def test_write_to_file(self):
persons = MagicMock()
households = MagicMock()
allocator = HouseholdAllocator(households, persons)
allocator.write(person_file='persons_file',
household_file='households_file')
persons.to_csv.assert_called_once_with('persons_file')
households.to_csv.assert_called_once_with('households_file')
def _mock_allocated(self):
def mock_person(serialno, age, sex, income):
return {
'serial_number': serialno,
'age': age,
'sex': sex,
'individual_income': income
}
allocated_persons = pandas.DataFrame([
mock_person('b', '35-64', 'F', '40k+'),
mock_person('b', '35-64', 'M', 'None'),
])
def mock_household(serialno, num_people, num_vehicles, income, count,
tract):
return {
'serial_number': serialno,
'num_people': num_people,
'num_vehicles': num_vehicles,
'household_income': income,
'count': count,
'tract': tract,
}
allocated_households = pandas.DataFrame([
mock_household('b', '6+', '2', '40k+', count=2, tract='tract1'),
mock_household('b', '6+', '2', '40k+', count=2, tract='tract2'),
])
return HouseholdAllocator(allocated_households, allocated_persons)
def test_read_from_file(self):
read_csv = MagicMock(return_value=pandas.DataFrame())
with patch('pandas.read_csv', read_csv):
allocator = HouseholdAllocator.from_csvs('households_file',
'persons_file')
assert type(allocator) == HouseholdAllocator
read_csv.assert_any_call('households_file')
read_csv.assert_any_call('persons_file')
def download_tract_data(state_id, puma_id, output_dir, census_api_key,
puma_tract_mappings, households_data, persons_data):
'''Download tract data from the US Census' API.
Initilize an allocator, capable of allocating PUMS households as best as possible based on
marginal census (currently tract) data using a cvx-solver.
Args:
state_id: 2-digit state fips code
puma_id: 5-digit puma code
output_dir: dir to write outWriter the generated bayesian nets to
census_api_key: key used to download data from the U.S. Census
puma_tract_mappings: filepath to the puma-tract mappings
households_data: pums households data frame
persons_data: pums persons data frame
Returns:
An allocator described above.
'''
marginal_path = os.path.join(
output_dir, FILE_PATTERN.format(state_id, puma_id, 'marginals.csv'))
try: # Already have marginals file
marginals = Marginals.from_csv(marginal_path)
except Exception: # Download marginal data from the Census API
with builtins.open(puma_tract_mappings) as csv_file:
csv_reader = csv.DictReader(csv_file)
marginals = Marginals.from_census_data(csv_reader,
census_api_key,
state=state_id,
pumas=puma_id)
if len(marginals.data) <= 1:
logging.exception(
'Couldn\'t fetch data from the census. Check your API key')
raise CensusFetchException()
else:
logging.info(
'Writing outWriter marginal file for state: %s, puma: %s',
state_id, puma_id)
marginals.write(marginal_path)
'''With the above marginal controls (tract data), the methods in allocation.py
allocate discrete PUMS households to the subject PUMA.'''
try:
allocator = HouseholdAllocator.from_cleaned_data(
marginals=marginals,
households_data=households_data,
persons_data=persons_data)
except Exception as e:
logging.exception('Error Allocating state: %s, puma: %s\n%s', state_id,
puma_id, e)
__builtin__.exit()
return marginals, allocator
def test_from_cleaned_data(self):
# Prepare pums data
households_data = CleanedData(self._mock_household_data())
persons_data = CleanedData(self._mock_person_data())
# Prepare marginals
marginals = Marginals(self._mock_tract_data())
allocator = HouseholdAllocator.from_cleaned_data(
marginals, households_data, persons_data)
self.assertTrue(allocator)
expected_shape = (114, 17)
self.assertEqual(allocator.allocated_households.shape, expected_shape)
expected_columns = [
u'serial_number', u'num_people', u'num_vehicles',
u'household_weight', u'num_people_1', u'num_people_2',
u'num_people_3', u'num_vehicles_0', u'num_vehicles_1',
u'num_vehicles_2', u'num_vehicles_3+', u'age_0-17', u'age_18-34',
u'age_65+', u'age_35-64', u'count', u'tract'
]
self.assertEqual(set(allocator.allocated_households.columns.tolist()),
set(expected_columns))