-
Notifications
You must be signed in to change notification settings - Fork 3
/
preprocess.py
159 lines (129 loc) · 5.14 KB
/
preprocess.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
import sys
import os
import gzip
import cPickle as pkl
import pandas
import numpy
from subprocess import Popen, PIPE
chosen_frequency = 10
tokenizer_cmd = ['/usr/bin/perl', 'tokenizer.perl', '-l', 'en', '-q', '-']
def arg_passing(argv):
i = 1
arg_dict = {}
while i < len(argv) - 1:
arg_dict[argv[i]] = argv[i+1]
i += 2
return arg_dict
def normalize(seqs):
for i, s in enumerate(seqs):
words = s.split()
if len(words) < 1:
seqs[i] = 'null'
return seqs
def tokenize(sentences):
print 'Tokenizing..',
text = "\n".join(sentences)
tokenizer = Popen(tokenizer_cmd, stdin=PIPE, stdout=PIPE) # pass string to perl function for tokenizing
tok_text, _ = tokenizer.communicate(text)
toks = tok_text.split('\n')[:-1]
print 'Done'
return toks
def grab_data(context, codesnippet, dictionary):
context = tokenize(context)
codesnippet = tokenize(codesnippet)
seqs = [[None] * len(context), [None] * len(codesnippet)]
for i, sentences in enumerate([context, codesnippet]):
for idx, ss in enumerate(sentences):
words = ss.strip().lower().split()
seqs[i][idx] = [dictionary[w] if w in dictionary else 0 for w in words]
if len(seqs[i][idx]) == 0:
print 'len 0: ', i, idx
return seqs[0], seqs[1]
def build_dict(sentences):
sentences = tokenize(sentences)
print 'Building dictionary..'
wordcount = dict()
for ss in sentences:
words = ss.strip().lower().split()
for w in words:
if w not in wordcount:
wordcount[w] = 1
else:
wordcount[w] += 1
counts = wordcount.values()
keys = wordcount.keys()
sorted_idx = numpy.argsort(counts)[::-1]
counts = numpy.array(counts)
print 'number of words in dictionary:', len(keys)
worddict = dict()
for idx, ss in enumerate(sorted_idx):
worddict[keys[ss]] = idx+1 # leave 0 (UNK)
pos = 0
for i, c in enumerate(sorted_idx):
if counts[c] >= chosen_frequency: pos = i
print numpy.sum(counts), ' total words, ', pos, 'words with frequency >=', chosen_frequency
return worddict
def clean_sen(sen):
sen = ''.join([c if ord(c) < 128 and ord(c) > 32 else ' ' for c in sen])
return sen
def main():
# read data from csv
args = arg_passing(sys.argv)
path = args['-path']
project = args['-project']
data_path = path + project + '.csv'
data = pandas.read_csv(data_path)
# the following line drops created column which is only present in Tawosi dataset. This column contains the creation
# date-time of each bug, which is used to sort the issues but is not needed here.
data = data.drop(['created'], axis=1, errors='ignore')
data = data.values
labels = data[:, 1].astype('float32')
context = normalize(data[:, 2].astype('str'))
codesnippet = normalize(data[:, 3].astype('str'))
binaryFeat = data[:, 4:].astype('float32')
for i in range(len(context)):
if context[i] is None:
context[i] = 'None'
else:
context[i] = clean_sen(context[i])
for i in range(len(codesnippet)):
if codesnippet[i] is None:
codesnippet[i] = 'None'
else:
codesnippet[i] = clean_sen(codesnippet[i])
if not os.path.isfile('files/' + project + '_3sets.txt'):
raise Exception('expected file not found! file= files/' + project + '_3sets.txt')
f = open('files/' + project + '_3sets.txt', 'r')
train_ids, valid_ids, test_ids = [], [], []
count = -2
for line in f:
if count == -2:
count += 1
continue
count += 1
ls = line.split()
if ls[0] == '1':
train_ids.append(count)
if ls[1] == '1':
valid_ids.append(count)
if ls[2] == '1':
test_ids.append(count)
print 'ntrain, nvalid, ntest: ', len(train_ids), len(valid_ids), len(test_ids)
#preprocess data and packing
train_context, train_codesnippet, train_binaryFeat, train_labels = context[train_ids], codesnippet[train_ids], binaryFeat[train_ids], labels[train_ids]
valid_context, valid_codesnippet, valid_binaryFeat, valid_labels = context[valid_ids], codesnippet[valid_ids], binaryFeat[valid_ids], labels[valid_ids]
test_context, test_codesnippet, test_binaryFeat, test_labels = context[test_ids], codesnippet[test_ids], binaryFeat[test_ids], labels[test_ids]
dictionary = build_dict(numpy.concatenate([train_context, train_codesnippet]))
f = gzip.open('files/' + project + '.dict.pkl.gz', 'wb')
pkl.dump(dictionary, f, -1)
f.close()
train_t, train_d = grab_data(train_context, train_codesnippet, dictionary)
valid_t, valid_d = grab_data(valid_context, valid_codesnippet, dictionary)
test_t, test_d = grab_data(test_context, test_codesnippet, dictionary)
f = gzip.open('files/' + project + '.pkl.gz', 'wb')
pkl.dump((train_t, train_d,train_binaryFeat, train_labels,
valid_t, valid_d,valid_binaryFeat, valid_labels,
test_t, test_d,test_binaryFeat, test_labels), f, -1)
f.close()
if __name__ == '__main__':
main()