def main():
args = parse_arguments()
dataset = load_dataset(args.dataset_path)
users = dataset['data']['users']
# prepare id and name substitutions
# id substitution = hash(id)
id_substitutions = {
user['_id']: {
'substitution': sha256_hash(user['_id']),
'regex': user['_id']
}
for user in users
}
name_substitutions = {
user['name']: {
# name substitution = id substitution
'substitution': id_substitutions[user['_id']]['substitution'],
'regex': get_regex(user['name'])
}
for user in users
}
substitutions = [v['substitution'] for v in id_substitutions.values()]
# replace the user ids in actions
for action in dataset['data']['actions']:
if 'user' in action:
userid = action['user']
if userid not in id_substitutions:
id_substitutions[userid] = {
'substitution': sha256_hash(userid),
'regex': userid
}
action['user'] = id_substitutions[userid]['substitution']
# replace the user ids in appInstanceResources
for app_instance_res in dataset['data']['appInstanceResources']:
if 'user' in app_instance_res:
userid = app_instance_res['user']
if userid not in id_substitutions:
id_substitutions[userid] = {
'substitution': sha256_hash(userid),
'regex': userid
}
app_instance_res['user'] = id_substitutions[userid]['substitution']
# generically search and replace occurrences of usernames and ids
find_and_replace(dataset, name_substitutions)
find_and_replace(dataset, id_substitutions)
# the users fields are now just the hashes
dataset['data']['users'] = substitutions
save_dataset(dataset, args.output_path)
def main():
args = parse_arguments()
dataset = load_dataset(args.dataset_path)
# write your dataset changes here
save_dataset(dataset, args.output_path)
def main():
args = parse_arguments([{'name': 'positive_integer', 'type': int}])
dataset = load_dataset(args.dataset_path)
if args.positive_integer < 0:
raise Exception
save_dataset(dataset, args.output_path)
def main():
args = parse_arguments([{'name': 'fields', 'type': str}])
dataset = load_dataset(args.dataset_path)
fields = json.loads(args.fields)
iterate_and_suppress(dataset, fields)
save_dataset(dataset, args.output_path)
def main():
args = parse_arguments([{'name': 'password', 'type': str}])
dataset = load_dataset(args.dataset_path)
if args.password != 'PASSWORD':
raise Exception
save_dataset(dataset, args.output_path)
def main():
args = parse_arguments([{'name': 'fields', 'type': str}])
dataset = load_dataset(args.dataset_path)
fields = json.loads(args.fields)
relevant_arrays = find_selected_arrays(dataset, fields)
for array in relevant_arrays:
shuffle_attributes(array['array'], array['field_selection']),
save_dataset(dataset, args.output_path)
def main():
args = parse_arguments()
dataset = load_dataset(args.dataset_path)
for action in dataset['data']['actions']:
if 'user' in action:
action['user'] = sha256_hash(action['user'])
for appInstanceResource in dataset['data']['appInstanceResources']:
if 'user' in appInstanceResource:
appInstanceResource['user'] = sha256_hash(appInstanceResource['user'])
save_dataset(dataset, args.output_path)
def main():
args = parse_arguments()
dataset = load_dataset(args.dataset_path)
# hash the 'user' id for every action
for action in dataset['data']['actions']:
if 'user' in action:
action['user'] = sha256_hash(action['user'])
# hash the 'user' id for every appInstanceResource
for appInstanceResource in dataset['data']['appInstanceResources']:
if 'user' in appInstanceResource:
appInstanceResource['user'] = sha256_hash(
appInstanceResource['user'])
# hash the user '_id' for every user and remove every other attribute
new_users = []
for user in dataset['data']['users']:
if '_id' in user:
new_users.append({'_id': sha256_hash(user['_id'])})
dataset['data']['users'] = new_users
save_dataset(dataset, args.output_path)
def main():
args = parse_arguments([{'name': 'k', 'type': int}])
dataset = load_dataset(args.dataset_path)
actions = dataset['data']['actions']
# get the geolocations for each user
geolocations_per_user = {}
for action in actions:
if 'geolocation' in action:
user = action['user']
if user not in geolocations_per_user:
geolocations_per_user[user] = []
geoloc = action['geolocation']
country = geoloc.get('country')
region = geoloc.get('region')
city = geoloc.get('city')
geolocations_per_user[user].append((country, region, city))
# get most represented geolocation for each user
geolocation_mapping = {}
for user, geolocations in geolocations_per_user.items():
most_common = Counter(geolocations).most_common(1)
if len(most_common) == 1:
[((country, region, city), _)] = most_common
geolocation_mapping[user] = {
'country': country,
'region': region,
'city': city,
}
# group users by country then region then city
grouped = {}
for user, geo in geolocation_mapping.items():
country = geo['country']
region = geo['region']
city = geo['city']
if country not in grouped:
grouped[country] = {'count': 0, 'regions': {}}
country_group = grouped[country]
country_group['count'] += 1
country_regions = country_group['regions']
if region not in country_regions:
country_regions[region] = {'count': 0, 'cities': {}}
region_group = country_regions[region]
region_group['count'] += 1
region_cities = region_group['cities']
if city not in region_cities:
region_cities[city] = {'count': 0, 'users': []}
city_group = region_cities[city]
city_group['count'] += 1
city_group['users'].append(user)
# remove each value that is not represented at least k times
for country, country_group in grouped.items():
country_user_count = country_group.get('count', 0)
regions = country_group.get('regions', {})
for region, region_group in regions.items():
region_user_count = region_group.get('count', 0)
cities = region_group.get('cities', {})
for city, city_group in cities.items():
city_user_count = city_group.get('count', 0)
users = city_group.get('users', [])
if city_user_count < args.k:
for user in users:
geolocation_mapping[user]['city'] = ''
if region_user_count < args.k:
for user in users:
geolocation_mapping[user]['region'] = ''
if country_user_count < args.k:
for user in users:
geolocation_mapping[user]['country'] = ''
# update with new values
for action in actions:
if 'geolocation' in action:
user = action['user']
action['geolocation'] = geolocation_mapping[user]
save_dataset(dataset, args.output_path)
