@Hubert Dudek​ thanks for your response! I was able to use what you proposed above to generate the schema. The issue is that the schema sets all attributes to STRING values and renames them numerically ('_c0', '_c1', etc.). Although this allows us to then write stream, it does not accurately portray the schema or the attribute names. If I instead use:

csvFile = spark.read.options(header = 'True',inferSchema='True').csv(data_source)

I can then alter the attribute names with:

json_str = json.dumps(schema_json)
json_str = json_str.replace(" ", "")
schema_json = json.loads(json_str)

We can then create the ddl and pass it as the schema to the readStream.