Resolved test case failure: test_custom_search (test_examples.ExamplesTestCase)

Switched from float to Decimal computation in tophashtags.py unit test. This ensures uniform precision on all platforms. Without this fix precision is different on OS X 10.8 than it is on CentOS 6.4, OS X 10.9, and Windows Server 2008 R2.

Author: David Noble

examples/twitted/twitted/bin/tophashtags.py

@@ -16,6 +16,7 @@
 
 import csv, sys, urllib, re
 
+
 # Tees output to a logfile for debugging
 class Logger:
     def __init__(self, filename, buf = None):
@@ -36,6 +37,7 @@ def write(self, message):
             self.buf.write(message)
             self.buf.flush()
 
+
 # Tees input as it is being read, also logging it to a file
 class Reader:
     def __init__(self, buf, filename = None):
@@ -65,6 +67,7 @@ def readline(self):
         # Return to the caller
         return line
 
+
 def output_results(results, mvdelim = '\n', output = sys.stdout):
     """Given a list of dictionaries, each representing
     a single result, and an optional list of fields,
@@ -91,6 +94,7 @@ def output_results(results, mvdelim = '\n', output = sys.stdout):
     writer.writerow(dict(zip(fields, fields)))
     writer.writerows(results)
 
+
 def read_input(buf, has_header = True):
     """Read the input from the given buffer (or stdin if no buffer)
     is supplied. An optional header may be present as well"""
@@ -132,6 +136,7 @@ def read_input(buf, has_header = True):
 
     return buf, header
 
+
 def encode_mv(vals):
     """For multivalues, values are wrapped in '$' and separated using ';'
     Literal '$' values are represented with '$$'"""
@@ -144,6 +149,7 @@ def encode_mv(vals):
 
     return s
 
+
 def main(argv):
     stdin_wrapper = Reader(sys.stdin)
     buf, settings = read_input(stdin_wrapper, has_header = True)
@@ -167,12 +173,13 @@ def main(argv):
 
     num_hashtags = sum(hashtags.values())
 
+    from decimal import Decimal
     results = []
     for k, v in hashtags.iteritems():
         results.append({
             "hashtag": k, 
             "count": v, 
-            "percentage": (float(v) / float(num_hashtags))
+            "percentage": (Decimal(v) / Decimal(num_hashtags))
         })
 
     # And output it to the next stage of the pipeline

tests/data/custom_search/tophashtags.baseline

@@ -1,4 +1,4 @@
 count,hashtag,percentage
-1,#2,0.333333333333
-1,#redsox,0.333333333333
-1,#mlb,0.333333333333
+1,#2,0.3333333333333333333333333333
+1,#redsox,0.3333333333333333333333333333
+1,#mlb,0.3333333333333333333333333333