+      w2 : [(w2, 1), (w2, 1)...],
+      ...}
+    """
+    mapping = {}
+    for pairs in pairs_list:
+        for p in pairs:
+            if p[0] in mapping:
+                mapping[p[0]].append(p)
+            else:
+                mapping[p[0]] = [p]
+    return mapping
+
+def count_words(mapping):
+    """
+    Takes a mapping of the form (word, [(word, 1), (word, 1)...)])
+    and returns a pair (word, frequency), where frequency is the
+    sum of all the reported occurrences
+    """
+    def add(x, y):
+        return x+y
+
+    return (mapping[0], reduce(add, (pair[1] for pair in mapping[1])))
+
+#
+# Auxiliary functions
+#
+def read_file(path_to_file):
+    with open(path_to_file) as f:
+        data = f.read()
+    return data
+
+def sort(word_freq):
+    return sorted(word_freq, key=operator.itemgetter(1), reverse=True)
+
+#
+# The main function
+#
+splits = map(split_words, partition(read_file(sys.argv[1]), 200))
+splits_per_word = regroup(splits)
+word_freqs = sort(map(count_words, splits_per_word.items()))
+
+for (w, c) in word_freqs[0:25]:
+    print(w, ' - ', c)