aws glue job to import dynamodb data - amazon-dynamodb

Question

We are trying to do DynamoDB migration from prod account to stage account.
In the source account, we are making use of "Export" feature of DDB to put the compressed .json.gz files into destination S3 bucket.
We have written a glue script which will read the exported .json.gz files and writes it to DDB table.
We are making the code generic, so we should be able to migrate any DDB table from prod to stage account.
As part of that process, while testing we are facing issues when we are trying to write a NUMBER SET data to target DDB table.
Following is the sample snippet which is raising ValidationException when trying to insert into DDB
from decimal import Decimal
def number_set(datavalue):
# datavalue will be ['0', '1']
set_of_values = set()
for value in datavalue:
set_of_values.add(Decimal(value))
return set_of_values
When running the code, we are getting following ValidationException
An error occurred while calling o82.pyWriteDynamicFrame. Supplied AttributeValue is empty, must contain exactly one of the supported datatypes (Service: AmazonDynamoDBv2; Status Code: 400; Error Code: ValidationException; Request ID: UKEU70T0BLIKN0K2OL4RU56TGVVV4KQNSO5AEMVJF66Q9ASUAAJG; Proxy: null)
However, if instead of Decimal(value) if we use int(value) then no ValidationException is being thrown and the job succeeds.
I feel that write_dynamic_frame_from_options will try to infer schema based on the values the element contains, if the element has "int" values then the datatype would be "NS", but if the element contains all "Decimal type" values, then it is not able to infer the datatype.
The glue job we have written is
dyf = glue_context.create_dynamic_frame_from_options(
connection_type="s3",
connection_options={
"paths": [file_path]
},
format="json",
transformation_ctx = "dyf",
recurse = True,
)
def number_set(datavalue):
list_of_values = []
for value in datavalue:
list_of_values.append(Decimal(value))
print("list of values ")
print(list_of_values)
return set(list_of_values)
def parse_list(datavalue):
list_of_values = []
for object in datavalue:
list_of_values.append(generic_conversion(object))
return list_of_values
def generic_conversion(value_dict):
for datatype,datavalue in value_dict.items():
if datatype == 'N':
value = Decimal(datavalue)
elif datatype == 'S':
value = datavalue
elif datatype == 'NS':
value = number_set(datavalue)
elif datatype == 'BOOL':
value = datavalue
elif datatype == 'M':
value = construct_map(datavalue)
elif datatype == 'B':
value = datavalue.encode('ascii')
elif datatype == 'L':
value = parse_list(datavalue)
return value
def construct_map(row_dict):
ddb_row = {}
for key,value_dict in row_dict.items():
# value is a dict with key as N or S
# if N then use Decimal type
ddb_row[key] = generic_conversion(value_dict)
return ddb_row
def map_function(rec):
row_dict = rec["Item"]
return construct_map(row_dict)
mapped_dyF = Map.apply(frame = dyf, f = map_function, transformation_ctx = "mapped_dyF")
datasink2 = glue_context.write_dynamic_frame_from_options(
frame=mapped_dyF,
connection_type="dynamodb",
connection_options={
"dynamodb.region": "us-east-1",
"dynamodb.output.tableName": destination_table,
"dynamodb.throughput.write.percent": "0.5"
},
transformation_ctx = "datasink2"
)
can anyone help us in how can we unblock from this situation?
Record that we are trying to insert
{
"region": {
"S": "to_delete"
},
"date": {
"N": "20210916"
},
"number_set": {
"NS": [
"0",
"1"
]
},
"test": {
"BOOL": false
},
"map": {
"M": {
"test": {
"S": "value"
},
"test2": {
"S": "value"
},
"nestedmap": {
"M": {
"key": {
"S": "value"
},
"nestedmap1": {
"M": {
"key1": {
"N": "0"
}
}
}
}
}
}
},
"binary": {
"B": "QUFBY2Q="
},
"list": {
"L": [
{
"S": "abc"
},
{
"S": "def"
},
{
"N": "123"
},
{
"M": {
"key2": {
"S": "value2"
},
"nestedmaplist": {
"M": {
"key3": {
"S": "value3"
}
}
}
}
}
]
}
}