def testGetSampledUciData(self, mock_read_csv):
mock_df = mock.MagicMock()
mock_df.sample = mock.MagicMock()
mock_read_csv.return_value = mock_df
sample = 0.45 # Arbitrary number.
utils.get_uci_min_diff_datasets(sample=sample)
mock_read_csv.assert_called_once()
mock_df.sample.assert_called_once_with(
frac=sample, replace=True, random_state=1)
def testGetMinDiffDatasetsDefaults(self, mock_get_uci_data):
_ = utils.get_uci_min_diff_datasets()
mock_get_uci_data.assert_has_calls(
[mock.call(split='train', sample=None),
mock.call(split='train')],
any_order=True)
def testUciDataCached(self, mock_read_csv):
utils.get_uci_min_diff_datasets()
mock_read_csv.assert_called_once()
# pandas.read_csv should not be called again for the same split ('train')
mock_read_csv.reset_mock()
utils.get_uci_min_diff_datasets()
mock_read_csv.assert_not_called()
# pandas.read_csv should be called again for a different split
mock_read_csv.reset_mock()
utils.get_uci_min_diff_datasets(split='test')
mock_read_csv.assert_called_once()
# pandas.read_csv should not be called again (both splits have been cached)
mock_read_csv.reset_mock()
utils.get_uci_min_diff_datasets(split='train')
utils.get_uci_min_diff_datasets(split='test')
mock_read_csv.assert_not_called()
def testGetMinDiffDatasets(self, mock_get_uci_data, mock_df_to_dataset):
mock_md_ds1 = mock.MagicMock()
mock_md_ds1.batch.return_value = 'batched_d1'
mock_md_ds2 = mock.MagicMock()
mock_md_ds2.batch.return_value = 'batched_d2'
mock_df_to_dataset.side_effect = ['og_ds', mock_md_ds1, mock_md_ds2]
sample = 0.56 # Arbitrary sample value.
split = 'fake_split'
original_batch_size = 19 # Arbitrary size.
min_diff_batch_size = 23 # Different arbitrary size.
res = utils.get_uci_min_diff_datasets(
split=split,
sample=sample,
original_batch_size=original_batch_size,
min_diff_batch_size=min_diff_batch_size)
# Assert outputs come from the right place.
expected_res = ('og_ds', 'batched_d1', 'batched_d2')
self.assertEqual(res, expected_res)
# Assert proper calls have been made.
self.assertEqual(mock_get_uci_data.call_count, 2)
mock_get_uci_data.assert_has_calls(
[mock.call(split=split, sample=sample),
mock.call(split=split)],
any_order=True)
self.assertEqual(mock_df_to_dataset.call_count, 3)
mock_df_to_dataset.assert_has_calls([
mock.call(mock.ANY, shuffle=True, batch_size=original_batch_size),
mock.call(mock.ANY, shuffle=True),
mock.call(mock.ANY, shuffle=True),
])
mock_md_ds1.batch.assert_called_once_with(min_diff_batch_size,
drop_remainder=True)
mock_md_ds2.batch.assert_called_once_with(min_diff_batch_size,
drop_remainder=True)
def testGetUciDataWithBadSplitRaisesError(self):
with self.assertRaisesRegex(
ValueError, 'split must be.*train.*test.*given.*bad_split'):
utils.get_uci_min_diff_datasets('bad_split')
def testGetTestUciData(self, mock_read_csv):
utils.get_uci_min_diff_datasets(split='test')
mock_read_csv.assert_called_once_with(
utils._UCI_DATA_URL_TEMPLATE.format('test'),
names=utils._UCI_COLUMN_NAMES,
header=None)
def testGetTrainUciDataAsDefault(self, mock_read_csv):
_ = utils.get_uci_min_diff_datasets()
mock_read_csv.assert_called_once_with(
utils._UCI_DATA_URL_TEMPLATE.format('data'),
names=utils._UCI_COLUMN_NAMES,
header=None)