fix(idempotency): sorting keys before hashing

Author: heitorlessa

aws_lambda_powertools/utilities/idempotency/persistence/base.py

@@ -223,7 +223,7 @@ def _generate_hash(self, data: Any) -> str:
 
         """
         data = getattr(data, "raw_event", data)  # could be a data class depending on decorator order
-        hashed_data = self.hash_function(json.dumps(data, cls=Encoder).encode())
+        hashed_data = self.hash_function(json.dumps(data, cls=Encoder, sort_keys=True).encode())
         return hashed_data.hexdigest()
 
     def _validate_payload(self, data: Dict[str, Any], data_record: DataRecord) -> None:
@@ -310,7 +310,7 @@ def save_success(self, data: Dict[str, Any], result: dict) -> None:
         result: dict
             The response from function
         """
-        response_data = json.dumps(result, cls=Encoder)
+        response_data = json.dumps(result, cls=Encoder, sort_keys=True)
 
         data_record = DataRecord(
             idempotency_key=self._get_hashed_idempotency_key(data=data),

tests/functional/idempotency/conftest.py

@@ -21,6 +21,10 @@
 TABLE_NAME = "TEST_TABLE"
 
 
+def serialize(data):
+    return json.dumps(data, sort_keys=True, cls=Encoder)
+
+
 @pytest.fixture(scope="module")
 def config() -> Config:
     return Config(region_name="us-east-1")
@@ -62,12 +66,12 @@ def lambda_response():
 
 @pytest.fixture(scope="module")
 def serialized_lambda_response(lambda_response):
-    return json.dumps(lambda_response, cls=Encoder)
+    return serialize(lambda_response)
 
 
 @pytest.fixture(scope="module")
 def deserialized_lambda_response(lambda_response):
-    return json.loads(json.dumps(lambda_response, cls=Encoder))
+    return json.loads(serialize(lambda_response))
 
 
 @pytest.fixture
@@ -144,20 +148,20 @@ def expected_params_put_item_with_validation(hashed_idempotency_key, hashed_vali
 def hashed_idempotency_key(lambda_apigw_event, default_jmespath, lambda_context):
     compiled_jmespath = jmespath.compile(default_jmespath)
     data = compiled_jmespath.search(lambda_apigw_event)
-    return "test-func#" + hashlib.md5(json.dumps(data).encode()).hexdigest()
+    return "test-func#" + hashlib.md5(serialize(data).encode()).hexdigest()
 
 
 @pytest.fixture
 def hashed_idempotency_key_with_envelope(lambda_apigw_event):
     event = extract_data_from_envelope(
         data=lambda_apigw_event, envelope=envelopes.API_GATEWAY_HTTP, jmespath_options={}
     )
-    return "test-func#" + hashlib.md5(json.dumps(event).encode()).hexdigest()
+    return "test-func#" + hashlib.md5(serialize(event).encode()).hexdigest()
 
 
 @pytest.fixture
 def hashed_validation_key(lambda_apigw_event):
-    return hashlib.md5(json.dumps(lambda_apigw_event["requestContext"]).encode()).hexdigest()
+    return hashlib.md5(serialize(lambda_apigw_event["requestContext"]).encode()).hexdigest()
 
 
 @pytest.fixture

tests/functional/idempotency/test_idempotency.py

@@ -21,6 +21,7 @@
 from aws_lambda_powertools.utilities.idempotency.idempotency import idempotent, idempotent_function
 from aws_lambda_powertools.utilities.idempotency.persistence.base import BasePersistenceLayer, DataRecord
 from aws_lambda_powertools.utilities.validation import envelopes, validator
+from tests.functional.idempotency.conftest import serialize
 from tests.functional.utils import load_event
 
 TABLE_NAME = "TEST_TABLE"
@@ -741,7 +742,7 @@ def test_default_no_raise_on_missing_idempotency_key(
     hashed_key = persistence_store._get_hashed_idempotency_key({})
 
     # THEN return the hash of None
-    expected_value = "test-func#" + md5(json.dumps(None).encode()).hexdigest()
+    expected_value = "test-func#" + md5(serialize(None).encode()).hexdigest()
     assert expected_value == hashed_key
 
 
@@ -785,7 +786,7 @@ def test_jmespath_with_powertools_json(
     expected_value = [sub_attr_value, key_attr_value]
     api_gateway_proxy_event = {
         "requestContext": {"authorizer": {"claims": {"sub": sub_attr_value}}},
-        "body": json.dumps({"id": key_attr_value}),
+        "body": serialize({"id": key_attr_value}),
     }
 
     # WHEN calling _get_hashed_idempotency_key
@@ -869,7 +870,7 @@ def _delete_record(self, data_record: DataRecord) -> None:
 def test_idempotent_lambda_event_source(lambda_context):
     # Scenario to validate that we can use the event_source decorator before or after the idempotent decorator
     mock_event = load_event("apiGatewayProxyV2Event.json")
-    persistence_layer = MockPersistenceLayer("test-func#" + hashlib.md5(json.dumps(mock_event).encode()).hexdigest())
+    persistence_layer = MockPersistenceLayer("test-func#" + hashlib.md5(serialize(mock_event).encode()).hexdigest())
     expected_result = {"message": "Foo"}
 
     # GIVEN an event_source decorator
@@ -889,7 +890,7 @@ def lambda_handler(event, _):
 def test_idempotent_function():
     # Scenario to validate we can use idempotent_function with any function
     mock_event = {"data": "value"}
-    persistence_layer = MockPersistenceLayer("test-func#" + hashlib.md5(json.dumps(mock_event).encode()).hexdigest())
+    persistence_layer = MockPersistenceLayer("test-func#" + hashlib.md5(serialize(mock_event).encode()).hexdigest())
     expected_result = {"message": "Foo"}
 
     @idempotent_function(persistence_store=persistence_layer, data_keyword_argument="record")
@@ -906,7 +907,7 @@ def test_idempotent_function_arbitrary_args_kwargs():
     # Scenario to validate we can use idempotent_function with a function
     # with an arbitrary number of args and kwargs
     mock_event = {"data": "value"}
-    persistence_layer = MockPersistenceLayer("test-func#" + hashlib.md5(json.dumps(mock_event).encode()).hexdigest())
+    persistence_layer = MockPersistenceLayer("test-func#" + hashlib.md5(serialize(mock_event).encode()).hexdigest())
     expected_result = {"message": "Foo"}
 
     @idempotent_function(persistence_store=persistence_layer, data_keyword_argument="record")
@@ -921,7 +922,7 @@ def record_handler(arg_one, arg_two, record, is_record):
 
 def test_idempotent_function_invalid_data_kwarg():
     mock_event = {"data": "value"}
-    persistence_layer = MockPersistenceLayer("test-func#" + hashlib.md5(json.dumps(mock_event).encode()).hexdigest())
+    persistence_layer = MockPersistenceLayer("test-func#" + hashlib.md5(serialize(mock_event).encode()).hexdigest())
     expected_result = {"message": "Foo"}
     keyword_argument = "payload"
 
@@ -938,7 +939,7 @@ def record_handler(record):
 
 def test_idempotent_function_arg_instead_of_kwarg():
     mock_event = {"data": "value"}
-    persistence_layer = MockPersistenceLayer("test-func#" + hashlib.md5(json.dumps(mock_event).encode()).hexdigest())
+    persistence_layer = MockPersistenceLayer("test-func#" + hashlib.md5(serialize(mock_event).encode()).hexdigest())
     expected_result = {"message": "Foo"}
     keyword_argument = "record"
 
@@ -956,7 +957,7 @@ def record_handler(record):
 def test_idempotent_function_and_lambda_handler(lambda_context):
     # Scenario to validate we can use both idempotent_function and idempotent decorators
     mock_event = {"data": "value"}
-    persistence_layer = MockPersistenceLayer("test-func#" + hashlib.md5(json.dumps(mock_event).encode()).hexdigest())
+    persistence_layer = MockPersistenceLayer("test-func#" + hashlib.md5(serialize(mock_event).encode()).hexdigest())
     expected_result = {"message": "Foo"}
 
     @idempotent_function(persistence_store=persistence_layer, data_keyword_argument="record")
@@ -976,3 +977,20 @@ def lambda_handler(event, _):
     # THEN we expect the function and lambda handler to execute successfully
     assert fn_result == expected_result
     assert handler_result == expected_result
+
+
+def test_idempotent_data_sorting():
+    # Scenario to validate same data in different order hashes to the same idempotency key
+    data_one = {"data": "test message 1", "more_data": "more data 1"}
+    data_two = {"more_data": "more data 1", "data": "test message 1"}
+
+    # Assertion will happen in MockPersistenceLayer
+    persistence_layer = MockPersistenceLayer("test-func#" + hashlib.md5(json.dumps(data_one).encode()).hexdigest())
+
+    # GIVEN
+    @idempotent_function(data_keyword_argument="payload", persistence_store=persistence_layer)
+    def dummy(payload):
+        return {"message": "hello"}
+
+    # WHEN
+    dummy(payload=data_two)