def test_run_data_processing_pipeline_happy_path(
mocker: MockFixture, aml_pipeline_mocks): # NOQA: F811, E501
""" Test Processing Pipeline build
happy Path
"""
# Load mocks from fixture
(workspace, aml_compute, mock_workspace_get, mock_pipeline_publish) =\
aml_pipeline_mocks
mock_args = mocker.patch('argparse.ArgumentParser.parse_args')
mock_args.return_value = edict({
"aml_pipeline_name": "aml_pipeline_name_test", # NOQA: E501
"output_pipeline_id_file": "w",
"skip_preprocessing_execution": False
})
mock_aml_pipelines_list = mocker.patch(
'azureml.pipeline.core.PublishedPipeline.list') # NOQA: E501
type(mock_pipeline_publish).name = mocker.PropertyMock(
return_value='aml_pipeline_name_test') # NOQA: E501
type(mock_pipeline_publish).id = mocker.PropertyMock(return_value='1')
mock_aml_pipelines_list.return_value = [mock_pipeline_publish]
mocker.patch('azureml.core.Experiment')
# run as module (start from if __name__ == "__main__")
runpy.run_module('ml_service.pipelines.'
'run_data_processing_pipeline',
run_name='__main__')
def test_find_pipeline_by_name(mocker: MockFixture, workspace):
mock_aml_pipelines = mocker.patch(
'azureml.pipeline.core.PublishedPipeline.list')
mock_aml_pipeline = mocker.patch('azureml.pipeline.core.PublishedPipeline')
name = mocker.PropertyMock(return_value="mock_pipeline_name")
version = mocker.PropertyMock(return_value="2020-01-12T14:12:06.000")
type(mock_aml_pipeline).name = name
type(mock_aml_pipeline).version = version
mock_aml_pipelines.return_value = {mock_aml_pipeline}
matched_pipelines = find_pipeline_by_name(
aml_workspace=workspace, pipeline_name="mock_pipeline_name")
assert matched_pipelines
async def test_ok_wo_sql(self, model_manager: BaseModelManager, mocker: MockFixture):
fake_fetch = mocker.Mock()
mocked_get_sql = mocker.patch.object(model_manager, 'get_sql')
mocked_connection = CoroutineMock()
mocker.patch.object(model_manager, 'sql')
mocker.patch.object(
BaseModelManager,
'engine',
mocker.PropertyMock(
return_value=mocker.Mock(
acquire=async_context_mock(return_value=mocked_connection)
)
)
)
mocked_run_query_with_connection = mocker.patch.object(
model_manager, 'run_query_with_connection', CoroutineMock()
)
compared_return = await model_manager.run_query(fetch=fake_fetch)
mocked_get_sql.assert_called_once_with()
assert compared_return == mocked_run_query_with_connection.return_value
mocked_run_query_with_connection.assert_called_once_with(
mocked_connection, mocked_get_sql.return_value, fake_fetch
)
def test_get_aml_compute_already_exists(mocker: MockFixture, workspace):
expected_compute = mocker.patch('azureml.core.compute.ComputeTarget',
spec=True)
compute_targets = mocker.PropertyMock(
return_value={"compute_target": expected_compute})
type(workspace).compute_targets = compute_targets
compute = get_compute(workspace=workspace, compute_name="compute_target")
assert compute == expected_compute
async def test_acquire_ok(self, model_manager: BaseModelManager, mocker: MockFixture):
fake_sql = mocker.Mock()
fake_fetch = mocker.Mock()
mocked_run_query_with_connection = mocker.patch.object(
model_manager, 'run_query_with_connection', CoroutineMock()
)
mocked_connection = CoroutineMock()
mocker.patch.object(
BaseModelManager,
'engine',
mocker.PropertyMock(
return_value=mocker.Mock(
acquire=async_context_mock(return_value=mocked_connection)
)
)
)
await model_manager.run_query(fake_sql, fake_fetch)
mocked_run_query_with_connection.assert_called_once_with(mocked_connection, fake_sql, fake_fetch)
def test_build_data_processing_os_cmd_pipeline_default_datastore(
mocker: MockFixture,
environment_vars, # NOQA: F811
aml_pipeline_mocks): # NOQA: F811
""" Test Processing Pipeline build
test branch get default datastore
"""
# Load mocks and data objects from fixture
(workspace, aml_compute, mock_workspace_get, mock_pipeline_publish) =\
aml_pipeline_mocks
# Reset datastore_name from env_variables
type(environment_vars).datastore_name =\
mocker.PropertyMock(return_value=None)
# From https://realpython.com/python-data-structures/#collectionsnamedtuple-convenient-data-objects # NOQA: E501
def_datastore_name = 'default_datastore'
Def_datastore = namedtuple("Def_datastore", "name")
ddi = Def_datastore(def_datastore_name)
# Mock return value of aml_workspace.get_default_datastore().name
mocker.patch('azureml.core.Workspace.get_default_datastore',
return_value=ddi)
# Create Spies
spy_pythonscriptstep_create =\
mocker.patch('azureml.pipeline.steps.PythonScriptStep',
wraps=PythonScriptStep)
spy_pipeline_create = mocker.patch('azureml.pipeline.core.Pipeline',
wraps=Pipeline)
# run as module (start from if __name__ == "__main__")
runpy.run_module(
'ml_service.pipelines.'
'build_data_processing_os_cmd_pipeline',
run_name='__main__')
# Assertions
# Load Mocked environment variables
e = Env()
# Check if the correct workspace retrieved
mock_workspace_get.assert_called_with(name=e.workspace_name,
resource_group=e.resource_group,
subscription_id=e.subscription_id)
# Check if PythonScriptStep instantiation was called correctly
spy_pythonscriptstep_create.\
assert_called_once_with(allow_reuse=False,
runconfig=ANY,
arguments=ANY,
source_directory=e.sources_directory_train,
script_name="preprocess/"
"preprocess_os_cmd_aml.py",
name="Preprocess Data with OS cmd",
compute_target=ANY)
assert spy_pythonscriptstep_create.call_args[1]['arguments'][0] ==\
'--dataset_name'
assert spy_pythonscriptstep_create.call_args[1]['arguments'][1] ==\
e.dataset_name
assert spy_pythonscriptstep_create.call_args[1]['arguments'][2] ==\
'--datastore_name'
assert spy_pythonscriptstep_create.call_args[1]['arguments'][3] ==\
def_datastore_name
# Check if Pipeline instantiation was called correctly
spy_pipeline_create.assert_called_once_with(workspace=workspace, steps=ANY)
assert spy_pipeline_create.call_args[1]['steps'][0]._runconfig ==\
spy_pythonscriptstep_create.call_args[1]['runconfig']
assert spy_pipeline_create.call_args[1]['steps'][0]._compute_target ==\
spy_pythonscriptstep_create.call_args[1]['compute_target']
# Check if Pipeline publish was called with arguments
mock_pipeline_publish.\
assert_called_once_with(name=e.preprocessing_pipeline_name,
description="Data preprocessing"
" OS cmd pipeline",
version=e.build_id)
def environment_vars(mocker: MockFixture):
""" environment_vars
Auto fixture - will be always present if imported by unit tests
Sets basic env variables needed.
Prevents developer's .env file from being used
It is possible to overwrite their values within a test case with
type(environment_vars).<env_variable_name> =\
mocker.PropertyMock(return_value="<env_variable_value>")
"""
# In search for the instance attribute mock
environment_vars = MagicMock(spec=Env)
type(environment_vars).sources_directory_train =\
mocker.PropertyMock(return_value="ml_model")
type(environment_vars).aml_env_name =\
mocker.PropertyMock(return_value="utest_env_name")
type(environment_vars).aml_env_train_conda_dep_file =\
mocker.PropertyMock(return_value="utest_conda_dependencies.yml")
type(environment_vars).dataset_name =\
mocker.PropertyMock(return_value="utest_dataset")
type(environment_vars).subscription_id =\
mocker.PropertyMock(return_value="utest_subid")
type(environment_vars).resource_group =\
mocker.PropertyMock(return_value="utest_rg")
type(environment_vars).workspace_name =\
mocker.PropertyMock(return_value="utest_workspace")
type(environment_vars).compute_name =\
mocker.PropertyMock(return_value="utest_computename")
type(environment_vars).vm_size =\
mocker.PropertyMock(return_value="utest_vm_size")
type(environment_vars).rebuild_env =\
mocker.PropertyMock(return_value="False")
type(environment_vars).datastore_name =\
mocker.PropertyMock(return_value="utest_datastore")
type(environment_vars).processed_dataset_name =\
mocker.PropertyMock(return_value="utest_processed_dataset")
type(environment_vars).app_insights_connection_string =\
mocker.PropertyMock(return_value="utest_connectionstring")
type(environment_vars).log_level =\
mocker.PropertyMock(return_value="DEBUG")
type(environment_vars).log_sampling_rate =\
mocker.PropertyMock(return_value="1.0")
type(environment_vars).trace_sampling_rate =\
mocker.PropertyMock(return_value="1.0")
type(environment_vars).metrics_export_interval =\
mocker.PropertyMock(return_value="15")
# Returns mocked Env() instance every time it is instanciated
mocker.patch('ml_service.util.env_variables.Env.__new__',
return_value=environment_vars)
return environment_vars
def test_get_aml_compute_throws_exception(mocker: MockFixture, workspace):
compute_targets = mocker.PropertyMock(
side_effect=ComputeTargetException("Error while creating compute!"))
type(workspace).compute_targets = compute_targets
with pytest.raises(SystemExit):
get_compute(workspace=workspace, compute_name="compute_target")
def test_mocker_has_property_mock_class_as_attribute_for_instantiation():
from pytest_mock import mock_module, MockFixture
mocker = MockFixture()
assert isinstance(mocker.PropertyMock(), mock_module.PropertyMock)
