import json import numpy as np import matplotlib.pyplot as plt import pandas as pd import seaborn as sns from matplotlib.dates import MONTHLY, WEEKLY, DateFormatter, rrulewrapper, RRuleLocator from pandas.plotting import register_matplotlib_converters import re from mongoengine import connect, DoesNotExist from pycoshark.mongomodels import Commit, FileAction, File, Project, VCSSystem, Hunk, Issue, IssueSystem, Refactoring from pycoshark.utils import create_mongodb_uri_string, java_filename_filter from bson import json_util from scipy import stats from scipy import optimize from autorank._util import _confidence_interval from itertools import chain with open('labels.json', 'r') as file: completed = json_util.loads(file.read()) confidence_level = 1-(0.05/45) for result in completed: labels = result['lines_manual'] first_labels = next(iter(labels.values())) lines = [] for cur_lines in first_labels.values(): lines.extend(cur_lines) label_dict = {} label_dict_consensus = {} for line in lines: label_dict[line] = {'bugfix':0, 'whitespace':0, 'documentation':0, 'refactoring':0, 'test':0, 'unrelated':0} for user, user_labels in labels.items(): for label_type, label_lines in user_labels.items(): if label_type=='bug': label_type='bugfix' print(result) for label_line in label_lines: label_dict[label_line][label_type] += 1 for line, line_labels in label_dict.items(): has_consensus = False for label_type, count in line_labels.items(): if count>=3: # we need this part due to our training issues if has_consensus: if count>consensus_count: consensus_count = count label_dict_consensus[line] = label_type label_dict_consensus[line] = label_type has_consensus = True consensus_count = count if not has_consensus: is_java_file = java_filename_filter(result['file'], production_only=False) is_prod_file = java_filename_filter(result['file'], production_only=True) is_test_file = is_java_file and not is_prod_file if is_test_file and line_labels['test']+line_labels['whitespace']+line_labels['documentation']>=4: label_dict_consensus[line] = 'test_doc_whitespace' elif line_labels['bugfix']==0: label_dict_consensus[line] = 'no_bugfix' else: label_dict_consensus[line] = 'None' result['label_dict'] = label_dict result['label_dict_consensus'] = label_dict_consensus label_dict_by_type = {} for line, label in label_dict_consensus.items(): if label not in label_dict_by_type: label_dict_by_type[label] = [] label_dict_by_type[label].append(line) result['label_dict_by_type'] = label_dict_by_type commit_ratios_all = {} commit_ratios_java = {} commit_ratios_prod = {} for result in completed: if result['revision_hash'] not in commit_ratios_all: commit_ratios_all[result['revision_hash']] = {} #commit_ratios_java[result['revision_hash']] = {} #commit_ratios_prod[result['revision_hash']] = {} for label, lines in result['label_dict_by_type'].items(): if label not in commit_ratios_all[result['revision_hash']]: commit_ratios_all[result['revision_hash']][label] = 0 commit_ratios_all[result['revision_hash']][label] += len(lines) if java_filename_filter(result['file'], production_only=False): if result['revision_hash'] not in commit_ratios_java: commit_ratios_java[result['revision_hash']] = {} if label not in commit_ratios_java[result['revision_hash']]: commit_ratios_java[result['revision_hash']][label] = 0 commit_ratios_java[result['revision_hash']][label] += len(lines) if java_filename_filter(result['file'], production_only=True): if result['revision_hash'] not in commit_ratios_prod: commit_ratios_prod[result['revision_hash']] = {} if label not in commit_ratios_prod[result['revision_hash']]: commit_ratios_prod[result['revision_hash']][label] = 0 commit_ratios_prod[result['revision_hash']][label] += len(lines) commit_ratios_all_df = pd.DataFrame(commit_ratios_all).fillna(0).transpose() commit_ratios_all_df = commit_ratios_all_df.div(commit_ratios_all_df.sum(axis=1), axis=0) commit_ratios_java_df = pd.DataFrame(commit_ratios_java).fillna(0).transpose() commit_ratios_java_df = commit_ratios_java_df.div(commit_ratios_java_df.sum(axis=1), axis=0).fillna(0) commit_ratios_prod_df = pd.DataFrame(commit_ratios_prod).fillna(0).transpose() commit_ratios_prod_df = commit_ratios_prod_df.div(commit_ratios_prod_df.sum(axis=1), axis=0).fillna(0) commit_ratios_prod_df['total_lines'] = pd.DataFrame(commit_ratios_prod).fillna(0).transpose().sum(axis=1) for column in commit_ratios_all_df.columns: if column not in commit_ratios_java_df.columns: commit_ratios_java_df[column] = 0.0 if column not in commit_ratios_prod_df.columns: commit_ratios_prod_df[column] = 0.0 # create stats table col_order = ['bugfix', 'test', 'documentation', 'refactoring', 'unrelated', 'whitespace', 'test_doc_whitespace', 'no_bugfix', 'None'] box_labels = ['Bugfix', 'Test', 'Documentation', 'Refactoring', 'Unrelated Improvement', 'Whitespace', 'Test/Doc/Whitespace', 'No Bugfix', 'No Consensus'] for col, label in zip(col_order, box_labels): lower_all, upper_all = _confidence_interval(commit_ratios_all_df[col], 1-confidence_level, is_normal=False) lower_prod, upper_prod = _confidence_interval(commit_ratios_prod_df[col], 1-confidence_level, is_normal=False) gt_zero = sum(commit_ratios_prod_df[col]>0)/len(commit_ratios_prod_df) print("%22s , %3d , %3d , [%3d, %3d] , %3d , %3d , [%3d, %3d] , %3d" % (label, 100*commit_ratios_all_df[col].median(), 100*stats.median_abs_deviation(commit_ratios_all_df[col], scale=1/1.4826), 100*lower_all, 100*upper_all, 100*commit_ratios_prod_df[col].median(), 100*stats.median_abs_deviation(commit_ratios_prod_df[col], scale=1/1.4826), 100*lower_prod, 100*upper_prod, 100*gt_zero))