def loading_data_all(FLAGS):
# load all data from FLAGS path
# split data to training and testing, only load testing data
commits_ = extract_commit(path_file=FLAGS.path)
filter_commits = filtering_commit(commits=commits_,
num_file=FLAGS.code_file,
num_hunk=FLAGS.code_hunk,
num_loc=FLAGS.code_line,
size_line=FLAGS.code_length)
msgs_, codes_ = extract_msg(commits=filter_commits), extract_code(
commits=filter_commits)
dict_msg_, dict_code_ = dictionary(data=msgs_), dictionary(data=codes_)
pad_msg = mapping_commit_msg(msgs=msgs_,
max_length=FLAGS.msg_length,
dict_msg=dict_msg_)
pad_added_code = mapping_commit_code(type="added",
commits=filter_commits,
max_hunk=FLAGS.code_hunk,
max_code_line=FLAGS.code_line,
max_code_length=FLAGS.code_length,
dict_code=dict_code_)
pad_removed_code = mapping_commit_code(type="removed",
commits=filter_commits,
max_hunk=FLAGS.code_hunk,
max_code_line=FLAGS.code_line,
max_code_length=FLAGS.code_length,
dict_code=dict_code_)
labels = load_label_commits(commits=filter_commits)
return pad_msg, pad_added_code, pad_removed_code, labels
def load_data_type(path, FLAGS):
commits_ = extract_commit_july(path_file=path)
msgs_, codes_ = extract_msg(commits=commits_), extract_code(
commits=commits_)
dict_msg_, dict_code_ = dictionary(data=msgs_), dictionary(data=codes_)
print len(commits_), len(dict_msg_), len(dict_code_)
pad_msg = mapping_commit_msg(msgs=msgs_,
max_length=FLAGS.msg_length,
dict_msg=dict_msg_)
pad_added_code = mapping_commit_code(type="added",
commits=commits_,
max_hunk=FLAGS.code_hunk,
max_code_line=FLAGS.code_line,
max_code_length=FLAGS.code_length,
dict_code=dict_code_)
pad_removed_code = mapping_commit_code(type="removed",
commits=commits_,
max_hunk=FLAGS.code_hunk,
max_code_line=FLAGS.code_line,
max_code_length=FLAGS.code_length,
dict_code=dict_code_)
labels = load_label_commits(commits=commits_)
return pad_msg, pad_added_code, pad_removed_code, labels, dict_msg_, dict_code_
def loading_data_lstm(FLAGS):
print FLAGS.model
if "msg" in FLAGS.model:
commits_ = extract_commit(path_file=FLAGS.path)
filter_commits = filtering_commit(commits=commits_, num_file=FLAGS.code_file, num_hunk=FLAGS.code_hunk,
num_loc=FLAGS.code_line,
size_line=FLAGS.code_length)
msgs_, codes_ = extract_msg(commits=filter_commits), extract_code(commits=filter_commits)
dict_msg_, dict_code_ = dictionary(data=msgs_), dictionary(data=codes_)
pad_msg = mapping_commit_msg(msgs=msgs_, max_length=FLAGS.msg_length, dict_msg=dict_msg_)
pad_added_code = mapping_commit_code(type="added", commits=filter_commits, max_hunk=FLAGS.code_hunk,
max_code_line=FLAGS.code_line,
max_code_length=FLAGS.code_length, dict_code=dict_code_)
pad_removed_code = mapping_commit_code(type="removed", commits=filter_commits, max_hunk=FLAGS.code_hunk,
max_code_line=FLAGS.code_line,
max_code_length=FLAGS.code_length, dict_code=dict_code_)
labels = load_label_commits(commits=filter_commits)
elif "all" in FLAGS.model:
commits_ = extract_commit(path_file=FLAGS.path)
filter_commits = filtering_commit(commits=commits_, num_file=FLAGS.code_file, num_hunk=FLAGS.code_hunk,
num_loc=FLAGS.code_line,
size_line=FLAGS.code_length)
msgs_, codes_ = extract_msg(commits=filter_commits), extract_code(commits=filter_commits)
all_lines = add_two_list(list1=msgs_, list2=codes_)
msgs_ = all_lines
dict_msg_, dict_code_ = dictionary(data=msgs_), dictionary(data=codes_)
pad_msg = mapping_commit_msg(msgs=msgs_, max_length=FLAGS.msg_length, dict_msg=dict_msg_)
pad_added_code = mapping_commit_code(type="added", commits=filter_commits, max_hunk=FLAGS.code_hunk,
max_code_line=FLAGS.code_line,
max_code_length=FLAGS.code_length, dict_code=dict_code_)
pad_removed_code = mapping_commit_code(type="removed", commits=filter_commits, max_hunk=FLAGS.code_hunk,
max_code_line=FLAGS.code_line,
max_code_length=FLAGS.code_length, dict_code=dict_code_)
labels = load_label_commits(commits=filter_commits)
elif "code" in FLAGS.model:
commits_ = extract_commit(path_file=FLAGS.path)
filter_commits = filtering_commit(commits=commits_, num_file=FLAGS.code_file, num_hunk=FLAGS.code_hunk,
num_loc=FLAGS.code_line,
size_line=FLAGS.code_length)
msgs_, codes_ = extract_msg(commits=filter_commits), extract_code(commits=filter_commits)
msgs_ = codes_
dict_msg_, dict_code_ = dictionary(data=msgs_), dictionary(data=codes_)
pad_msg = mapping_commit_msg(msgs=msgs_, max_length=FLAGS.msg_length, dict_msg=dict_msg_)
pad_added_code = mapping_commit_code(type="added", commits=filter_commits, max_hunk=FLAGS.code_hunk,
max_code_line=FLAGS.code_line,
max_code_length=FLAGS.code_length, dict_code=dict_code_)
pad_removed_code = mapping_commit_code(type="removed", commits=filter_commits, max_hunk=FLAGS.code_hunk,
max_code_line=FLAGS.code_line,
max_code_length=FLAGS.code_length, dict_code=dict_code_)
labels = load_label_commits(commits=filter_commits)
else:
print "You need to type correct model"
exit()
kf = KFold(n_splits=FLAGS.folds, random_state=FLAGS.seed)
for train_index, test_index in kf.split(filter_commits):
X_train_msg, X_test_msg = np.array(get_items(items=pad_msg, indexes=train_index)), \
np.array(get_items(items=pad_msg, indexes=test_index))
X_train_added_code, X_test_added_code = np.array(get_items(items=pad_added_code, indexes=train_index)), \
np.array(get_items(items=pad_added_code, indexes=test_index))
X_train_removed_code, X_test_removed_code = np.array(get_items(items=pad_removed_code, indexes=train_index)), \
np.array(get_items(items=pad_removed_code, indexes=test_index))
y_train, y_test = np.array(get_items(items=labels, indexes=train_index)), \
np.array(get_items(items=labels, indexes=test_index))
return X_test_msg, X_test_added_code, X_test_removed_code, y_test
if __name__ == "__main__":
tf = model_parameters()
FLAGS = tf.flags.FLAGS
print_params(tf)
type = "msg"
msgs_train, codes_train, commit_train = loading_training_data(FLAGS=FLAGS,
type=type)
path_testing_data = "./data/test_data/merging_markus_sasha.txt"
msgs_test, codes_test, commit_test = loading_testing_data(
path_file=path_testing_data, FLAGS=FLAGS, type=type)
msgs_train_code = msgs_train + codes_train
dict_train = dictionary(data=msgs_train_code)
pad_msg_train = mapping_commit_msg(msgs=msgs_train,
max_length=FLAGS.msg_length,
dict_msg=dict_train)
pad_msg_test = mapping_commit_msg(msgs=msgs_test,
max_length=FLAGS.msg_length,
dict_msg=dict_train)
labels_train, labels_test = load_label_commits(
commits=commit_train), load_label_commits(commits=commit_test)
labels_train, labels_test = convert_to_binary(
labels_train), convert_to_binary(labels_test)
Y_train, Y_test = labels_train, labels_test
# name = "lstm_cnn_msg"
# name = "lstm_cnn_code"
# name = "lstm_cnn_all"
# name = "cnn_msg"
# name = "cnn_code"
def loading_baseline_july(tf, folds, random_state):
FLAGS = tf.flags.FLAGS
commits_ = extract_commit_july(path_file=FLAGS.path)
filter_commits = commits_
print len(commits_)
kf = KFold(n_splits=folds, random_state=random_state)
idx_folds = list()
for train_index, test_index in kf.split(filter_commits):
idx = dict()
idx["train"], idx["test"] = train_index, test_index
idx_folds.append(idx)
if "msg" in FLAGS.model:
msgs_, codes_ = extract_msg(commits=filter_commits), extract_code(
commits=filter_commits)
elif "all" in FLAGS.model:
msgs_, codes_ = extract_msg(commits=filter_commits), extract_code(
commits=filter_commits)
all_lines = add_two_list(list1=msgs_, list2=codes_)
msgs_ = all_lines
elif "code" in FLAGS.model:
msgs_, codes_ = extract_msg(commits=filter_commits), extract_code(
commits=filter_commits)
msgs_ = codes_
else:
print "You need to type correct model"
exit()
dict_msg_, dict_code_ = dictionary(data=msgs_), dictionary(data=codes_)
pad_msg = mapping_commit_msg(msgs=msgs_,
max_length=FLAGS.msg_length,
dict_msg=dict_msg_)
labels = load_label_commits(commits=filter_commits)
labels = convert_to_binary(labels)
# path_file = "./statistical_test_prob/true_label.txt"
# write_file(path_file=path_file, data=labels)
# exit()
print pad_msg.shape, labels.shape, len(dict_msg_)
cntfold = 0
pred_dict = dict()
pred_dict_list = list()
for i in xrange(cntfold, len(idx_folds)):
idx = idx_folds[i]
train_index, test_index = idx["train"], idx["test"]
X_train_msg, X_test_msg = np.array(get_items(items=pad_msg, indexes=train_index)), \
np.array(get_items(items=pad_msg, indexes=test_index))
Y_train, Y_test = np.array(get_items(items=labels, indexes=train_index)), \
np.array(get_items(items=labels, indexes=test_index))
if FLAGS.model == "lstm_cnn_all" or FLAGS.model == "lstm_cnn_msg" \
or FLAGS.model == "lstm_cnn_code" or FLAGS.model == "cnn_all" \
or FLAGS.model == "cnn_msg" or FLAGS.model == "cnn_code":
# path_model = "./keras_model/%s_%s.h5" % (FLAGS.model, str(cntfold))
path_model = "./keras_model/test_%s_%s.h5" % (FLAGS.model,
str(cntfold))
# path_model = "./keras_model/%s_%s_testing.h5" % (FLAGS.model, str(cntfold))
model = load_model(path_model)
else:
print "You need to give correct model name"
exit()
y_pred = model.predict(X_test_msg, batch_size=FLAGS.batch_size)
y_pred = np.ravel(y_pred)
pred_dict.update(make_dictionary(y_pred=y_pred, y_index=test_index))
y_pred = y_pred.tolist()
pred_dict_list += y_pred
# print len(pred_dict_list)
# exit()
# path_file = "./statistical_test_prob/" + FLAGS.model + ".txt"
# write_file(path_file=path_file, data=sorted_dict(dict=pred_dict))
path_file = "./statistical_test_prob/" + FLAGS.model + "_checking.txt"
write_file(path_file=path_file, data=pred_dict_list)
path_test = "./data/test_data/markus_translated.out"
# path_test = "./data/test_data/sasha_translated.out"
commits_test = extract_commit(path_file=path_test)
filter_commits_test = filtering_commit(commits=commits_test,
num_file=code_file,
num_hunk=code_hunk,
num_loc=code_line,
size_line=code_length)
msgs_test, codes_test = extract_msg(
commits=filter_commits_test), extract_code(commits=filter_commits_test)
all_lines_test = add_two_list(list1=msgs_test, list2=codes_test)
msgs_test = msgs_test
dict_msg_ = dict_msg_train.update(dict_code_train)
pad_msg_test = mapping_commit_msg(msgs=msgs_test,
max_length=msg_length,
dict_msg=dict_msg_)
labels = load_label_commits(commits=filter_commits_test)
labels = convert_to_binary(labels)
model_name = "cnn_all"
model_name = "lstm_all"
model_name = "bi_lstm_all"
model_name = "lstm_cnn_all"
print path_test, model_name
model_path = "./lstm_model_ver2/" + model_name + "_0.h5"
model = load_model(model_path)
y_pred = model.predict(pad_msg_test, batch_size=32)
y_pred = np.ravel(y_pred)
y_pred[y_pred > 0.5] = 1
y_pred[y_pred <= 0.5] = 0
commits_ = extract_commit(path_file=FLAGS.path)
filter_commits = filtering_commit(commits=commits_,
num_file=FLAGS.code_file,
num_hunk=FLAGS.code_hunk,
num_loc=FLAGS.code_line,
size_line=FLAGS.code_length)
msgs_, codes_ = extract_msg(commits=filter_commits), extract_code(
commits=filter_commits)
msgs_ = codes_
else:
print "You need to type correct model"
exit()
dict_msg_, dict_code_ = dictionary(data=msgs_), dictionary(data=codes_)
pad_msg = mapping_commit_msg(msgs=msgs_,
max_length=FLAGS.msg_length,
dict_msg=dict_msg_)
labels = load_label_commits(commits=filter_commits)
labels = convert_to_binary(labels)
print pad_msg.shape, labels.shape, labels.shape, len(dict_msg_)
y_pred = model.predict(pad_msg, batch_size=FLAGS.batch_size)
y_pred = np.ravel(y_pred)
y_pred[y_pred > 0.5] = 1
y_pred[y_pred <= 0.5] = 0
accuracy = accuracy_score(y_true=labels, y_pred=y_pred)
precision = precision_score(y_true=labels, y_pred=y_pred)
recall = recall_score(y_true=labels, y_pred=y_pred)
f1 = f1_score(y_true=labels, y_pred=y_pred)
auc = auc_score(y_true=labels, y_pred=y_pred)
def running_baseline_july(tf, folds, random_state):
FLAGS = tf.flags.FLAGS
commits_ = extract_commit_july(path_file=FLAGS.path)
filter_commits = commits_
print len(commits_)
kf = KFold(n_splits=folds, random_state=random_state)
idx_folds = list()
for train_index, test_index in kf.split(filter_commits):
idx = dict()
idx["train"], idx["test"] = train_index, test_index
idx_folds.append(idx)
if "msg" in FLAGS.model:
msgs_, codes_ = extract_msg(commits=filter_commits), extract_code(commits=filter_commits)
elif "all" in FLAGS.model:
msgs_, codes_ = extract_msg(commits=filter_commits), extract_code(commits=filter_commits)
all_lines = add_two_list(list1=msgs_, list2=codes_)
msgs_ = all_lines
elif "code" in FLAGS.model:
msgs_, codes_ = extract_msg(commits=filter_commits), extract_code(commits=filter_commits)
msgs_ = codes_
else:
print "You need to type correct model"
exit()
dict_msg_, dict_code_ = dictionary(data=msgs_), dictionary(data=codes_)
pad_msg = mapping_commit_msg(msgs=msgs_, max_length=FLAGS.msg_length, dict_msg=dict_msg_)
labels = load_label_commits(commits=filter_commits)
labels = convert_to_binary(labels)
print pad_msg.shape, labels.shape, len(dict_msg_)
# exit()
timestamp = str(int(time.time()))
accuracy, precision, recall, f1, auc = list(), list(), list(), list(), list()
cntfold = 0
pred_dict, pred_dict_prob = dict(), dict()
for i in xrange(cntfold, len(idx_folds)):
idx = idx_folds[i]
train_index, test_index = idx["train"], idx["test"]
X_train_msg, X_test_msg = np.array(get_items(items=pad_msg, indexes=train_index)), \
np.array(get_items(items=pad_msg, indexes=test_index))
Y_train, Y_test = np.array(get_items(items=labels, indexes=train_index)), \
np.array(get_items(items=labels, indexes=test_index))
if FLAGS.model == "lstm_cnn_msg" or FLAGS.model == "lstm_cnn_code" or FLAGS.model == "lstm_cnn_all":
model = lstm_cnn(x_train=X_train_msg, y_train=Y_train, x_test=X_test_msg,
y_test=Y_test, dictionary_size=len(dict_msg_), FLAGS=FLAGS)
elif FLAGS.model == "cnn_msg" or FLAGS.model == "cnn_code" or FLAGS.model == "cnn_all":
model = cnn_model(x_train=X_train_msg, y_train=Y_train, x_test=X_test_msg,
y_test=Y_test, dictionary_size=len(dict_msg_), FLAGS=FLAGS)
else:
print "You need to give correct model name"
exit()
# model.save("./keras_model/" + FLAGS.model + "_" + str(cntfold) + ".h5")
# model.save("./keras_model/" + FLAGS.model + "_" + str(cntfold) + "_testing.h5")
# model.save("./keras_model/test_" + FLAGS.model + "_" + str(cntfold) + ".h5")
model.save("./keras_model/newres_funcalls_" + FLAGS.model + "_" + str(cntfold) + ".h5")
y_pred = model.predict(X_test_msg, batch_size=FLAGS.batch_size)
y_pred = np.ravel(y_pred)
y_pred_tolist = y_pred.tolist()
data_fold = [str(i) + "\t" + str(l) for i, l in zip(test_index, y_pred)]
path_file = "./statistical_test/newres_funcalls_%s_fold_%s.txt" % (FLAGS.model, str(cntfold))
write_file(path_file=path_file, data=data_fold)
y_pred[y_pred > 0.5] = 1
y_pred[y_pred <= 0.5] = 0
pred_dict.update(make_dictionary(y_pred=y_pred, y_index=test_index))
accuracy.append(accuracy_score(y_true=Y_test, y_pred=y_pred))
precision.append(precision_score(y_true=Y_test, y_pred=y_pred))
recall.append(recall_score(y_true=Y_test, y_pred=y_pred))
f1.append(f1_score(y_true=Y_test, y_pred=y_pred))
auc.append(auc_score(y_true=Y_test, y_pred=y_pred))
print "accuracy", accuracy_score(y_true=Y_test, y_pred=y_pred)
print "precision", precision_score(y_true=Y_test, y_pred=y_pred)
print "recall", recall_score(y_true=Y_test, y_pred=y_pred)
print "f1", f1_score(y_true=Y_test, y_pred=y_pred)
cntfold += 1
break