test_hyperopt.py 文件源码

def df_train(spark_context, hive_context):
    # TODO: Use some fixture dataset representing real-ish data? But
    # it needs to be pretty small
    return spark_context.parallelize(
        _make_q('abc') + _make_q('def') + _make_q('ghi') + _make_q('jkl')
        + _make_q('mno') + _make_q('pqr') + _make_q('stu')
    ).toDF(['wikiid', 'norm_query_id', 'query', 'label', 'features'])