def padding_train_test_commits(train, test, params):
# training data
# -------------------------------------------------------------------------------------------------------
# -------------------------------------------------------------------------------------------------------
train_msgs, train_codes = extract_msg(commits=train), extract_code(
commits=test)
train_dict_msg, train_dict_code = dictionary(data=train_msgs), dictionary(
data=train_codes)
train_pad_msg = mapping_commit_msg(msgs=train_msgs,
max_length=params.msg_length,
dict_msg=train_dict_msg)
train_pad_added_code = mapping_commit_code(
type="added",
commits=train,
max_hunk=params.code_hunk,
max_code_line=params.code_line,
max_code_length=params.code_length,
dict_code=train_dict_code)
train_pad_removed_code = mapping_commit_code(
type="removed",
commits=train,
max_hunk=params.code_hunk,
max_code_line=params.code_line,
max_code_length=params.code_length,
dict_code=train_dict_code)
train_labels = load_label_commits(commits=train)
train_data = (train_labels, train_pad_msg, train_pad_added_code,
train_pad_removed_code)
dict_commit = (train_dict_msg, train_dict_code)
# testing data
# -------------------------------------------------------------------------------------------------------
# -------------------------------------------------------------------------------------------------------
test_msgs, test_codes = extract_msg(commits=test), extract_code(
commits=test)
test_pad_msg = mapping_commit_msg(msgs=test_msgs,
max_length=params.msg_length,
dict_msg=train_dict_msg)
test_pad_added_code = mapping_commit_code(
type="added",
commits=test,
max_hunk=params.code_hunk,
max_code_line=params.code_line,
max_code_length=params.code_length,
dict_code=train_dict_code)
test_pad_removed_code = mapping_commit_code(
type="removed",
commits=test,
max_hunk=params.code_hunk,
max_code_line=params.code_line,
max_code_length=params.code_length,
dict_code=train_dict_code)
test_labels = load_label_commits(commits=test)
test_data = (test_labels, test_pad_msg, test_pad_added_code,
test_pad_removed_code)
return train_data, test_data, dict_commit
def padding_testing_commit(commits, dictionary, params):
msgs = extract_msg(commits=commits)
codes = extract_code(commits=commits)
ids = extract_id(commits=commits) #patch id
dict_msg, dict_code = dictionary
# padding commit message
pad_msg = mapping_commit_msg(msgs=msgs,
max_length=params.msg_length,
dict_msg=dict_msg)
# padding commit code
pad_added_code = mapping_commit_code(type_="added",
commits=commits,
max_file=params.code_file,
max_hunk=params.code_hunk,
max_code_line=params.code_line,
max_code_length=params.code_length,
dict_code=dict_code)
pad_removed_code = mapping_commit_code(type_="removed",
commits=commits,
max_file=params.code_file,
max_hunk=params.code_hunk,
max_code_line=params.code_line,
max_code_length=params.code_length,
dict_code=dict_code)
labels = load_label_commits(commits=commits)
return pad_msg, pad_added_code, pad_removed_code, labels, ids
def padding_commit_topwords(commits, params):
codes = extract_code(commits=commits)
dict_code = dictionary(data=codes)
# padding commit code
pad_added_code = mapping_commit_code(type="added",
commits=commits,
max_hunk=params.code_hunk,
max_code_line=params.code_line,
max_code_length=params.code_length,
dict_code=dict_code)
pad_removed_code = mapping_commit_code(type="removed",
commits=commits,
max_hunk=params.code_hunk,
max_code_line=params.code_line,
max_code_length=params.code_length,
dict_code=dict_code)
return pad_added_code, pad_removed_code, dict_code
def padding_pred_commit(commits, params, dict_msg, dict_code):
msgs, codes = extract_msg(commits=commits), extract_code(commits=commits)
# padding commit message
pad_msg = mapping_commit_msg(msgs=msgs,
max_length=params.msg_length,
dict_msg=dict_msg)
# padding commit code
pad_added_code = mapping_commit_code(type="added",
commits=commits,
max_hunk=params.code_hunk,
max_code_line=params.code_line,
max_code_length=params.code_length,
dict_code=dict_code)
pad_removed_code = mapping_commit_code(type="removed",
commits=commits,
max_hunk=params.code_hunk,
max_code_line=params.code_line,
max_code_length=params.code_length,
dict_code=dict_code)
labels = load_label_commits(commits=commits)
return pad_msg, pad_added_code, pad_removed_code, labels
#path_input_data1 = "test_data.out"
#path_input_data2 = "training_data.out"
#path_output_data = "./try_data/dict_try.pkl"
#show_dict = True
params = read_args().parse_args()
path_input_data1 = params.text_path1
path_input_data2 = params.text_path2
path_output_data = params.dict_path
show_dict = params.print
test_data = extract_commit(path_file=path_input_data1)
train_data = extract_commit(path_file=path_input_data2)
whole_data = train_data + test_data # add train data and test data together
#whole_data = test_data
msgs, codes = extract_msg(whole_data), extract_code(whole_data)
dict_msg, dict_code = dictionary(data=msgs), dictionary(data=codes)
#print(len(msgs))
#print (msgs[1])
#print(len(codes))
#print(codes[1])
print("the number of different tokens in message part is : {n}".format(n=len(dict_msg)))
print("the number of different tokens in code part is : {n}".format(n=len(dict_code)))
dict_whole = (dict_msg, dict_code)
with open(path_output_data, 'wb') as handle:
pickle.dump(dict_whole, handle, protocol=pickle.HIGHEST_PROTOCOL)
if show_dict == True:
