Amazon web services – AWS Glue – Create Dynamic Frame for Dynamo DB Table having columns with mixed type

DynamoDB Table:

+------------------+---------+---------------------+
|        id        |Column A |             Column B|
+------------------+---------+---------------------+
|1                 |  "155.2"|                 400 |
|2                 |      100|                 200 |
|3                 |  "455.2"|               305.5 |
|4                 |  "312.3"|                 350 |
+------------------+---------+---------------------+

Notice that in Column A. We have numbers stored as strings except for id = 2.

Following code is used to read the table contents into a Dynamic Frame:

def create_dynamic_frame(table_name):
    ddb_s3_bucket = <some-s3-bucket>
    ddb_table_arn = <some-table-arn>
    connection_options = {
        "dynamodb.export": "ddb",
        "dynamodb.unnestDDBJson": True,
        "dynamodb.tableArn": ddb_table_arn,
        "dynamodb.s3.bucket": ddb_s3_bucket,
        "dynamodb.s3.prefix": 'temporary/ddbexport/'
    }
    dynamic_frame = glueContext.create_dynamic_frame.from_options(
        connection_type="dynamodb",
        connection_options=connection_options,
        transformation_ctx="dynamic_frame",
    )
    return dynamic_frame

dyf = create_dynamic_frame('test-table')

The output of show on the created Dynamic Frame: dyf.toDf().show()

+------------------+---------+---------------------+
|        id        |Column A |             Column B|
+------------------+---------+---------------------+
|1                 |     null|                 400 |
|2                 |     100 |                 200 |
|3                 |     null|               305.5 |
|4                 |     null|                 350 |
+------------------+---------+---------------------+

The output of dyf.toDf().printSchema():

root
 |-- id: string (nullable = true)
 |-- Column A: string (nullable = true)
 |-- Column B: string (nullable = true)

Notice that the string values in Column A are null. I was under the impression that Glue keeps both types in the column and you can use resolveChoice to then cast to whichever type you would want.

Is there a way I can resolve the type in the Glue-DDB connector?

I tried to resolve the types using resolveChoice:

resolved_dyf = dyf.resolveChoice(specs = [("Column A", "cast:string")])

This did not work, since the values in dyf itself are null