how does TypeDeserialization work with DynamoDBStreamEvent

[Muthuveerappanv]

Trying to de-serialize dynamodb stream event in lambda, when using the below, number types get transformed to string

# event: DynamoDBStreamEvent
for record in event.records:
        if record.event_name == DynamoDBRecordEventName.REMOVE:
          output = {k: v.get_value for k, v in data.items()} 

so tried using type deserializer from boto3, but below didn't work as v is a object and not a dict

from boto3.dynamodb.types import TypeDeserializer
deserializer = TypeDeserializer()
# event: DynamoDBStreamEvent
for record in event.records:
        if record.event_name == DynamoDBRecordEventName.REMOVE:
           output = {k: deserializer.deserialize(v) for k, v in data.items()}
# ERROR - AttributeError: 'AttributeValue' object has no attribute 'keys'

What worked - was passing the _data dict field to the serializer

from boto3.dynamodb.types import TypeDeserializer
deserializer = TypeDeserializer()
# event: DynamoDBStreamEvent
for record in event.records:
        if record.event_name == DynamoDBRecordEventName.REMOVE:
           output = {k: deserializer.deserialize(v._data) for k, v in data.items()}

Is there a way to derserialize the AttributeValue class using - using TypeDeserializer
(When we have a huge data-set or dynamic values coming in, looking up the AttributeValueType for each value becomes tedious)

[michaelbrewer]

@Muthuveerappanv would it be possible to post a more complete example with an event, so i can reproduce your issue and see if we can make a fix for it or use the existing api to do what you want.

Otherwise get_value was added to AttributeValue to do the "pythonic value of the DynamoDB type" that the TypeDeserializer expects

[Muthuveerappanv]

ddb_stream_deserialize.zip

**See if this helps. ☝🏻 **
get_value helps get value of individual attribute. What if I want to transform the entire record.
As highlighted in the example, getting the _data from the object, helps get the dict and from there the deserializer can work, but i wanted to check if that was the intended way from a powertools data model perspective

from_dict = {k: deserializer.deserialize(v) for k, v in record.dynamodb._data["OldImage"].items()}
            print("Works with Dict", from_dict)

[michaelbrewer]

makes sense. And it might not be obvious using _data is the best choice. Maybe a helper function with docs might help.