def compute_repo_ccp(repo_file):
rep = pd.read_csv(repo_file)
rep['y2016_hit_rate'] = 1.0 * rep.y2016_hits / rep.y2016_commits
rep['y2017_hit_rate'] = 1.0 * rep.y2017_hits / rep.y2017_commits
rep['y2018_hit_rate'] = 1.0 * rep.y2018_hits / rep.y2018_commits
rep['y2019_hit_rate'] = 1.0 * rep.y2019_hits / rep.y2019_commits
rep['y2016_ccp'] = rep.y2016_hit_rate.map(
lambda x: ccp_estimator.estimate_positives(x))
rep['y2017_ccp'] = rep.y2017_hit_rate.map(
lambda x: ccp_estimator.estimate_positives(x))
rep['y2018_ccp'] = rep.y2018_hit_rate.map(
lambda x: ccp_estimator.estimate_positives(x))
rep['y2019_ccp'] = rep.y2019_hit_rate.map(
lambda x: ccp_estimator.estimate_positives(x))
rep['y2019_hit_rate_rnd'] = rep.y2019_hit_rate.map(lambda x: round(x, 2))
# extract to repo utils
trep = rep[rep.fork == False]
trep = trep[trep.y2019_hit_rate.map(
lambda x: ccp_estimator.is_in_range(x))]
trep = trep.sort_values(['y2019_hit_rate'], ascending=False)
y2019_hit_rate_10p = trep.iloc[int(90 * len(trep) / 100)].y2019_hit_rate
rep['quality_group'] = rep.apply(
lambda x: 'Others'
if x.y2019_hit_rate > y2019_hit_rate_10p else 'Top 10',
axis=1)
rep['dev_num_group'] = pd.cut(rep.authors, [
0,
rep.authors.quantile(0.25),
rep.authors.quantile(0.75),
float("inf")
],
labels=["few", "intermediate", "numerous"])
rep['start_year'] = rep.start_time.map(lambda x: int(x[:4]))
rep['age'] = ANALYZED_YEAR - rep.start_year
rep['age_group'] = pd.cut(rep.start_year, [
0, GITHUB_START_YEAR - 1,
rep[rep.start_year >= GITHUB_START_YEAR].start_year.quantile(0.25),
rep[rep.start_year >= GITHUB_START_YEAR].start_year.quantile(0.75),
float("inf")
],
labels=['prehistory', "old", "medium", "young"])
rep['y2019_ccp_in_valid_range'] = rep.y2019_hit_rate.map(
lambda x: ccp_estimator.is_in_range(x))
# Updating the file with the CCP values
rep.to_csv(repo_file, index=False)
return rep
def onboarding_ccp_cochange(repo_file_quality_per_year,
repo_file_onboarding_per_year):
"""
Note that repo_file_quality_per_year uses bug hit ratio and not ccp.
For change analysis it doesn't matter.
:param repo_file_quality_per_year:
:return:
"""
key = 'repo_name'
repo_file_quality_per_year_df = build_repo_per_year_df(
repo_file_quality_per_year, key=key)
repo_file_onboarding_per_year_df = build_repo_per_year_df(
repo_file_onboarding_per_year, key=key)
repo_file_onboarding_per_year_df = repo_file_onboarding_per_year_df[
repo_file_onboarding_per_year_df.comming_developers > 9]
per_year_df = pd.merge(repo_file_quality_per_year_df,
repo_file_onboarding_per_year_df,
on=[key, 'year'])
per_year_df['ccp'] = per_year_df.corrective_commits_ratio.map(
lambda x: ccp_estimator.estimate_positives(x))
cochange_analysis(per_year_df,
first_metric='ccp',
second_metric='comming_involved_developers_ratio',
first_the_higher_the_better=False,
second_the_higher_the_better=True,
first_sig_threshold=0.1,
second_sig_threshold=0.1,
key=key)
def churn_ccp_cochange_by_age(repo_file_quality_per_year,
repo_file_churn_per_year):
"""
Note that repo_file_quality_per_year uses bug hit ratio and not ccp.
For change analysis it doesn't matter.
:param repo_file_quality_per_year:
:return:
"""
key = 'repo_name'
control_variables = ['age_group']
repo_file_quality_per_year_df = build_repo_per_year_df(
repo_file_quality_per_year,
key=key,
control_variables=control_variables)
repo_file_churn_per_year_df = build_repo_per_year_df(
repo_file_churn_per_year, key=key, control_variables=control_variables)
repo_file_churn_per_year_df = repo_file_churn_per_year_df[(
repo_file_churn_per_year_df.base_year_developers > 9)]
per_year_df = pd.merge(repo_file_quality_per_year_df,
repo_file_churn_per_year_df,
on=[key, 'year'] + control_variables)
per_year_df['ccp'] = per_year_df.corrective_commits_ratio.map(
lambda x: ccp_estimator.estimate_positives(x))
cochange_analysis_by_value(per_year_df,
first_metric='ccp',
second_metric='continuing_developers_ratio',
first_the_higher_the_better=False,
second_the_higher_the_better=True,
first_sig_threshold=0.1,
second_sig_threshold=0.1,
fixed_variable='age_group',
fixed_values=['old', 'medium', 'young'],
key=key,
control_variables=control_variables)
def plot_duration_by_length():
df = get_dataset()
df['CCP'] = df['corrective_rate'].map(
lambda x: round(ccp_estimator.estimate_positives(x), 2))
fig = plot_deciles(df,
grouping_column='line_count',
metric_column='CCP',
title='CCP by Line Count Deciles',
xaxis_title='Number of Lines',
output_file=None)
fig.show()
print("Perason corrective rate and line count",
df.corr()['line_count']['corrective_rate'])
def onboarding_ccp_cochange_by_lang(repo_file_quality_per_year,
repo_file_onboarding_per_year):
"""
Note that repo_file_quality_per_year uses bug hit ratio and not ccp.
For change analysis it doesn't matter.
:param repo_file_quality_per_year:
:return:
"""
key = 'repo_name'
fixed_variable = 'language'
control_variables = [fixed_variable]
repo_file_quality_per_year_df = build_repo_per_year_df(
repo_file_quality_per_year,
key=key,
control_variables=control_variables)
repo_file_churn_per_year_df = build_repo_per_year_df(
repo_file_onboarding_per_year,
key=key,
control_variables=control_variables)
repo_file_churn_per_year_df = repo_file_churn_per_year_df[
(repo_file_churn_per_year_df.comming_developers > 9)
& (repo_file_churn_per_year_df.language.isin(lang_name))]
per_year_df = pd.merge(repo_file_quality_per_year_df,
repo_file_churn_per_year_df,
on=[key, 'year'] + control_variables)
per_year_df['ccp'] = per_year_df.corrective_commits_ratio.map(
lambda x: ccp_estimator.estimate_positives(x))
cochange_analysis_by_value(
per_year_df,
first_metric='ccp',
second_metric='comming_involved_developers_ratio',
first_the_higher_the_better=False,
second_the_higher_the_better=True,
first_sig_threshold=0.1,
second_sig_threshold=0.1,
fixed_variable=fixed_variable,
fixed_values=lang_name,
key=key,
control_variables=control_variables)
def generete_computed_values(df: pd.DataFrame):
print()
print("% Computed values")
print(r"\newcommand \filesNum {" + "{:,d}".format(len(df)) + " }")
print(r"\newcommand \reposNum {" + "{:,d}".format(df.repo_name.nunique()) +
" }")
print(r"\newcommand \meanCCP {" + str(round(get_average_ccp(), 2)) + " }")
print(r"\newcommand \hotspotCCPThreshold {" + "{:,.2f}".format(
ccp_estimator.estimate_positives(get_hotspots_corrective_threshold()))
+ " }")
print(r"\newcommand \reducedRiskRatio {" +
str(round(len(df[df.quality_group == 'reduced_risk']) /
len(df), 2)) + " }")
print(r"\newcommand \shortLengthThreshold {" +
"{:,.0f}".format(get_short_length_threshold()) + " }")
print(r"\newcommand \longLengthThreshold {" +
"{:,.0f}".format(get_long_length_threshold()) + " }")
print(r"\newcommand \robustSmellNum {" + str(len(get_robust_smells())) +
" }")
print()