Bug: The KinesisFirehoseRecordMetadata model is typed incorrectly and forces a pydantic ValidationError.

[ahitchin]

Expected Behaviour

1. Overview

If you are using a data processor Lambda Function for a Kinesis Firehose Delivery Stream, you should be able to use the KinesisFirehoseModel model to validate events without encountering a pydantic ValidationError.

Current Behaviour

1. Overview

When you validate a Kinesis Firehose Delivery Stream event with the KinesisFirehoseModel model, it fails because the kinesis record metadata's subsequenceNumber (records[].kinesisRecordMetadata.subsequenceNumber) attribute has an incorrect type. This forces pydantic to throw a ValidationError that should not happen.

Specifically, the KinesisFirehoseModel model starts validating its sub-structures. Once it validates its kinesis records' metadata with the KinesisFirehoseRecordMetadata model, it fails. This is because sub-sequence numbers are integers, but the model's subsequenceNumber attribute is expecting a string.

2. ValidationError Output

pydantic_core._pydantic_core.ValidationError: 1 validation error for KinesisFirehoseModel
records.0.kinesisRecordMetadata.subsequenceNumber
  Input should be a valid string [type=string_type, input_value=0, input_type=int]
    For further information visit https://errors.pydantic.dev/2.4/v/string_type

Code snippet

#!/usr/bin/env python

from aws_lambda_powertools.utilities.parser.models import KinesisFirehoseModel


def main() -> None:
    """Code snippet that shows the previously mentioned validation error."""
    # Create a kinesis firehose delivery stream event
    firehose_event: dict = {
        "invocationId": "00000000-0000-0000-0000-000000000000",
        "sourceKinesisStreamArn": "arn:aws-us-gov:kinesis:us-gov-west-1:000000000000:stream/A",
        "deliveryStreamArn": "arn:aws-us-gov:firehose:us-gov-west-1:000000000000:deliverystream/A",
        "region": "us-gov-west-1",
        "records": [
            {
                "recordId": "00000000000000000000000000000000000000000000000000000000000000",
                "approximateArrivalTimestamp": 1697943843714,
                "data": "YnVnIHJlcG9ydA==",
                "kinesisRecordMetadata": {
                    "sequenceNumber": "00000000000000000000000000000000000000000000000000000000",
                    "subsequenceNumber": 0,
                    "partitionKey": "00000000000000000000000000000000",
                    "shardId": "shardId-000000000000",
                    "approximateArrivalTimestamp": 1697943843714
                }
            }
        ]
    }

    # Create a model from the event (this fails)
    model: KinesisFirehoseModel = KinesisFirehoseModel.model_validate(firehose_event)
    print(model)

    return None

if __name__ == "__main__":
    main()

Possible Solution

1. Overview

Update the Kinesis Firehose pydantic model and data class with the correct type hints.

2. Problematic Code

• https://github.com/aws-powertools/powertools-lambda-python/blob/develop/aws_lambda_powertools/utilities/data_classes/kinesis_firehose_event.py#L203-L209
• https://github.com/aws-powertools/powertools-lambda-python/blob/develop/aws_lambda_powertools/utilities/parser/models/kinesis_firehose.py#L8-L13

3. Proposal for the Pydantic Model

class KinesisFirehoseRecordMetadata(BaseModel):
    shardId: str
    partitionKey: str
    approximateArrivalTimestamp: PositiveInt
    sequenceNumber: str
    subsequenceNumber: int

4. Proposal for the Data Class Dictionary Wrapper

class KinesisFirehoseRecordMetadata(DictWrapper):
    @property
    def subsequence_number(self) -> int:
        """Kinesis stream sub-sequence number; present only when Kinesis Stream is source

        Note: this will only be present for Kinesis streams using record aggregation
        """
        return self._metadata["subsequenceNumber"]

5. Examples From Other Amazon/AWS Packages

• https://github.com/awslabs/amazon-kinesis-client-net/blob/master/ClientLibrary/inputs/Record.cs#L27-L32
• https://github.com/aws/aws-lambda-go/blob/main/events/firehose.go#L43-L49
• https://github.com/awslabs/amazon-kinesis-client/blob/master/amazon-kinesis-client/src/main/java/software/amazon/kinesis/retrieval/kpl/ExtendedSequenceNumber.java#L37-L39

Steps to Reproduce

1. Overview

1. Have an event from a Kinesis Firehose Delivery Stream.
2. Validate the event with the KinesisFirehoseModel pydantic model.
3. Encounter an error when the KinesisFirehoseRecordMetadata fails to validate.

2. Sample Event

{
  "invocationId": "00000000-0000-0000-0000-000000000000",
  "sourceKinesisStreamArn": "arn:aws-us-gov:kinesis:us-gov-west-1:000000000000:stream/A",
  "deliveryStreamArn": "arn:aws-us-gov:firehose:us-gov-west-1:000000000000:deliverystream/A",
  "region": "us-gov-west-1",
  "records": [
    {
      "recordId": "00000000000000000000000000000000000000000000000000000000000000",
      "approximateArrivalTimestamp": 1697943843714,
      "data": "YnVnIHJlcG9ydA==",
      "kinesisRecordMetadata": {
        "sequenceNumber": "00000000000000000000000000000000000000000000000000000000",
        "subsequenceNumber": 0,
        "partitionKey": "00000000000000000000000000000000",
        "shardId": "shardId-000000000000",
        "approximateArrivalTimestamp": 1697943843714
      }
    }
  ]
}

Powertools for AWS Lambda (Python) version

2.26.0

AWS Lambda function runtime

3.11

Packaging format used

PyPi

Debugging logs

N/A

[boring-cyborg]

Thanks for opening your first issue here! We'll come back to you as soon as we can.
In the meantime, check out the #python channel on our Powertools for AWS Lambda Discord: Invite link

[leandrodamascena]

Hi @ahitchin! Thanks for opening this issue and I can confirm it's a bug. Checking the links you sent and here this field should actually be integer instead of str.

Would you like to submit a PR to fix this? We'd love to have your first contribution here. If so, change the class, the test and also the [event] (https://github.com/aws-powertools/powertools-lambda-python/blob/develop/tests/events/kinesisFirehoseKinesisEvent.json). You can do the same for the data class test.