def test_raise_assertion_error_with_duplicate_features(self, tmp_path):
config_STRING = """
raw_data_dir: "dummy"
dataset_name: "dummy"
base_features:
- name: "TIME"
dtype: DATETIME
transforming_features:
- name: "weekday"
index: 1
dtype: STRING
dependencies:
- "TIME"
- name: "weekday"
index: 2
dtype: STRING
"""
config_path = tmp_path / "tmp.yaml"
write_str_to_file(config_STRING, config_path)
with AssertRaises(AssertionError) as assert_raises:
feature_config_helper.FeatureConfigHelper(config_path)
error_message = assert_raises.expected_exception_found
assert_eq(
True,
error_message.args[0].startswith(
"There are duplicate objects in the list: "
),
)
def test_extract_config_1(self, tmp_path):
subset_features = ["e"]
expected_STRING = """
raw_data_dir: "dummy"
dataset_name: "dummy"
base_features:
- name: "a"
dtype: STRING
transforming_features:
- name: "b"
index: 1
dtype: STRING
dependencies:
- "a"
- name: "c"
index: 2
dtype: STRING
dependencies:
- "a"
- "b"
- name: "e"
index: 4
dtype: STRING
dependencies:
- "c"
"""
new_config_path = str(tmp_path / "new_tmp.yaml")
write_str_to_file(expected_STRING, new_config_path)
new_config = self.fm_helper.extract_config(selected_features=subset_features)
assert_eq(parse_feature_config(new_config_path), new_config)
def test_raise_value_error_with_invalid_indexes(self, tmp_path):
invalid_index_STRING = """
# invalid config with indexes are not continuous
raw_data_dir: "dummy"
dataset_name: "dummy"
base_features:
- name: "TIME"
dtype: "DATETIME"
transforming_features:
- name: "weekday"
index: 1
dtype: INT32
dependencies:
- "TIME"
- name: "hour"
index: 1
dtype: INT32
dependencies:
- "TIME"
"""
config_path = tmp_path / "tmp.yaml"
write_str_to_file(invalid_index_STRING, config_path)
with AssertRaises(ValueError) as assert_raises:
feature_config_helper.FeatureConfigHelper(config_path)
error_message = assert_raises.expected_exception_found
assert_eq(
True,
error_message.args[0].startswith(
"Feature indexes must be a list of increasing positive integers. "
"Got indexes = [1, 1]"
),
)
def setup_class(cls, tmp_path):
feature_config_str = """
raw_data_dir: "dummy"
dataset_name: "dummy"
base_features:
- name: "a"
dtype: STRING
transforming_features:
- name: "b"
index: 1
dtype: STRING
dependencies:
- "a"
- name: "e"
index: 4
dtype: STRING
dependencies:
- "c"
- name: "c"
index: 2
dtype: STRING
dependencies:
- "a"
- "b"
- name: "d"
index: 3
dtype: STRING
dependencies:
- "a"
"""
config_path = tmp_path / "feature_config_str.yaml"
write_str_to_file(feature_config_str, config_path)
cls.fm_helper = feature_config_helper.FeatureConfigHelper(config_path)
def test_raise_value_error_with_invalid_dependencies(self, tmp_path):
invalid_dependency_STRING = """
raw_data_dir: "dummy"
dataset_name: "dummy"
base_features:
- name: "TIME"
dtype: DATETIME
transforming_features:
- name: "weekday"
index: 1
dtype: STRING
dependencies:
- "date"
"""
config_path = tmp_path / "tmp.yaml"
write_str_to_file(invalid_dependency_STRING, config_path)
with AssertRaises(AssertionError) as assert_raises:
feature_config_helper.FeatureConfigHelper(config_path)
error_message = assert_raises.expected_exception_found
assert_eq(
True,
error_message.args[0].startswith(
"Feature weekday depends on feature date that is undefined."
),
)
def test_from_lines_in_txt(self, tmp_path):
file_content = """a
d
c
"""
txt_path = tmp_path / "foo.txt"
write_str_to_file(file_content, txt_path)
got = data_processing.CategoryEncoder.from_lines_in_txt(txt_path).get_encoded(
self.series
)
expected = pd.Series([0, 3, 2, 1])
pd.testing.assert_series_equal(expected, got)
def test_from_mapping_in_csv(self, tmp_path):
file_content = """
a,0
d, 2
c,5
"""
txt_path = tmp_path / "bar.txt"
write_str_to_file(file_content, txt_path)
got = data_processing.CategoryEncoder.from_mapping_in_csv(txt_path).get_encoded(
self.series
)
expected = pd.Series([0, 6, 5, 2])
pd.testing.assert_series_equal(expected, got)
def setup_class(self, tmp_path):
pb_str = """
config_name: "foo"
data_loader:
cls_name: "bar"
feature_config_path: "rab"
features_to_model: ["c"]
label_col: "d"
train_filters: []
validation_filters: []
model_wrapper:
cls_name: "bobar"
model_analysis:
metrics: ["a"]
by_features: ["b"]
"""
self.pipeline_config_path = tmp_path / "pipeline_config.yaml"
write_str_to_file(pb_str, self.pipeline_config_path)
def setup_class(cls, tmp_path):
yaml_str = """
raw_data_dir: dummy
dataset_name: dummy
base_features:
- name: a
dtype: INT32
- name: d1
dtype: DATETIME
transforming_features:
- name: b
index: 1
dependencies:
- a
dtype: INT32
- name: c
index: 2
dependencies:
- b
dtype: INT32
- name: d
index: 3
dependencies:
- a
dtype: INT32
- name: e
index: 4
dependencies:
- c
dtype: INT32
- name: d2
index: 5
dependencies:
- d1
dtype: DATETIME
"""
yaml_config_path = str(tmp_path / "tmp.yaml")
write_str_to_file(yaml_str, yaml_config_path)
cls.fm = _DummyFeatureManager(yaml_config_path)
cls.fm2 = _DummyFeatureManager2(yaml_config_path)
cls.fm.initialize_dataframe()
cls.fm2.initialize_dataframe()
def setup_class(self, tmp_path):
# feature manager config
dataset_dir = tmp_path / "dataset"
fm_pb_str = f"""
raw_data_dir: "{dataset_dir}"
dataset_name: "dummy"
base_features:
- name: "a"
dtype: INT32
transforming_features:
- name: "b"
index: 1
dtype: INT32
- name: "label"
index: 2
dtype: INT32
- name: "is_train"
index: 3
dtype: BOOL
- name: "is_validation"
index: 4
dtype: BOOL
"""
fm_pb_path = tmp_path / "feature_config.yaml"
write_str_to_file(fm_pb_str, fm_pb_path)
# create fake data
df = pd.DataFrame(
data={
"a": [1, 2, 3, 4, 5],
"b": [6, 7, 8, 9, 0],
"c": [-1, -1, -1, -1, -1],
"label": [0, 1, 1, 0, 1],
"is_train": [True, False, True, True, False],
"is_validation": [False, True, False, False, True],
})
dataset_path = BaseFeatureManager(fm_pb_path).get_dataset_path()
Path(dataset_path).parent.mkdir(parents=True)
df.to_csv(dataset_path, index=False)
# pipeline config
pipeline_config = f"""
config_name: "dummy"
data_loader:
cls_name: "tabml.data_loaders.BaseDataLoader"
feature_config_path: "{fm_pb_path}"
label_col: "label"
features_to_model: ["a", "b"]
train_filters: ["is_train"]
validation_filters: ["is_validation"]
model_wrapper:
cls_name: "a"
model_analysis:
metrics: ["foo"]
by_features: ["bar"]
by_label: "bar"
training_size: 50
"""
pipeline_config_path = tmp_path / "pipeline_config.yaml"
write_str_to_file(pipeline_config, pipeline_config_path)
self.config = parse_pipeline_config(pipeline_config_path)