def test_basic_sum_one_group(self):
"""
Test summing for a single group on a matrix which is small enough to
verify the results by hand
"""
group_definition = [(0, "even"), (1, "odd"), (2, "even"), (3, "odd")]
rows = [[1, 2, 3, 4], [2, 3, 4, 5], [3, 4, 5, 6], [4, 5, 6, 7]]
matrix = numpy.array(rows)
row_grouper = RowGrouper(group_definition)
group_sum = row_grouper.sum_one_group(matrix, "even")
self.assertEqual(group_sum.tolist(), [4, 6, 8, 10])
def test_sum_one_group_with_all_group_and_matrix_type_combinations(self):
"""
Tests the `sum_one_group` method with every combination of group type
and matrix type
"""
test_cases = product(self.get_group_definitions(), self.get_matrices())
for (group_name, group_definition), (matrix_name, matrix) in test_cases:
# Use `subTest` when we upgrade to Python 3
# with self.subTest(matrix=matrix_name, group=group_name):
row_grouper = RowGrouper(group_definition)
# Calculate sums for all groups the boring way using pure Python
expected_values = self.sum_rows_by_group(group_definition, matrix)
# Check the `sum_one_group` gives the expected answer for all groups
for offset, group_id in enumerate(row_grouper.ids):
expected_value = expected_values[offset]
value = row_grouper.sum_one_group(matrix, group_id)
# We need to round floats to account for differences between
# numpy and Python float rounding
self.assertEqual(
round_floats(value.tolist()), round_floats(expected_value)
)