def get_dataframe_from_repo(repo, num=100):
"""Create pandas dataframe of contributors by country.
Args:
repo - a full GitHub repo URL
num - number of contributors to analyze per repo
Returns:
df - a pandas dataframe of contributors by country
num_contributors - total number of contributors
"""
# get contributors
repo_ending_string = extract_github_owner_and_repo(repo)
contributors = get_contributors(repo_ending_string, num)
num_contributors = len(contributors)
# get count of countries
country_list = []
for contributor in contributors:
location = get_contributor_location(contributor)
country = get_country_from_location(location)
country_list.append(country)
country_counter = Counter(country_list)
# convert counter to pandas dataframe
df = pd.DataFrame.from_records(country_counter.most_common(),
columns=["country", "contributor_count"])
return df, num_contributors
def test_get_country_from_location_world_cities(self):
"""test get_country_from_location on world city names."""
assert get_country_from_location("Tokyo") == "Japan"
assert get_country_from_location("London") == "United Kingdom"
assert get_country_from_location("Jakarta") == "Indonesia"
assert get_country_from_location("Beijing") == "China"
assert get_country_from_location("Washington D.C.") == "United States"
assert get_country_from_location("Toronto, ON") == "Canada"
def print_by_contributor(software_name,
contributors,
output_csv=False,
pypi_data=None):
"""Print location results by contributor.
Print contributors and countries to terminal window. If output csv is set
to true, then also output results to a csv file.
Args:
software_name - name of package or repo
contributors - a list of contributors
output_csv - whether to output a csv.
pypi_data - a pypi data object.
Returns:
null
"""
# create csv if output_csv specified
if output_csv:
# unique current time timestamp to create unique filename
timestamp = time.strftime("%Y%m%d-%H%M%S")
create_csv("contributor", timestamp)
print("CONTRIBUTOR, LOCATION")
if pypi_data is not None:
print("* indicates PyPI maintainer")
print("---------------------")
for contributor in contributors:
location = get_contributor_location(contributor)
country = get_country_from_location(location)
if output_csv:
add_committer_to_csv("contributor", software_name, timestamp,
contributor, location, country)
try:
# Check if pypi_data is not None, indicating a PyPI package scan
if pypi_data is not None and contributor in pypi_data[
"pypi_maintainers"]:
print(contributor, "*", "|", location, "|", country)
else:
print(contributor, "|", location, "|", country)
except UnicodeEncodeError:
print(contributor, "| error")
def print_by_country(contributors):
"""Print contributors aggregated by country.
Print contributor county by country to terminal window.
Args:
contributors: a list of contributors
Returns:
null
"""
print("COUNTRY | # OF CONTRIBUTORS")
print("---------------------------")
country_list = []
for contributor in contributors:
location = get_contributor_location(contributor)
country = get_country_from_location(location)
country_list.append(country)
country_counter = Counter(country_list)
for country, count in country_counter.most_common():
print(country, count)
def scan_multiple_repos(input_file="repos.txt", num=100):
"""Create csv of data for multiple repos.
Scan through repos provided in repos.txt and create a single csv that
stores all contributor-related data for each contributor in each repo.
Args:
input_file - file containing repo list
num - max number of contributors to analyze per repo
Returns:
None
"""
# create csv to store multi-repo scan results
timestamp = time.strftime("%Y%m%d-%H%M%S")
create_csv("multirepo", timestamp)
# open file that contains repos to scan and append contributors for each
# repo to csv. Also, repos.txt must contain repo names, one repo per line.
with open(input_file, "r") as input_repos:
for repo in input_repos:
# Skip blank lines
if repo == "":
continue
# strip blank space before extracting owner and repo name
repo_ending_string = extract_github_owner_and_repo(repo.strip())
contributors = get_contributors(repo_ending_string, num)
for contributor in contributors:
location = get_contributor_location(contributor)
country = get_country_from_location(location)
add_committer_to_csv(
"multirepo",
repo_ending_string,
timestamp,
contributor,
location,
country,
)
def test_get_country_from_location_dataset_pull_geographies(self):
"""tests of get_gountry_from_location() that fail as of 2/14/2021"""
assert get_country_from_location("Saclay") == "France"
assert get_country_from_location("Warszawa") == "Poland"
assert get_country_from_location("brookline, ma") == "United States"
assert get_country_from_location("Greater Los Angeles Area") == "United States"
assert get_country_from_location("Forschungszentrum") == "Germany"
assert get_country_from_location("Montigny-lès-Metz") == "France"
assert get_country_from_location("roudnice nad labem, czech republic") == "Czech Republic"
assert get_country_from_location("Berlin/Florence") == "Germany"
assert get_country_from_location("Greater Seattle Area") == "United States"
assert get_country_from_location("Flanders, Europe, Earth") == "Belgium"
assert get_country_from_location("Wrocław") == "Poland"
def test_get_country_from_location_corner_case_geographies(self):
"""test get_country_from_location on unusual geographies."""
assert get_country_from_location("Palestine") == "Palestine"
assert get_country_from_location("San Francisco Bay Area") == "United States"
assert get_country_from_location("EU") == "None"
assert get_country_from_location("Canary Islands") == "Spain"
assert get_country_from_location("Earth") == "None"
assert get_country_from_location("Sydney") == "Australia"
assert get_country_from_location("Amsterdam") == "Netherlands"
assert get_country_from_location("NYC") == "United States"
assert get_country_from_location("Barcelona") == "Spain"
assert get_country_from_location("Kerala") == "India"
assert get_country_from_location("Hyderabad") == "India"
assert get_country_from_location("Vancouver") == "Canada"
assert get_country_from_location("Jiangxi") == "China"
assert get_country_from_location("San Francisco") == "United States"
assert get_country_from_location("New York") == "United States"
assert get_country_from_location("Saint Petersburg") == "Russia"
assert get_country_from_location("England") == "United Kingdom"
assert get_country_from_location("Athens") == "Greece"
assert get_country_from_location("Europe") == "None"
assert get_country_from_location("Lima") == "Peru"
assert get_country_from_location("Bay Area") == "United States"
assert get_country_from_location("EU") == "None"
assert get_country_from_location("Canary Islands") == "Spain"
assert get_country_from_location("waterloo") == "United Kingdom"
assert get_country_from_location("Europe/Berlin") == "None"
assert get_country_from_location("York") == "United Kingdom"
assert get_country_from_location("München") == "Germany"
assert get_country_from_location("Montreal, CA") == "Canada"
assert get_country_from_location("Florianópolis") == "Brazil"
assert get_country_from_location("Montréal") == "Canada"
assert get_country_from_location("Bangalore") == "India"
assert get_country_from_location("Dublin") == "Ireland"
assert get_country_from_location("Santiago de Querétaro, México") == "Mexico"
assert get_country_from_location("Jülich") == "Germany"
assert get_country_from_location("Victoria, BC") == "Canada"
assert get_country_from_location("Waterloo, ON") == "Canada"
assert get_country_from_location("Falls Church, Virginia") == "United States"
assert get_country_from_location("Amsterdam, the Netherlands") == "Netherlands"
assert get_country_from_location("BeiJing") == "China"
assert get_country_from_location("Edinburgh, Scotland") == "United Kingdom"
assert get_country_from_location("Medellín, Colombia") == "Colombia"
assert get_country_from_location("La Jolla, CA.") == "United States"
assert get_country_from_location("beijing") == "China"
assert get_country_from_location("Pemberton, British Columbia") == "Canada"
assert get_country_from_location("Timi»ôoara") == "Romania"
assert get_country_from_location("PRC") == "China"
assert get_country_from_location("Amsterdam, The Netherlands") == "Netherlands"
assert get_country_from_location("Oxford") == "United Kingdom"
assert get_country_from_location("S√£o Paulo") == "Brazil"
assert get_country_from_location("Kyiv") == "Ukraine"
assert get_country_from_location("Vancouver, BC") == "Canada"
assert get_country_from_location("N.H.") == "United States"
assert get_country_from_location("Sri-City, Andhra Pradesh") == "India"
assert get_country_from_location("Scotland") == "United Kingdom"
assert get_country_from_location("Geneva") == "Switzerland"
assert get_country_from_location("Rotterdam, the Netherlands") == "Netherlands"
assert get_country_from_location("Milan") == "Italy"
assert get_country_from_location("Republic of Korea") == "South Korea"
assert get_country_from_location("Brasília, Brazil.") == "Brazil"
assert get_country_from_location("beijing") == "China"
assert get_country_from_location("Zürich") == "Switzerland"
assert get_country_from_location("Kitchener, Ontario") == "Canada"
assert get_country_from_location("Montréal, QC") == "Canada"
assert get_country_from_location("Glasgow, Scotland") == "United Kingdom"
assert (
get_country_from_location("28 rue du Dr Roux 75015 Paris, FRANCE")
== "France"
)
assert get_country_from_location("Kraków") == "Poland"
assert get_country_from_location("İstanbul") == "Turkey"
assert get_country_from_location("Russian Federation") == "Russia"
assert get_country_from_location("Newcastle, NSW") == "Australia"
assert get_country_from_location("Australia, Victoria") == "Australia"
assert get_country_from_location("Perth, Western Australia ") == "Australia"
assert get_country_from_location("Gdańsk") == "Poland"
assert get_country_from_location("SF") == "United States"
assert get_country_from_location("Hyderabad (India)") == "India"
assert get_country_from_location("BITS Pilani, Rajasthan") == "India"
assert get_country_from_location("Sri-City, Andhra Pradesh") == "India"
def test_get_country_from_location_country_abbreviations(self):
"""test get_country_from_location on country abbreviations."""
assert get_country_from_location("USA") == "United States"
assert get_country_from_location("Cambridge, UK") == "United Kingdom"
assert get_country_from_location("UK") == "United Kingdom"
def test_get_country_from_location_standard_order_no_comma(self):
"""test get_country_from_location on standard order pairs without comma."""
assert get_country_from_location("Menlo Park CA") == "United States"
def test_get_country_from_location_nonstandard_order(self):
"""test get_country_from_location on non-standard order pairs."""
assert get_country_from_location("Russia, Moscow") == "Russia"
assert get_country_from_location("Russia, Nizhny Novgorod") == "Russia"
def test_get_country_from_location_standard_order_with_comma(self):
"""test get_country_from_location on standard order pairs with comma."""
assert get_country_from_location("Wellington, New Zealand") == "New Zealand"
assert get_country_from_location("Jordan, Minnesota") == "United States"
assert get_country_from_location("Jordan, MN") == "United States"
assert get_country_from_location("Atlanta, Georgia") == "United States"
assert get_country_from_location("Atlanta, Ga") == "United States"
assert get_country_from_location("London, England") == "United Kingdom"
assert get_country_from_location("Prague, Czech Republic") == "Czech Republic"
assert get_country_from_location("Virginia, USA") == "United States"
assert get_country_from_location("Naperville, IL") == "United States"
assert get_country_from_location("Toronto, Ontario, Canada") == "Canada"
assert get_country_from_location("Berlin, DE") == "Germany"
assert get_country_from_location("CSU Sacramento") == "United States"
assert get_country_from_location("Philadelphia, PA") == "United States"