The lo•boto•my library allows one to mock the low-level boto3 client libraries efficiently, especially in more complex scenario testing situations, using configuration-based response definitions. The benefit is a separation of the configuration from the test execution, which cleans up the test invocation process.
lobotomy is available on pypi and installable via pip:
$ pip install lobotomyor via poetry as a development dependency in a project:
$ poetry add lobotomy -DCreate a configuration file in YAML, TOML, or JSON format with a root clients
key. Beneath that key add the desired client calls mocked responses by service
as shown here for a session.client('s3').get_object() mocked response:
clients:
s3:
get_object:
Body: 'The contents of my S3 file.'
LastModified: '2020-12-01T01:02:03Z'Then in the test notice that the lobotomy patching process handles the rest, including casting the configuration values specified in the call above into the more complex data types returned for the specific call.
import lobotomy
import pathlib
import boto3
import datetime
my_directory = pathlib.Path(__file__).parent
@lobotomy.patch(my_directory.joinpath("test_lobotomy.yaml"))
def test_lobotomy(lobotomized: lobotomy.Lobotomy):
"""
Should return the mocked get_object response generated from the
configuration data specified in the lobotomy patch above. By default
the patch(...) applies to the boto3.Session object, so calling
boto3.Session() will create a lobotomy.Session instead of a normal
boto3.Session. From there the low-level client interface is designed
to match normal usage.
"""
s3_client = boto3.Session().client("s3")
# Lobotomy will validate that you have specified the required keys
# in your request, so Bucket and Key have to be supplied here even though
# they are not meaningful values in this particular test scenario.
response = s3_client.get_object(Bucket="foo", Key="bar")
expected = b"The contents of my S3 file."
assert response["Body"].read() == expected, """
Expected the mocked response body data to be returned as a
StreamingBody object with blob/bytes contents. The lobotomy
library introspects boto to properly convert the string body
value in the configuration file into the expected return format
for the particular call.
"""
expected = datetime.datetime(2020, 12, 1, 1, 2, 3, 0, datetime.timezone.utc)
assert response["LastModified"] == expected, """
Expected the mocked response last modified value to be a timezone-aware
datetime value generated from the string timestamp value in the
configuration file to match how it would be returned by boto in
an actual response.
"""
call = lobotomized.get_service_call("s3", "get_object")
assert call.request["Bucket"] == "foo", """
Expected the s3.get_object method call arguments to have specified the bucket
as "foo".
"""
assert call.request["Key"] == "bar", """
Expected the s3.get_object method call argumets to have specified the key as
"bar".
"""Test scenarios can be written in YAML, TOML, or JSON formats. YAML or TOML are recommended unless copying output responses from JSON calls is easier for a given use-case. In the following example, we'll define the calls using YAML:
clients:
sts:
get_caller_identity:
Account: '987654321'
s3:
get_object:
Body: 'The contents of my S3 file.'
LastModified: '2020-11-01T12:23:34Z'All call responses are stored within the root clients attribute with service names as sub-attributes and then the service method call responses defined beneath the service name attributes. Multiple services and multiple methods per service are specified in this way by the hierarchical key lists.
If the contents of a method response definition are not a list, the same response will be returned for each call to that client method during the test. Specifying a list of responses will alter the behavior such that each successive call will iterate through that list of configured responses.
clients:
s3:
get_object:
- Body: 'The contents of my S3 file.'
LastModified: '2020-11-01T12:23:34Z'
- Body: 'Another S3 file.'
LastModified: '2020-11-03T12:23:34Z'The lobotomy library dynamically inspects boto for its response structure
and data types, such that they match the types for the normal boto response
for each client method call. In the case of the
session.client('s3').get_object() calls above, the Body would return
a StreamingBody object with the string converted to bytes to match the
normal output. Similarly, the LastModified would be converted to a
timezone-aware datetime object by lobotomy as well. This makes it easy
to specify primitive data types in the configuration file that are transformed
into their more complex counterparts when returned within the execution of
the actual code.
Once the configuration file has been specified, it is used within a test
via a lobotomy.patch as shown below.
import lobotomy
import pathlib
import boto3
import datetime
my_directory = pathlib.Path(__file__).parent
@lobotomy.patch(my_directory.joinpath("test_lobotomy.yaml"))
def test_lobotomy(lobotomized: lobotomy.Lobotomy):
"""
Should return the mocked get_object response generated from the
configuration data specified in the lobotomy patch above. By default
the patch(...) applies to the boto3.Session object, so calling
boto3.Session() will create a lobotomy.Session instead of a normal
boto3.Session. From there the low-level client interface is designed
to match normal usage.
"""
s3_client = boto3.Session().client("s3")
# Lobotomy will validate that you have specified the required keys
# in your request, so Bucket and Key have to be supplied here even though
# they are not meaningful values in this particular test scenario.
response = s3_client.get_object(Bucket="foo", Key="bar")
expected = b"The contents of my S3 file."
assert response['Body'].read() == expected, """
Expect the mocked response body data to be returned as a
StreamingBody object with blob/bytes contents. The lobotomy
library introspects boto to properly convert the string body
value in the configuration file into the expected return format
for the particular call.
"""
expected = datetime.datetime(2020, 12, 1, 1, 2, 3, 0, datetime.timezone.utc)
assert response['LastModified'] == expected, """
Expect the mocked response last modified value to be a timezeon-aware
datetime value generated from the string timestamp value in the
configuration file to match how it would be returned by boto in
an actual response.
"""The patching process replaces the boto3.Session class with a
lobotomy.Lobotomy object that contains the loaded configuration data.
When patched in this fashion, boto3.Session() calls will actually be
lobotomy.Lobotomy() calls that return lobotomy.Session objects. These
sessions have the interface of the boto3.Session object, but behave in
a way such that client responses are returned from the configuration data
instead of through interactivity with AWS.
For simple cases with little configuration, it is also possible to patch data stored directly within the Python code. The above test could be rewritten in this way as:
import lobotomy
configuration = {
"clients": {
"s3": {
"get_object": [
{
"Body": "The contents of my S3 file.",
"LastModified": "2020-11-01T12:23:34Z",
},
{
"Body": "Another S3 file.",
"LastModified": "2020-11-03T12:23:34Z",
},
],
},
},
}
@lobotomy.patch(data=configuration)
def test_lobotomy(lobotomized: lobotomy.Lobotomy):
"""..."""Although one of the benefits of lobotomy is the ability to streamline the tests files by reducing the response configuration, which can be a bit verbose inside Python files and that is the highly recommended approach.
A third option for simpler cases is to use the Lobotomy.add_call() method
to register calls and responses.
import lobotomy
@lobotomy.patch()
def test_lobotomy(lobotomized: lobotomy.Lobotomy):
"""..."""
lobotomized.add_call(
service_name="s3",
method_name="get_object",
response={
"Body": "The contents of my S3 file.",
"LastModified": "2020-11-01T12:23:34Z",
},
)
lobotomized.add_call(
service_name="s3",
method_name="get_object",
response={
"Body": "Another S3 file.",
"LastModified": "2020-11-03T12:23:34Z",
},
)In the case above no data or path was supplied to the lobotomy.patch() and instead
call responses were registered within the test function itself. The response argument
in the Lobotomy.add_call() method is optional. If omitted, a default one will be
created instead in the same fashion as one would be created via the CLI (see below).
Note that it is also possible to mix loading data and adding calls within the test function.
When using YAML files, lobotomy comes with custom YAML classes that can provide even more flexibility and ease in defining data. The following are the available modifiers and how to use them:
This modifier will convert YAML data into a JSON string in the creation of the response, which makes it easier to represent complex JSON data in the scenario data.
clients:
secretsmanager:
get_secret_value:
SecretString: !lobotomy.to_json
first: first_value
second: second_valueIn this example the !lobotomy.to_json YAML modifier instructs the lobotomy to
converts the object data beneath the SecretString attribute into a JSON string
as part of the response object. In this case then in the associated Python test:
import boto3
import lobotomy
import json
@lobotomy.patch(path="scenario.yaml")
def test_showing_to_json(lobotomized: lobotomy.Lobotomy):
"""Should expect a JSON string for the 'SecretString' value."""
client = boto3.Session().client("secretsmanager")
response = client.get_secret_value(SecretId="fake")
expected = {"first": "first_value", "second": "second_value"}
assert expected == json.loads(response["SecretValue"])the value is returned as the expected JSON string.
This modifier is used to inject the string contents of another file into the value of the associated attribute.
clients:
s3:
get_object:
Body: !lobotomy.inject_string './body.txt'Here the !lobotomy.inject_string YAML modifier instructs lobotomy to load the
contents of the external file ./body.txt into the Body attribute value where
the external file path is defined relative to the defining YAML file.
So in this case, if there's a body.txt file with the contents Hello lobotomy!,
the Python test would find this in reading the body:
import boto3
import lobotomy
import json
@lobotomy.patch(path="scenario.yaml")
def test_showing_inject_string(lobotomized: lobotomy.Lobotomy):
"""Should expect the body.txt to be injected into the Body response attribute."""
client = boto3.Session().client("s3")
response = client.get_object(Bucket="fake", Key="fake")
assert response["Body"].read() == b"Hello lobotomy!"The lobotomy library also has a command line interface to help streamline
the process of creating configuration files. The CLI has an add command
that can be used to auto-generate method call response configurations to
a new or existing configuration file. The values are meant to be replaced
and unused keys to be removed to streamline for testing, but it helps a lot
to get the full structure of the response in place and work from there instead
of having to look it up yourself.
For example, creating the file:
clients:
sts:
get_caller_identity:
Account: '987654321'
s3:
get_object:
- Body: 'The contents of my S3 file.'
LastModified: '2020-11-01T12:23:34Z'
- Body: 'Another S3 file.'
LastModified: '2020-11-03T12:23:34Z'could be done first through the CLI commands:
$ lobotomy add sts.get_caller_identity example.yamlAfter that command is executed, the example.yaml file will be
created and populated initially with:
clients:
sts:
get_caller_identity:
Account: '...'
Arn: '...'
UserId: '...'Notice the values are placeholders. We can adjust the values to what we want and remove the unnecessary keys for our particular case such that the file contents are then:
clients:
sts:
get_caller_identity:
Account: '987654321'Next add the first s3.get_object call:
$ lobotomy add s3.get_object example.yamlThe configuration file now looks like:
clients:
s3:
get_object:
AcceptRanges: '...'
Body: '...'
CacheControl: '...'
ContentDisposition: '...'
ContentEncoding: '...'
ContentLanguage: '...'
ContentLength: 1
ContentRange: '...'
ContentType: '...'
DeleteMarker: null
ETag: '...'
Expiration: '...'
Expires: '2020-11-04T14:37:18.042821Z'
LastModified: '2020-11-04T14:37:18.042821Z'
Metadata: null
MissingMeta: 1
ObjectLockLegalHoldStatus: '...'
ObjectLockMode: '...'
ObjectLockRetainUntilDate: '2020-11-04T14:37:18.042821Z'
PartsCount: 1
ReplicationStatus: '...'
RequestCharged: '...'
Restore: '...'
SSECustomerAlgorithm: '...'
SSECustomerKeyMD5: '...'
SSEKMSKeyId: '...'
ServerSideEncryption: '...'
StorageClass: '...'
TagCount: 1
VersionId: '...'
WebsiteRedirectLocation: '...'
sts:
get_caller_identity:
Account: '987654321'Of course, this is a simple case because we don't need much of the response structure in our simplified use-case, but hopefully you can see the value of being able to add the response structure so easily for more complex cases. Once again, the new call is adjusted to fit our particular needs:
clients:
s3:
get_object:
Body: 'The contents of my S3 file.'
LastModified: '2020-11-01T12:23:34Z'
sts:
get_caller_identity:
Account: '987654321'Adding the second s3.get_object call is identical:
$ lobotomy add s3.get_object example.yamlHowever, lobotomy notices the existing call there and so converts the
get_object response configuration to a list of responses for you:
clients:
s3:
Body: 'The contents of my S3 file.'
LastModified: '2020-11-01T12:23:34Z'
get_object:
AcceptRanges: '...'
Body: '...'
CacheControl: '...'
ContentDisposition: '...'
ContentEncoding: '...'
ContentLanguage: '...'
ContentLength: 1
ContentRange: '...'
ContentType: '...'
DeleteMarker: null
ETag: '...'
Expiration: '...'
Expires: '2020-11-04T14:42:51.077364Z'
LastModified: '2020-11-04T14:42:51.077364Z'
Metadata: null
MissingMeta: 1
ObjectLockLegalHoldStatus: '...'
ObjectLockMode: '...'
ObjectLockRetainUntilDate: '2020-11-04T14:42:51.077364Z'
PartsCount: 1
ReplicationStatus: '...'
RequestCharged: '...'
Restore: '...'
SSECustomerAlgorithm: '...'
SSECustomerKeyMD5: '...'
SSEKMSKeyId: '...'
ServerSideEncryption: '...'
StorageClass: '...'
TagCount: 1
VersionId: '...'
WebsiteRedirectLocation: '...'
sts:
get_caller_identity:
Account: '987654321'Finally, edit the new call response configuration and we end up with the configuration we were looking for:
clients:
sts:
get_caller_identity:
Account: '987654321'
s3:
get_object:
- Body: 'The contents of my S3 file.'
LastModified: '2020-11-01T12:23:34Z'
- Body: 'Another S3 file.'
LastModified: '2020-11-03T12:23:34Z'By default configuration files are rooted at the clients key within the
file. However, it is possible to specify a different root key prefix, which
is useful when co-locating lobotomy test configuration with other test
configuration data in the same file, or when co-locating multiple lobotomy
test configurations within the same file. To achieve that a prefix must be
specified during patching.
As an example, consider the configuration file:
lobotomy:
test_a:
clients:
sts:
get_caller_identity:
Account: '987654321'
test_b:
clients:
sts:
get_caller_identity:
Account: '123456678'
UserId: 'AIFASDJWISJAVHXME'In this case the prefixes are lobotomy.test_a and lobotomy.test_b.
To use these in a test the prefix must be specified in the patch:
import pathlib
import lobotomy
config_path = pathlib.Path(__file__).parent.joinpath("validation.yaml")
@lobotomy.patch(config_path, prefix="lobotomy.test_a")
def test_a(lobotomized: lobotomy.Lobotomy):
"""..."""
@lobotomy.patch(config_path, prefix="lobotomy.test_b")
def test_b(lobotomized: lobotomy.Lobotomy):
"""..."""The prefix can be specified as a . delimited string, or as a list/tuple.
The list/tuple is needed if the keys themselves contains ..
By default, lobotomy will patch boto3.Session. There are scenarios where
a different patch would be desired either to limit the scope of the patch or
to patch another library wrapping the boto3.Session call. In those cases,
specify the patch path argument:
import pathlib
import lobotomy
config_path = pathlib.Path(__file__).parent.joinpath("test.yaml")
@lobotomy.patch(config_path, patch_path="something.else.Session")
def test_another_patch_path(lobotomized: lobotomy.Lobotomy):
"""..."""In addition to the clients configurations described above, it is also possible
to configure the session values as well with the session: attribute. The
available configuration settings for the session are:
session:
profile_name: some-profile
region_name: us-west-2
available_profiles:
- some-profile
- some-other-profile
credentials:
access_key: A123KEY
secret_key: somesecretkeyvalue
token: theaccesstokenifset
method: how-the-credentials-were-loaded
clients:
...All of these values are optional and in cases where they are omitted, but the session requires having them, they will be defaulted.
There are cases where it is desirable to replace one or more service clients with non-lobotomy objects. Lobotomy supports that by adding client overrides to the patched lobotomy object:
from unittest.mock import MagicMock
import lobotomy
@lobotomy.patch()
def test_example(lobotomized: lobotomy.Lobotomy):
"""Should do something..."""
mock_dynamo_db_client = MagicMock()
lobotomized.add_client_override("dynamodb", mock_dynamo_db_client)
# continue testing...The lobotomy library mimics client error handling with a lobotomized version of the same interface used by the live clients. As such, handling and capturing errors works transparently. For example,
import boto3
import pytest
import lobotomy
@lobotomy.patch()
def test_client_errors(lobotomized: lobotomy.Lobotomy):
"""Should raise the specified error."""
lobotomized.add_error_call(
service_name="s3",
method_name="list_objects",
error_code="NoSuchBucket",
error_message="Hello...",
)
session = boto3.Session()
client = session.client("s3")
with pytest.raises(client.exceptions.NoSuchBucket):
client.list_objects(Bucket="foo")Or the generic client error handling with response codes can be used as well:
import boto3
import pytest
import lobotomy
@lobotomy.patch()
def test_client_errors(lobotomized: lobotomy.Lobotomy):
"""Should raise the specified error."""
lobotomized.add_call(
service_name="s3",
method_name="list_objects",
error_code="NoSuchBucket",
error_message="Hello...",
)
session = boto3.Session()
client = session.client("s3")
with pytest.raises(lobotomy.ClientError) as exception_info:
client.list_objects(Bucket="foo")
assert exception_info.value.response["Error"]["Code"] == "NoSuchBucket"These errors can also be specified in YAML files like this:
clients:
s3:
list_objects: !lobotomy.error
code: NoSuchBucket
message: Hello...The !lobotomy.error will load this as an error response with the same behavior as the
examples shown above.
For more advanced cases it is also possible to use any callable as the source for generating the response. For example, if I want the response to be different based on the arguments specified when making the call, I can create a function to return a result that reflects the inputs.
import typing
import boto3
import lobotomy
def _respond(*args, **kwargs) -> typing.Dict[str, typing.Any]:
return {
"Body": "{}.{}".format(kwargs["Bucket"], kwargs["Key"]).encode()
}
@lobotomy.patch()
def test_callable_responses(lobotomized: "lobotomy.Lobotomy"):
"""Should return the expected body value from the callable response."""
lobotomized.add_call("s3", "get_object", _respond)
session = boto3.Session()
client = session.client("s3")
response = client.get_object(Bucket="foo", Key="bar")
assert response["Body"].read() == b"foo.bar"Any callable is supported as long as it accepts the *args, **kwargs appropriate to
the calling client service request.