As noted in #357 , the %manage_spark user interface only provides 'scala' and 'python' language options. In order for Python 3 to be used and unencoded JSON dataframes to be returned to the local environment (via %spark -o), 'python3' should also be included.
