fix(event_handler): serialize pydantic/dataclasses in exception handler (#3455)

Co-authored-by: Leandro Damascena <lcdama@amazon.pt>

Author: heitorlessa

Makefile

@@ -84,7 +84,7 @@ complexity-baseline:
 	$(info Maintenability index)
 	poetry run radon mi aws_lambda_powertools
 	$(info Cyclomatic complexity index)
-	poetry run xenon --max-absolute C --max-modules A --max-average A aws_lambda_powertools
+	poetry run xenon --max-absolute C --max-modules A --max-average A aws_lambda_powertools --exclude aws_lambda_powertools/shared/json_encoder.py
 
 #
 # Use `poetry version <major>/<minor></patch>` for version bump

aws_lambda_powertools/event_handler/api_gateway.py

@@ -709,7 +709,7 @@ class ResponseBuilder(Generic[ResponseEventT]):
     def __init__(
         self,
         response: Response,
-        serializer: Callable[[Any], str] = json.dumps,
+        serializer: Callable[[Any], str] = partial(json.dumps, separators=(",", ":"), cls=Encoder),
         route: Optional[Route] = None,
     ):
         self.response = response

aws_lambda_powertools/shared/functions.py

@@ -1,7 +1,6 @@
 from __future__ import annotations
 
 import base64
-import dataclasses
 import itertools
 import logging
 import os
@@ -168,8 +167,86 @@ def extract_event_from_common_models(data: Any) -> Dict | Any:
         return data.raw_event
 
     # Is it a Pydantic Model?
-    if callable(getattr(data, "dict", None)):
-        return data.dict()
+    if is_pydantic(data):
+        return pydantic_to_dict(data)
 
-    # Is it a Dataclass? If not return as is
-    return dataclasses.asdict(data) if dataclasses.is_dataclass(data) else data
+    # Is it a Dataclass?
+    if is_dataclass(data):
+        return dataclass_to_dict(data)
+
+    # Return as is
+    return data
+
+
+def is_pydantic(data) -> bool:
+    """Whether data is a Pydantic model by checking common field available in v1/v2
+
+    Parameters
+    ----------
+    data: BaseModel
+        Pydantic model
+
+    Returns
+    -------
+    bool
+        Whether it's a Pydantic model
+    """
+    return getattr(data, "json", False)
+
+
+def is_dataclass(data) -> bool:
+    """Whether data is a dataclass
+
+    Parameters
+    ----------
+    data: dataclass
+        Dataclass obj
+
+    Returns
+    -------
+    bool
+        Whether it's a Dataclass
+    """
+    return getattr(data, "__dataclass_fields__", False)
+
+
+def pydantic_to_dict(data) -> dict:
+    """Dump Pydantic model v1 and v2 as dict.
+
+    Note we use lazy import since Pydantic is an optional dependency.
+
+    Parameters
+    ----------
+    data: BaseModel
+        Pydantic model
+
+    Returns
+    -------
+
+    dict:
+        Pydantic model serialized to dict
+    """
+    from aws_lambda_powertools.event_handler.openapi.compat import _model_dump
+
+    return _model_dump(data)
+
+
+def dataclass_to_dict(data) -> dict:
+    """Dump standard dataclass as dict.
+
+    Note we use lazy import to prevent bloating other code parts.
+
+    Parameters
+    ----------
+    data: dataclass
+        Dataclass
+
+    Returns
+    -------
+
+    dict:
+        Pydantic model serialized to dict
+    """
+    import dataclasses
+
+    return dataclasses.asdict(data)

aws_lambda_powertools/shared/json_encoder.py

@@ -2,15 +2,25 @@
 import json
 import math
 
+from aws_lambda_powertools.shared.functions import dataclass_to_dict, is_dataclass, is_pydantic, pydantic_to_dict
+
 
 class Encoder(json.JSONEncoder):
-    """
-    Custom JSON encoder to allow for serialization of Decimals, similar to the serializer used by Lambda internally.
+    """Custom JSON encoder to allow for serialization of Decimals, Pydantic and Dataclasses.
+
+    It's similar to the serializer used by Lambda internally.
     """
 
     def default(self, obj):
         if isinstance(obj, decimal.Decimal):
             if obj.is_nan():
                 return math.nan
             return str(obj)
+
+        if is_pydantic(obj):
+            return pydantic_to_dict(obj)
+
+        if is_dataclass(obj):
+            return dataclass_to_dict(obj)
+
         return super().default(obj)

tests/functional/event_handler/test_api_gateway.py

@@ -10,6 +10,7 @@
 from typing import Dict
 
 import pytest
+from pydantic import BaseModel
 
 from aws_lambda_powertools.event_handler import content_types
 from aws_lambda_powertools.event_handler.api_gateway import (
@@ -1465,7 +1466,6 @@ def test_exception_handler_with_data_validation():
 
     @app.exception_handler(RequestValidationError)
     def handle_validation_error(ex: RequestValidationError):
-        print(f"request path is '{app.current_event.path}'")
         return Response(
             status_code=422,
             content_type=content_types.TEXT_PLAIN,
@@ -1486,6 +1486,34 @@ def get_lambda(param: int):
     assert result["body"] == "Invalid data. Number of errors: 1"
 
 
+def test_exception_handler_with_data_validation_pydantic_response():
+    # GIVEN a resolver with an exception handler defined for RequestValidationError
+    app = ApiGatewayResolver(enable_validation=True)
+
+    class Err(BaseModel):
+        msg: str
+
+    @app.exception_handler(RequestValidationError)
+    def handle_validation_error(ex: RequestValidationError):
+        return Response(
+            status_code=422,
+            content_type=content_types.APPLICATION_JSON,
+            body=Err(msg=f"Invalid data. Number of errors: {len(ex.errors())}"),
+        )
+
+    @app.get("/my/path")
+    def get_lambda(param: int):
+        ...
+
+    # WHEN calling the event handler
+    # AND a RequestValidationError is raised
+    result = app(LOAD_GW_EVENT, {})
+
+    # THEN exception handler's pydantic response should be serialized correctly
+    assert result["statusCode"] == 422
+    assert result["body"] == '{"msg":"Invalid data. Number of errors: 1"}'
+
+
 def test_data_validation_error():
     # GIVEN a resolver without an exception handler
     app = ApiGatewayResolver(enable_validation=True)

tests/unit/test_json_encoder.py

@@ -1,7 +1,9 @@
 import decimal
 import json
+from dataclasses import dataclass
 
 import pytest
+from pydantic import BaseModel
 
 from aws_lambda_powertools.shared.json_encoder import Encoder
 
@@ -22,3 +24,35 @@ class CustomClass:
 
     with pytest.raises(TypeError):
         json.dumps({"val": CustomClass()}, cls=Encoder)
+
+
+def test_json_encode_pydantic():
+    # GIVEN a Pydantic model
+    class Model(BaseModel):
+        data: dict
+
+    data = {"msg": "hello"}
+    model = Model(data=data)
+
+    # WHEN json.dumps use our custom Encoder
+    result = json.dumps(model, cls=Encoder)
+
+    # THEN we should serialize successfully; not raise a TypeError
+    assert result == json.dumps({"data": data}, cls=Encoder)
+
+
+def test_json_encode_dataclasses():
+    # GIVEN a standard dataclass
+
+    @dataclass
+    class Model:
+        data: dict
+
+    data = {"msg": "hello"}
+    model = Model(data=data)
+
+    # WHEN json.dumps use our custom Encoder
+    result = json.dumps(model, cls=Encoder)
+
+    # THEN we should serialize successfully; not raise a TypeError
+    assert result == json.dumps({"data": data}, cls=Encoder)