Bug: idempotent functions return None on replay when the stored response is falsy (0, False, "", [], {})

Aberta
  • Algo errado?
#8,700 3 comentários 0 reações 1 responsável Ver no GitHub

Mantenedores costumam responder em até 2 dias

@thebadalov já está trabalhando nisso.

Desde 29/10/2026.

  • #8701 de @thebadalov — aberto

Avaliação

Esta issue ainda não foi avaliada.

Descrição

bug idempotency
Expected Behaviour

When an idempotent function returns a falsy value (0, False, "", [], {}), a retry with the same payload should return that same value from the persistence store.

Current Behaviour

The first call returns the value and it is stored correctly, but every replay returns None.

IdempotencyHandler._handle_for_status reads the stored response with:

serialized_response = self.output_serializer.from_dict(response_dict) if response_dict else None

so any falsy response_dict becomes None. #8090 pointed out this line together with the write path in _get_function_response. The write path was changed to is not None, but the read path still uses a truthiness check, so the value is saved and then dropped on replay.

Code snippet
from aws_lambda_powertools.utilities.idempotency import idempotent_function
from aws_lambda_powertools.utilities.idempotency.exceptions import (
    IdempotencyItemAlreadyExistsError,
    IdempotencyItemNotFoundError,
)
from aws_lambda_powertools.utilities.idempotency.persistence.base import BasePersistenceLayer, DataRecord


class InMemoryPersistenceLayer(BasePersistenceLayer):
    def __init__(self):
        super().__init__()
        self.records = {}

    def _get_record(self, idempotency_key) -> DataRecord:
        if idempotency_key not in self.records:
            raise IdempotencyItemNotFoundError
        return self.records[idempotency_key]

    def _put_record(self, data_record: DataRecord) -> None:
        if data_record.idempotency_key in self.records:
            raise IdempotencyItemAlreadyExistsError
        self.records[data_record.idempotency_key] = data_record

    def _update_record(self, data_record: DataRecord) -> None:
        self.records[data_record.idempotency_key] = data_record

    def _delete_record(self, data_record: DataRecord) -> None:
        self.records.pop(data_record.idempotency_key, None)


persistence_layer = InMemoryPersistenceLayer()


@idempotent_function(data_keyword_argument="order", persistence_store=persistence_layer)
def count_items(order):
    return len(order["items"])


order = {"order_id": "o-1", "items": []}
print(count_items(order=order))  # 0
print(count_items(order=order))  # None, expected 0

I see the same with CachePersistenceLayer on a real Redis server: 0, False, "", [] and {} all come back as None on replay, while None and truthy values are returned correctly.

Possible Solution

Use is not None, matching the write path:

serialized_response = self.output_serializer.from_dict(response_dict) if response_dict is not None else None

DataRecord.response_json_as_dict() already returns None when there is no stored data, and a function that returns None is stored as null, so that case keeps returning None.

Steps to Reproduce

Run the snippet above. Any persistence layer is affected, since the check is in the shared handler.

Powertools for AWS Lambda (Python) version

latest (3.37.0, also on develop)

AWS Lambda function runtime

3.13

Packaging format used

PyPi

Debugging logs
N/A, the replay returns None without an error
Linguagem predominante
Python
Estrelas
3.3k
Forks
511
Merge médio
3d 12h
PRs com merge (30d)
93

Preparar o ambiente

Primeiros passos

  1. Leia a issue inteira e depois o guia de contribuição do projeto.
  2. Comente na issue dizendo que vai assumir — evita que duas pessoas façam o mesmo trabalho.
  3. Faça um fork do repositório e trabalhe em uma branch.
  4. Abra um pull request que referencie o número da issue.

Mais de aws-powertools/powertools-lambda-python

Todas as issues de aws-powertools/powertools-lambda-python

Issues semelhantes

Mais issues de Python

Receba novas issues na sua caixa de entrada

Um resumo curto de issues do GitHub para quem está começando.