ashb commented on a change in pull request #4091: [AIRFLOW-2524] Update
SageMaker hook, operator and sensor for training, tuning and transform
URL: https://github.com/apache/incubator-airflow/pull/4091#discussion_r228727567
##########
File path: tests/contrib/hooks/test_sagemaker_hook.py
##########
@@ -233,94 +254,82 @@ class TestSageMakerHook(unittest.TestCase):
def setUp(self):
configuration.load_test_config()
- db.merge_conn(
- models.Connection(
- conn_id='sagemaker_test_conn_id',
- conn_type='sagemaker',
- login='access_id',
- password='access_key',
- extra=json.dumps(db_config)
- )
- )
+
+ @mock.patch.object(SageMakerHook, 'get_iam_conn')
Review comment:
We have had a problem with overly specific mocks in the past and prefer
using moto where possible for mocking AWS api calls.
For example this would become:
https://github.com/spulec/moto/blob/master/tests/test_iam/test_iam.py#L153-L167
```
@mock_iam
def test_expand_role(self):
conn = boto3.client('iam', region_name='us-east-1')
conn.create_role(RoleName="test-role",
AssumeRolePolicyDocument="some policy")
hook = SageMakerHook(aws_conn_id='sagemaker_test_conn_id')
arn = hook.expand_role(some_role)
self.assertEqual("arn:aws:iam::123456789012:role/test-role")
```
This way it still goes through all the Boto3 client-side validation etc,
which has caught some bugs for us in the past.
We should use this similarly for S3 operations, but I don't think moto
supports SageMaker yet, so we'd still have to mock those API calls.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
