def test_sga_multi_bins(self):
exp = self.getExperiment([self.derived_kpi_1['name']],
[self.derived_kpi_1])
dimension_to_bin = {
"normal_same": [
Bin("numerical", 1, 2, True, False),
Bin("numerical", 2, 3, True, False)
]
}
sga_result = exp.sga(dimension_to_bin)
self.assertEqual(len(sga_result), 2)
dimension_name = find_list_of_dicts_element(sga_result, "segment",
"[1, 2)", "dimension")
self.assertEqual(dimension_name, 'normal_same')
def test_sga_invalid_bin_list_type(self):
exp = self.getExperiment([self.derived_kpi_1['name']],
[self.derived_kpi_1])
dimension_to_bin = {"normal_same": Bin("numerical", 1, 10, True, True)}
with self.assertRaisesRegexp(
TypeError,
"Value of the input dict needs to be a list of Bin objects."):
exp.sga(dimension_to_bin)
def test_sga_invalid_dimension_name(self):
exp = self.getExperiment([self.derived_kpi_1['name']],
[self.derived_kpi_1])
dimension_to_bin = {
"dimension_does_not_exist": [Bin("numerical", 1, 10, True, True)]
}
with self.assertRaises(KeyError):
exp.sga(dimension_to_bin)
def test_sga_invalid_dimension_type(self):
exp = self.getExperiment([self.derived_kpi_1['name']],
[self.derived_kpi_1])
dimension_to_bin = {123: [Bin("numerical", 1, 10, True, True)]}
with self.assertRaisesRegexp(
TypeError,
"Key of the input dict needs to be string, indicating the name of dimension."
):
exp.sga(dimension_to_bin)
def test_sga_one_bin(self):
exp = self.getExperiment([self.derived_kpi_1['name']],
[self.derived_kpi_1])
dimension_to_bin = {
"normal_same": [Bin("numerical", 1, 10, True, True)]
}
sga_result = exp.sga(dimension_to_bin)
self.assertEqual(len(sga_result), 1)
subgroup_res = sga_result[0]
self.assertEqual(subgroup_res['dimension'], 'normal_same')
self.assertEqual(subgroup_res['segment'], '[1, 10]')
self.assertTrue("result" in subgroup_res)
def test_sga_not_valid_data_for_one_subgroup(self):
'''
It should not raise error if there is not enough data in one subgroup,
which means one subgroup might contain zero or only one variant. (e.g. "browser" = "hack name")
sga should ignore this subgroup and continue doing analysis with other subgroups.
'''
exp = self.getExperiment([self.derived_kpi_1['name']],
[self.derived_kpi_1])
dimension_to_bin = {
"normal_same": [
Bin("numerical", 1, 2, True, False),
Bin("numerical", 2, 3, True, False),
Bin("numerical", 999998, 999999, False, False)
], # bin which does not have any data
"feature": [
Bin("categorical", ["has"]),
Bin("categorical", ["non"]),
Bin("categorical", ["feature that only has one data point"
]) # bin which has only one data point
]
}
sga_result = exp.sga(dimension_to_bin)
self.assertEqual(len(sga_result), 4)
def test_sga_invalid_dimension_type(self):
exp = self.getExperiment([self.derived_kpi_1['name']],
[self.derived_kpi_1])
dimension_to_bin = {123: [Bin("numerical", 1, 10, True, True)]}
with self.assertRaises(TypeError):
exp.sga(dimension_to_bin)
def test_sga_invalid_bin_list(self):
exp = self.getExperiment([self.derived_kpi_1['name']],
[self.derived_kpi_1])
dimension_to_bin = {"normal_same": Bin("numerical", 1, 10, True, True)}
with self.assertRaises(TypeError):
exp.sga(dimension_to_bin)