def test_upload_existing_project_to_dbfs(dbfs_path_exists_mock): # pylint: disable=unused-argument
# Check that we don't upload the project if it already exists on DBFS
with mock.patch("mlflow.projects.databricks.DatabricksJobRunner._upload_to_dbfs")\
as upload_to_dbfs_mock:
dbfs_path_exists_mock.return_value = True
runner = DatabricksJobRunner(databricks_profile="DEFAULT")
runner._upload_project_to_dbfs(project_dir=TEST_PROJECT_DIR,
experiment_id=0)
assert upload_to_dbfs_mock.call_count == 0
def test_upload_project_to_dbfs(dbfs_root_mock, tmpdir, dbfs_path_exists_mock,
upload_to_dbfs_mock): # pylint: disable=unused-argument
# Upload project to a mock directory
dbfs_path_exists_mock.return_value = False
runner = DatabricksJobRunner(databricks_profile="DEFAULT")
dbfs_uri = runner._upload_project_to_dbfs(project_dir=TEST_PROJECT_DIR,
experiment_id=0)
# Get expected tar
local_tar_path = os.path.join(dbfs_root_mock, dbfs_uri.split("/dbfs/")[1])
expected_tar_path = str(tmpdir.join("expected.tar.gz"))
file_utils.make_tarfile(output_filename=expected_tar_path,
source_dir=TEST_PROJECT_DIR,
archive_name=databricks.DB_TARFILE_ARCHIVE_NAME)
# Extract the tarred project, verify its contents
assert filecmp.cmp(local_tar_path, expected_tar_path, shallow=False)
