Repository navigation
Bug: Incomplete validation of annotated parameter types (3.21.0 regression) #7600
Description
Expected Behaviour
Annotated parameter types using annotated_types should be validated by Pydantic. This used to work, but regressed starting from 3.21.0. Types are still validated, but constraint annotations are not.
I thought #7227 would have enabled using Pydantic Field annotations for this use case, and tried to switch to that, but that also doesn't work (in any Powertools version). Should I report this separately (perhaps as a feature request)?
Current Behaviour
Parameter types are validated, but constraints expressed using annotated_types are not enforced. Handler functions can be called with invalid parameter values.
Code snippet
from http import HTTPStatus
from typing import Annotated
from aws_lambda_powertools.event_handler import APIGatewayHttpResolver
from aws_lambda_powertools.event_handler.openapi.params import Body, Query
import pytest
# aws-lambda-powertools 3.20.0: ok
# aws-lambda-powertools 3.21.0 and 3.22.0: fail (value outside of range allowed)
from annotated_types import Interval
ConstrainedInt = Annotated[int, Interval(ge=0, le=100)]
# aws-lambda-powertools 3.20.0 and 3.21.0 and 3.22.0: error (Only one FieldInfo can be used per parameter)
#from pydantic.fields import Field
#ConstrainedInt = Annotated[int, Field(ge=0, le=100)]
app = APIGatewayHttpResolver(enable_validation=True)
@app.post('/body')
def handle_body(value: Annotated[ConstrainedInt, Body()]) -> None:
assert isinstance(value, int)
assert value >= 0
assert value <= 100
@app.post('/query')
def handle_body(value: Annotated[ConstrainedInt, Query()]) -> None:
assert isinstance(value, int)
assert value >= 0
assert value <= 100
def handler(event, context):
return app.resolve(event, context)
params = [
(0, HTTPStatus.OK),
(50, HTTPStatus.OK),
(100, HTTPStatus.OK),
(-1, HTTPStatus.UNPROCESSABLE_ENTITY),
(101, HTTPStatus.UNPROCESSABLE_ENTITY),
('not an int', HTTPStatus.UNPROCESSABLE_ENTITY),
]
@pytest.mark.parametrize('value,expected_status', params)
def test_handle_body(value, expected_status):
event = {
'rawPath': '/body',
'requestContext': {
'http': {
'method': 'POST'
},
'stage': '$default'
},
'body': str(value)
}
response = handler(event, object())
assert expected_status == response['statusCode']
@pytest.mark.parametrize('value,expected_status', params)
def test_handle_query(value, expected_status):
event = {
'rawPath': '/query',
'requestContext': {
'http': {
'method': 'POST'
},
'stage': '$default'
},
'queryStringParameters': {
'value': str(value)
}
}
response = handler(event, object())
assert expected_status == response['statusCode']Possible Solution
No response
Steps to Reproduce
Run code snippet: pytest test_annotated_types.py
Powertools for AWS Lambda (Python) version
3.21.0, 3.22.0
AWS Lambda function runtime
3.11
Packaging format used
Lambda Layers
Activity
leandrodamascena commentedon Oct 30, 2025
Hey @rtandy thanks for reporting this, I will take a look.
leandrodamascena commentedon Oct 30, 2025
Hi @rtandy, can you confirm if this code (exactly as you pasted it here) works with Powertools <= 3.20.0? Without checking the code and just running your script, it fails in any version of Powertools. I'll check it in more detail, but I just wanted to confirm before dive deep into it.
rtandy commentedon Oct 30, 2025
Hi @leandrodamascena, thanks for responding so quickly. Confirmed, copy-pasted the exact code snippet above. Environment is macOS Sequoia 15.7.1, Python 3.11.14, Powertools 3.20.0, Pydantic 2.11.10.
And confirmed it fails with 3.21.0 and 3.22.0 with Pydantic 2.11 or 2.12.
leandrodamascena commentedon Oct 30, 2025
Ohh thanks for confirming.. I was using Pydantic 2.12+ and I think this is related with this: #7552.
Regarding the bug you opened, I think this is related with PR https://github.com/aws-powertools/powertools-lambda-python/pull/7227/files not checking the constraints. I'll take a look into it.
leandrodamascena commentedon Oct 31, 2025
Thanks for reporting this @rtandy ! I've identified and fixed the regression.
The problem was introduced in PR #7227 when we switched from
FieldInfo.from_annotation()to usingcopy()in thecopy_field_info()function. While this change correctly preserved discriminator functionality, thecopydoesn't copy theannotated_typesconstraint metadata (likeInterval) during the field copying process.When you use
Annotated[int, Interval(ge=0, le=100), Body()], the annotation contains multiple layers of metadata. The old implementation was doing a simple shallow copy and directly assigning the new annotation, which meant nested constraint metadata fromannotated_typeswas being discarded.The fix recursively extracts all constraint metadata from nested
Annotatedtypes while filtering outFieldInfoinstances. This way, we preserve both the discriminator functionality from PR #7227 and theannotated_typesconstraints that were working before.I've also added tests to ensure Interval constraints work correctly with both Body and Query parameters, covering valid values, boundary conditions, and validation failures.
I'll be sending a PR soon and launching it by Wednesday the 5th. Is it okay for you to wait until then?
rtandy commentedon Oct 31, 2025
No rush at all - no problem to stay pinned to the older version for a while. Thank you very much for handling it so quickly!
leandrodamascena commentedon Oct 31, 2025
Hey @rtandy! I was wondering if you can install Powertools from #7609 and run your tests to see if this fix the problem.
Thanks
rtandy commentedon Oct 31, 2025
Hi @leandrodamascena, your branch passed all my tests for things that used to work before 3.21. (Noting that Pydantic
Fieldannotations still don't work; but they never did before either, so no change.) Thank you!powertools-for-aws-oss-automation commentedon Nov 3, 2025
Warning
This issue is now closed. Please be mindful that future comments are hard for our team to see.
If you need more assistance, please either reopen the issue, or open a new issue referencing this one.
If you wish to keep having a conversation with other community members under this issue feel free to do so.
github-actions commentedon Nov 5, 2025
This is now released under 3.22.1 version!