def test_dyad_aggression_by_year(self):
rv = PandasGdeltHelper.dyad_aggression_by_year('series')
assert rv is not None
logging.warning("These tests will all be dependent on using the sample data.")
assert rv[18] == 1.0
assert rv.index[18] == ('ALBGOV', 'CUB', 1982, False)
# From here... sanity check but not quite the same as accessing one row by MultiIndex
assert 1982 == rv.index.levels[2][rv.index.labels[2][27]]
df = PandasGdeltHelper.dyad_aggression_by_year()
assert df is not None
assert 'AUS' == df.loc(0)[25]['actor1code']
def test_events_from_local_files(self):
if LONG_TEST_TOLERANCE < 120:
self.skipTest('Test takes 2 minutes or more. Would be better to work on some smaller '
'stubbed files.')
helper = PandasGdeltHelper(table_name='events', data_source='local')
rv = helper.events()
assert rv is not None
logging.warning("This isn't testing that the data being used is local data.")
def test_events_default(self):
if LONG_TEST_TOLERANCE < 120:
self.skipTest('Test takes 2 minutes or more. Would be better to work on some smaller '
'stubbed files.')
#... and that would necessitate changing the interface to allow passed-in settings,
# which is a very good thing.
helper = PandasGdeltHelper(table_name='events')
rv = helper.events()
assert rv is not None
def prepare_data(self):
self._events_data = PandasGdeltHelper.events()
columns_for_X = self._events_data[['fractiondate', 'goldsteinscale']]
X = np.reshape(np.array(columns_for_X),
(self._events_data.shape[0], -1))
y = np.reshape(np.array(self._events_data.avgtone),
(self._events_data.shape[0], -1))
(
self._X_train,
self._X_test,
self._y_train,
self._y_test,
) = train_test_split(X, y, test_size=0.33)
assert (self._events_data is not None)
def test_dyad_events_by_year(self):
"""This is my present thinking about what the 'conventional' path going forward should be:
Instantiate the helper class with a table name and load the table.
However I don't yet know how to deal with other functionality that doesn't pertain to a
table in the datafiles. Thus, for now, something like test_dyad_aggression_by_year will
still use the class methods rather than instance methods. The obvious solution is to
somehow make the PandasGdeltHelper class smart enough to handle 'virtual tables'
like dyad_aggression_by_year that requires some sort of data munging.
That said I need to think whether the logic (what are aggressive codes?) needs to live in
the data helper or in the classification file."""
helper = PandasGdeltHelper('dyad_events_by_year')
assert helper is not None
self.skipTest('fetch() method is not yet fully implemented.')
rv = helper.fetch()
assert rv.shape[0] > 0
assert rv.shape[1] > 0
assert rv is not None
def test_country_features(self):
df = PandasGdeltHelper.country_features()
new_columns = ['proportion_actor1', 'aggregate_relationships',]
assert True == all([df[c] is not None for c in new_columns])
def test_events_from_sample_data(self):
helper = PandasGdeltHelper(table_name='events', data_source='sample')
rv = helper.events()
assert rv is not None
logging.warning("This isn't testing that the data being used is sample data.")