def test_create_ml_object(self):
ml_object = MLObject()
ml_object.set_type("0.0.1", MLSchemaTypes.DATAPATH)
self.assertIsNotNone(ml_object)
self.assertTrue(ml_object.schema_version == "0.0.1")
self.assertTrue(ml_object.schema_type == MLSchemaTypes.DATAPATH.name.lower())
def test_create_stub_nested_object(self):
ml_object = MLObject()
ml_object.set_type('0.0.1', MLSchemaTypes.DATAPATH)
self.assertTrue(len(ml_object) == 13)
self.assertIsNone(ml_object.run_date)
self.assertTrue(len(ml_object.connection) == 3)
self.assertIsNone(ml_object.connection.endpoint)
def test_live_interface_samples(self):
MLSchema.populate_registry()
print("Testing Keras")
loaded_object, errors = MLObject.create_object_from_string(
SampleSubmissions.FULL_SUBMISSIONS.COMPONENT_KERAS)
pprint(errors)
self.assertTrue(len(errors) == 0)
self.assertIsNotNone(loaded_object.get_schema())
print("Testing IBM")
loaded_object, errors = MLObject.create_object_from_string(
SampleSubmissions.FULL_SUBMISSIONS.COMPONENT_IBM)
pprint(errors)
self.assertTrue(len(errors) == 0)
self.assertIsNotNone(loaded_object.get_schema())
print("Testing OpenVino")
loaded_object, errors = MLObject.create_object_from_string(
SampleSubmissions.FULL_SUBMISSIONS.COMPONENT_OPENVINO)
pprint(errors)
self.assertTrue(len(errors) == 0)
self.assertIsNotNone(loaded_object.get_schema())
def test_create_bad_semver(self):
with self.assertRaises(ValueError):
ml_object = MLObject()
ml_object.set_type(schema_version="0.0.x", schema_type=MLSchemaTypes.BASE)
with self.assertRaises(ValueError):
ml_object = MLObject()
ml_object.set_type(schema_version=None, schema_type=MLSchemaTypes.BASE)
def test_load_object_from_disk(self):
MLSchema.populate_registry()
file_path = Path('tests/data/0/0/1/datapath.yaml')
ml_object, _ = MLObject.create_object_from_file(file_path)
self.assertIsNotNone(ml_object.run_date)
self.assertIsNotNone(ml_object.connection.endpoint)
def test_all_data(self):
MLSchema.populate_registry()
all_data_files = list(Path('tests').glob('data/0/0/1/*.yaml'))
self.assertTrue(len(all_data_files) > 1)
for data_file in all_data_files:
print(data_file)
loaded_object, errors = MLObject.create_object_from_file(data_file)
self.assertTrue(len(errors) == 0)
self.assertIsNotNone(loaded_object.get_schema())
def test_code_gen_string(self):
code_gen_string = MLObject._code_gen_string(
"0.0.1", MLSchemaTypes.RUNCONFIG, "prefix"
)
fields_re = re.compile(r"^prefix.", flags=re.MULTILINE | re.DOTALL)
self.assertTrue(len(fields_re.findall(code_gen_string)) == 54)
hints_re = re.compile(r"^# prefix.", flags=re.MULTILINE | re.DOTALL)
self.assertTrue(len(fields_re.findall(code_gen_string)) == 54)
code_gen_string = MLObject._code_gen_string(
"0.0.1", MLSchemaTypes.RUNCONFIG, "prefix", type_hints=False
)
fields_re = re.compile(r"^prefix.", flags=re.MULTILINE | re.DOTALL)
self.assertTrue(len(fields_re.findall(code_gen_string)) == 54)
hints_re = re.compile(r"^# prefix.", flags=re.MULTILINE | re.DOTALL)
self.assertTrue(len(hints_re.findall(code_gen_string)) == 0)
def test_load_live_data(self):
MLSchema.populate_registry()
all_data_files = list(Path("tests").glob("data/**/*.yaml"))
for data_file in all_data_files:
this_text = data_file.read_text(encoding="utf-8")
this_dict = convert_yaml_to_dict(this_text)
if this_dict["schema_version"] == "0.0.1":
continue
# print(data_file)
loaded_object, errors = MLObject.create_object_from_file(data_file)
self.assertTrue(len(errors) == 0)
self.assertIsNotNone(loaded_object.get_schema())
def test_load_live_data(self):
MLSchema.populate_registry()
all_data_files = list(Path('tests').glob('data/**/*.yaml'))
for data_file in all_data_files:
this_text = data_file.read_text()
this_dict = convert_yaml_to_dict(this_text)
if this_dict['schema_version'] == '0.0.1':
continue
print(data_file)
loaded_object, errors = MLObject.create_object_from_file(data_file)
self.assertTrue(len(errors) == 0)
self.assertIsNotNone(loaded_object.get_schema())
def test_load_object_with_missing_field_from_variable(self):
WRONG_DATAPATH = """
schema_version: 0.0.1
schema_type: datapath
run_id: 6a9a5931-1c1d-47cc-aaf3-ad8b03f70575
step_id: 0c98f080-4760-46be-b35f-7dbb5e2a88c2
run_date: 1970-01-01 00:00:00.00000
# data_store: I_am_a_datastore_name
storage_connection_type: AWS_BLOB
connection:
# endpoint: S3://mybucket/puppy.jpg
access_key_id: AKIAIOSFODNN7EXAMPLE
secret_access_key: wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY"""
ml_object, errors = MLObject.create_object_from_string(WRONG_DATAPATH)
self.assertTrue(ml_object is None)
self.assertTrue(len(errors) == 2)
self.assertTrue(errors["data_store"][0] == "Field may not be null.")
self.assertTrue(errors["connection"]["endpoint"][0] == "Field may not be null.")
def attach_step_info(
self,
mlobject: MLObject,
workflow_version,
workflow_node_id,
step_name: str,
step_type: str,
):
if step_type not in ["input", "execution", "output", "log"]:
raise ValueError(
f"Error when saving '{mlobject.get_schema_name()}', the step_type must be from ['input', 'execution', 'output', 'log']."
)
run_info_id = build_vertex_id(
step_name,
step_type,
mlobject.run_id,
mlobject.run_date,
workflow_version,
self._workflow_partition_id,
)
mlobject_dict = mlobject.dict_without_internal_variables()
property_string = convert_to_property_strings(mlobject_dict)
raw_content = encode_raw_object_for_db(mlobject)
add_run_info_query = f"""g.addV('id', '{run_info_id}'){property_string}.property('raw_content', '{raw_content}').property('workflow_node_id', '{workflow_node_id}').property('workflow_partition_id', '{self._workflow_partition_id}')"""
self.execute_query(add_run_info_query)
self.execute_query(
sQuery(
"g.V('id', '%s').out().hasId('%s').addE('results').to(g.V('%s')).executionProfile()",
[workflow_node_id, step_name, run_info_id],
))
self.execute_query(
sQuery(
"g.V('id', '%s').out().hasId('%s').addE('root').from(g.V('%s')).executionProfile()",
[workflow_node_id, step_name, run_info_id],
))
return run_info_id
def test_create_stub_base_object(self):
ml_object = MLObject()
ml_object.set_type('0.0.1', MLSchemaTypes.BASE)
self.assertIsNone(ml_object['run_date'])
self.assertTrue(len(ml_object) == 10)
def test_create_bad_schema_type(self):
with self.assertRaises(KeyError):
ml_object = MLObject()
ml_object.set_type(schema_version='0.0.1', schema_type='foo')
def test_load_and_save_file(self):
run_id = uuid.uuid4()
save_path = Path(self.test_dir.name) / str(run_id)
save_path.mkdir()
datapath_object = MLObject()
datapath_object.set_type('0.0.1', MLSchemaTypes.DATAPATH)
datapath_object.run_id = run_id
datapath_object.step_id = uuid.uuid4()
datapath_object.run_date = datetime.datetime.now()
datapath_object.data_store = None # This is an intentional bug
# This is an intentional bug (Should be AWS_BLOB)
datapath_object.storage_connection_type = 'AWS_BLOB_OBJECT'
datapath_object.connection.endpoint = None # Another intentional bug
datapath_object.connection.access_key_id = 'AKIAIOSFODNN7EXAMPLE'
datapath_object.connection.secret_access_key = 'wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY'
response, errors = datapath_object.save(save_path)
self.assertFalse(response)
self.assertTrue(len(errors) == 3)
self.assertTrue(len(list(Path(save_path).glob('*'))) == 0)
datapath_object.storage_connection_type = 'AWS_BLOB'
response, errors = datapath_object.save(save_path)
self.assertFalse(response)
self.assertTrue(len(errors) == 2)
self.assertTrue(len(list(Path(save_path).glob('*'))) == 0)
datapath_object.connection.endpoint = 'http://s3.amazon.com/BUCKET'
response, errors = datapath_object.save(save_path)
self.assertFalse(response)
self.assertTrue(len(errors) == 1)
self.assertTrue(len(list(Path(save_path).glob('*'))) == 0)
datapath_object.data_store = 'BUCKET NAME'
response, errors = datapath_object.save(save_path)
self.assertTrue(response)
self.assertTrue(len(errors) == 0)
path = Path(save_path)
all_files = list(path.glob('*'))
self.assertTrue(len(all_files) == 1)
ml_object, errors = MLObject.create_object_from_file(all_files[0])
self.assertTrue(len(ml_object) == 13)
self.assertTrue(len(errors) == 0)
self.assertTrue(datapath_object.data_store == ml_object.data_store)
self.assertTrue(datapath_object.storage_connection_type ==
ml_object.storage_connection_type)
self.assertTrue(datapath_object.connection.endpoint ==
ml_object.connection.endpoint)
def test_create_bad_schema_type(self):
with self.assertRaises(RegistryError):
ml_object = MLObject()
ml_object.set_type(schema_version="0.0.1", schema_type="foo")
def test_cascading_inheritence(self):
MLSchema.populate_registry()
mlobject = MLObject()
mlobject.set_type("0.0.1", "data_version_control")
mlobject.run_id = uuid.uuid4()
mlobject.step_id = uuid.uuid4()
mlobject.run_date = datetime.datetime.now()
mlobject.data_store = "I_am_a_datastore"
mlobject.storage_connection_type = "AWS_BLOB"
mlobject.connection.endpoint = "con_endpoint"
mlobject.connection.access_key_id = "AKIAIOSFODNN7EXAMPLE"
mlobject.connection.secret_access_key = (
"wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY")
mlobject.dvc_hash = "923caceea54b38177505632f5612cc569a49b22246e346a7"
mlobject.validate()
