python boto3 How to configure AWS s3select on parquet?

Question

I am trying to use AWS s3select feature to query a parquet file. According to the documentation it's supported but I have tried various configurations and can't get it to work. In each of the InputSerialization attempts I've shown commented-out, I've listed the error I'm receiving when I try that version. Can someone show me how to properly configure this?

import boto3

S3_BUCKET = 'myBucket'
KEY_LIST = "'0123','6789'"
S3_FILE = 'myFolder/myFile.parquet'

s3 = boto3.client('s3')

r = s3.select_object_content(
        Bucket=S3_BUCKET,
        Key=S3_FILE,
        ExpressionType='SQL',
        Expression="select \"Record\" from s3object s where s.\"Key\" in [" + KEY_LIST + "]",
#        InputSerialization={}, # (MissingRequiredParameter) when calling the SelectObjectContent operation: InputSerialization is required
#        InputSerialization={'CompressionType': { 'NONE' }},    # Invalid type for parameter InputSerialization.CompressionType, value: {'NONE'}, type: <class 'set'>, valid types: <class 'str'>
#        InputSerialization={'Parquet': {}}, # Unknown parameter in InputSerialization: "Parquet", must be one of: CSV, CompressionType, JSON
#        InputSerialization={'CompressionType': { 'Snappy' }},    # Invalid type for parameter InputSerialization.CompressionType, value: {'Snappy'}, type: <class 'set'>, valid types: <class 'str'>

        OutputSerialization={'JSON': {}},
)

for event in r['Payload']:
    if 'Records' in event:
        records = event['Records']['Payload'].decode('utf-8')
        print(records)

Answer 1

I needed to upgrade my boto3 install to the latest version. After upgrading to 1.9.7, this version worked:

InputSerialization={'Parquet': {}},