add experiments directory + compare_approaches

data_construction/oracle_feature_envy/createLabels.py

@@ -4,6 +4,11 @@
 import fnmatch
 import os
 
+# This script is used to automatically create the labels from the answers collected by our survey.
+# It implements a vote descision between the different answers for a same instance.
+# For each system, it creates a file "data/labels/feature_envy/system-name.csv" containing all the 
+# instances of feature envy detected by our survey.
+
 systems = [
 		{
 		'name' : 'android-frameworks-opt-telephony',
@@ -82,7 +87,7 @@ def getCandidates():
 	smells = [candidates[i] for i in range(startIndex, endIndex) if getScore(answers[i]) >= 0.5]
 	startIndex = endIndex
 
-	labelFile = os.path.join(ROOT_DIR, 'data/labels/feature_envy/' + system['name'] + '.csv')
+	labelFile = os.path.join(ROOT_DIR, 'data/labels/feature_envy/' + system['name'] + '.txt')
 
 	with open(labelFile, 'w') as file:
 		for smell in smells:

detection_tools/confidence_metrics.py

@@ -152,7 +152,7 @@ def getHistFECM(systemName):
 
     FECM = {}
     for i, j in zip(*np.where(coOcc > 0)):
-        if classes[j] != entityUtils.getEmbeddingClass(methods[i])
+        if classes[j] != entityUtils.getEmbeddingClass(methods[i]):
             instanceName = methods[i] + ';' + classes[j]
             FECM[instanceName] = coOcc[i,j]
 

experiments/evaluate.py

@@ -0,0 +1,27 @@
+from __future__ import division
+
+
+def recall(detected, true):	
+	truePositive = [entity for entity in detected if entity in true]
+
+	if len(true) == 0:
+		return float('nan')
+
+	return len(truePositive) / len(true)
+
+def precision(detected, true):
+	truePositive = [entity for entity in detected if entity in true]
+
+	if len(detected) == 0:
+		return float('nan')
+
+	return len(truePositive) / len(detected)
+
+def f_measure(detected, true, alpha=0.5):
+	pre = precision(detected, true)
+	rec = recall(detected, true)
+
+	if ((pre == 0) & (rec == 0)):
+		return 0.0
+
+	return pre*rec/(alpha*rec + (1-alpha)*pre)

experiments/feature_envy/compare_approaches.py

@@ -0,0 +1,57 @@
+from context import dataUtils, evaluate, hist, incode, jdeodorant
+
+
+# This script is used to compare the performances of various feature envy detection approaches
+# on the systems considered in this study.
+
+parameters = {'hist': 2.0, 'incode': (3.0, 3.0, 3.0)}
+
+systems = [
+	'android-frameworks-opt-telephony',
+	'android-platform-support',
+	'apache-ant', 
+	'apache-tomcat',
+	'lucene',
+	'argouml',
+	'jedit',
+	'xerces-2_7_0'
+]
+
+
+for system in systems:
+	# Get occurrences manually detected on the considered system
+	true = dataUtils.getLabels(system, 'feature_envy')
+
+	# Compute performances for HIST
+	detected_hist = hist.getSmells(system, parameters['hist'])
+
+	precision_hist = evaluate.precision(detected_hist, true)
+	recall_hist = evaluate.recall(detected_hist, true)
+	f_measure_hist = evaluate.f_measure(detected_hist, true)
+
+	# Compute performances for InCode
+	detected_incode = incode.getSmells(system, *parameters['incode'])
+
+	precision_incode = evaluate.precision(detected_incode, true)
+	recall_incode = evaluate.recall(detected_incode, true)
+	f_measure_incode = evaluate.f_measure(detected_incode, true)
+
+	# Compute performances for JDeodorant
+	detected_jdeodorant = jdeodorant.getSmells(system)
+
+	precision_jdeodorant = evaluate.precision(detected_jdeodorant, true)
+	recall_jdeodorant = evaluate.recall(detected_jdeodorant, true)
+	f_measure_jdeodorant = evaluate.f_measure(detected_jdeodorant, true)
+
+
+	# Output results
+	print('           |precision  |recall  |f_measure')
+	print('-------------------------------------------------')
+	print('HIST       |' + "{0:.3f}".format(precision_hist) + '      |' + "{0:.3f}".format(recall_hist) + '   |' + "{0:.3f}".format(f_measure_hist))
+	print('-------------------------------------------------')
+	print('InCode     |' + "{0:.3f}".format(precision_incode) + '      |' + "{0:.3f}".format(recall_incode) + '   |' + "{0:.3f}".format(f_measure_incode))
+	print('-------------------------------------------------')
+	print('JDeodorant |' + "{0:.3f}".format(precision_jdeodorant) + '      |' + "{0:.3f}".format(recall_jdeodorant) + '   |' + "{0:.3f}".format(f_measure_jdeodorant))
+	print('\n\n')
+
+

experiments/feature_envy/context.py

@@ -0,0 +1,12 @@
+import os
+import sys
+
+ROOT_DIR = os.path.abspath(os.path.join(os.path.dirname(__file__), '../..'))
+sys.path.insert(0, ROOT_DIR)
+
+import utils.dataUtils as dataUtils
+import experiments.evaluate as evaluate
+
+import detection_tools.replication.feature_envy.hist       as hist
+import detection_tools.replication.feature_envy.incode     as incode
+import detection_tools.replication.feature_envy.jdeodorant as jdeodorant

tests/context.py

@@ -6,6 +6,8 @@
 
 import utils.dataUtils as dataUtils
 
+import experiments.evaluate as evaluate
+
 import detection_tools.confidence_metrics as cm
 
 import detection_tools.replication.feature_envy.hist       as hist_fe

tests/test_evaluate.py

@@ -0,0 +1,65 @@
+from context import evaluate
+
+import math
+import unittest
+
+class TestDecorGCCM(unittest.TestCase):
+	def setUp(self):
+		self.set_1 = ['a','b','c','d', 'e']
+		self.set_2 = ['b','e', 'f', 'g']
+		self.set_3 = ['c']
+		self.set_4 = []
+
+	def tearDown(self):
+		del self.set_1
+		del self.set_2
+		del self.set_3
+		del self.set_4
+
+
+	def test_precision(self):
+		self.assertEqual(evaluate.precision(self.set_1, self.set_1), 1.0)
+		self.assertEqual(evaluate.precision(self.set_1, self.set_2), 0.4)
+		self.assertEqual(evaluate.precision(self.set_1, self.set_3), 0.2)
+		self.assertEqual(evaluate.precision(self.set_1, self.set_4), 0.0)
+		self.assertEqual(evaluate.precision(self.set_2, self.set_1), 0.5)
+		self.assertEqual(evaluate.precision(self.set_2, self.set_2), 1.0)
+		self.assertEqual(evaluate.precision(self.set_2, self.set_3), 0.0)
+		self.assertEqual(evaluate.precision(self.set_3, self.set_1), 1.0)
+		self.assertEqual(evaluate.precision(self.set_3, self.set_4), 0.0)
+		self.assertTrue(math.isnan(evaluate.precision(self.set_4, self.set_1)))
+		self.assertTrue(math.isnan(evaluate.precision(self.set_4, self.set_3)))
+		self.assertTrue(math.isnan(evaluate.precision(self.set_4, self.set_4)))
+
+	def test_recall(self):
+		self.assertEqual(evaluate.recall(self.set_1, self.set_1), 1.0)
+		self.assertEqual(evaluate.recall(self.set_1, self.set_2), 0.5)
+		self.assertEqual(evaluate.recall(self.set_1, self.set_3), 1.0)
+		self.assertTrue(math.isnan(evaluate.recall(self.set_1, self.set_4)))
+		self.assertEqual(evaluate.recall(self.set_2, self.set_1), 0.4)
+		self.assertEqual(evaluate.recall(self.set_2, self.set_2), 1.0)
+		self.assertEqual(evaluate.recall(self.set_2, self.set_3), 0.0)
+		self.assertEqual(evaluate.recall(self.set_3, self.set_1), 0.2)
+		self.assertTrue(math.isnan(evaluate.recall(self.set_3, self.set_4)))
+		self.assertEqual(evaluate.recall(self.set_4, self.set_1), 0.0)
+		self.assertEqual(evaluate.recall(self.set_4, self.set_3), 0.0)
+		self.assertTrue(math.isnan(evaluate.recall(self.set_4, self.set_4)))
+
+	def test_f_measure(self):
+		self.assertEqual(evaluate.f_measure(self.set_1, self.set_1), 1.0)
+		self.assertEqual(evaluate.f_measure(self.set_1, self.set_2), 0.4444444444444445)
+		self.assertEqual(evaluate.f_measure(self.set_1, self.set_3), 0.33333333333333337)
+		self.assertTrue(math.isnan(evaluate.f_measure(self.set_1, self.set_4)))
+		self.assertEqual(evaluate.f_measure(self.set_2, self.set_1), 0.4444444444444445)
+		self.assertEqual(evaluate.f_measure(self.set_2, self.set_2), 1.0)
+		self.assertEqual(evaluate.f_measure(self.set_2, self.set_3), 0.0)
+		self.assertEqual(evaluate.f_measure(self.set_3, self.set_1), 0.33333333333333337)
+		self.assertTrue(math.isnan(evaluate.f_measure(self.set_3, self.set_4)))
+		self.assertTrue(math.isnan(evaluate.f_measure(self.set_4, self.set_1)))
+		self.assertTrue(math.isnan(evaluate.f_measure(self.set_4, self.set_3)))
+		self.assertTrue(math.isnan(evaluate.f_measure(self.set_4, self.set_4)))
+
+
+
+if __name__ == '__main__':
+	unittest.main()

utils/dataUtils.py

@@ -87,13 +87,22 @@ def getMethods(systemName):
 
     methods = []
     with open(systemMethodsFile, 'rb') as csvfile:
-        rdr = csv.reader(csvfile, delimiter=';')
+        reader = csv.reader(csvfile, delimiter=';')
 
-        for row in rdr:
+        for row in reader:
             methods.append(row[0])
 
     return methods
 
+# Get the hand-validated occurences reported in the considered system for antipattern in [god_class, feature_envy].
+def getLabels(systemName, antipattern):
+    if antipattern not in ['god_class', 'feature_envy']:
+        print(str(antipattern) + ' not valid antipattern name. Choose "god_class" or "feature_envy instead"')
+        return
+
+    labelFile = os.path.join(ROOT_DIR, 'data/labels/' + antipattern + '/' + systemName + '.txt')
+    with open(labelFile, 'r') as file:
+        return file.read().splitlines()
 
 
 #####     MERGED DETECTION INSTANCES GETTERS     #####