Python pyspark.sql.types.StructType() Examples

def _get_instance_usage_schema():
        """get instance usage schema."""

        # Initialize columns for all string fields
        columns = ["tenant_id", "user_id", "resource_uuid",
                   "geolocation", "region", "zone", "host", "project_id",
                   "aggregated_metric_name", "firstrecord_timestamp_string",
                   "lastrecord_timestamp_string",
                   "service_group", "service_id",
                   "usage_date", "usage_hour", "usage_minute",
                   "aggregation_period", "namespace", "pod_name", "app",
                   "container_name", "interface", "deployment", "daemon_set"]

        columns_struct_fields = [StructField(field_name, StringType(), True)
                                 for field_name in columns]

        # Add columns for non-string fields
        columns_struct_fields.append(StructField("firstrecord_timestamp_unix",
                                                 DoubleType(), True))
        columns_struct_fields.append(StructField("lastrecord_timestamp_unix",
                                                 DoubleType(), True))
        columns_struct_fields.append(StructField("quantity",
                                                 DoubleType(), True))
        columns_struct_fields.append(StructField("record_count",
                                                 DoubleType(), True))

        columns_struct_fields.append(StructField("processing_meta",
                                                 MapType(StringType(),
                                                         StringType(),
                                                         True),
                                                 True))
        schema = StructType(columns_struct_fields)

        return schema 

def create_df(spark_session):
    try:
        # Building a simple dataframe:
        schema = StructType([
            StructField("city", StringType(), True),
            StructField("country", StringType(), True),
            StructField("population", IntegerType(), True)])

        countries = ['Colombia   ', '[email protected]', 'Brazil', 'Spain']
        cities = ['Bogotá', 'New York', '   São Paulo   ', '~Madrid']
        population = [37800000, 19795791, 12341418, 6489162]

        # Dataframe:
        df = spark_session.createDataFrame(list(zip(cities, countries, population)), schema=schema)
        assert_spark_df(df)
        return df
    except RuntimeError:
        sys.exit(1) 

def test_seaprate_key_from_end(self):
        spark = SparkSession.builder.getOrCreate()
        sc = spark.sparkContext

        rdd = sc.parallelize([
            (100, 4000, "217.69.143.60"),
            (100, 4000, "217.69.143.60"),
            (1500, 54000, "192.168.30.2"),
            (200, 3000, "192.168.30.2"),
            (200, 3000, "192.168.30.2")
        ])

        config = Config(CONFIG_PATH)
        aggregation_processor = AggregationProcessor(config, StructType([packet_size, traffic, src_ip]))

        separate_key = aggregation_processor._get_separate_key_lambda()
        result = separate_key(rdd)
        self.assertListEqual(result.collect(),
                             [(('217.69.143.60',), (100, 4000)), (('217.69.143.60',), (100, 4000)),
                              (('192.168.30.2',), (1500, 54000)), (('192.168.30.2',), (200, 3000)),
                              (('192.168.30.2',), (200, 3000))], "Lists should be equal") 

def test_union_with_udt(self):
        from pyspark.sql.tests import ExamplePoint, ExamplePointUDT
        row1 = (1.0, ExamplePoint(1.0, 2.0))
        row2 = (2.0, ExamplePoint(3.0, 4.0))
        schema = StructType([StructField("label", DoubleType(), False),
                             StructField("point", ExamplePointUDT(), False)])
        df1 = self.spark.createDataFrame([row1], schema)
        df2 = self.spark.createDataFrame([row2], schema)

        result = df1.union(df2).orderBy("label").collect()
        self.assertEqual(
            result,
            [
                Row(label=1.0, point=ExamplePoint(1.0, 2.0)),
                Row(label=2.0, point=ExamplePoint(3.0, 4.0))
            ]
        ) 

def test_validate_with_correct_one_level_subtree(self):
        validator = TransformatoinsValidator(TransformationOperations({
            "country": "./GeoLite2-Country.mmdb",
            "city": "./GeoLite2-City.mmdb",
            "asn": "./GeoLite2-ASN.mmdb"
        }), self.data_structure_pyspark)

        syntaxtree = SyntaxTree()
        syntaxtree.operation = "mult"
        syntaxtree.children = ["packet_size", "sampling_rate"]

        fields = validator.validate([FieldTransformation("traffic", syntaxtree), "dst_ip"])

        self.assertEqual(fields, types.StructType([
            types.StructField('traffic', types.LongType()),
            types.StructField('dst_ip', types.StringType())
        ])) 

def test_validate_function_with_different_arguments_type(self):
        validator = TransformatoinsValidator(TransformationOperations({
            "country": "./GeoLite2-Country.mmdb",
            "city": "./GeoLite2-City.mmdb",
            "asn": "./GeoLite2-ASN.mmdb"
        }), self.data_structure_pyspark)

        main_syntax_tree = SyntaxTree()
        main_syntax_tree.operation = "truncate"
        main_syntax_tree.children = ["src_ip", "5"]

        fields = validator.validate([FieldTransformation("result", main_syntax_tree)])

        self.assertEqual(fields, types.StructType([
            types.StructField("result", types.StringType())
        ])) 

def test_build_lambda_for_reduce(self):
        test_input_rule = "Min(packet_size);Max(sampling_rate); Sum(traffic)"
        input_data_structure = StructType([StructField("sampling_rate", LongType()),
                                           StructField("packet_size", LongType()),
                                           StructField("traffic", LongType())])
        test_input_operation = "reduce"
        config = TestConfig({"processing": {"aggregations": {"operation_type": test_input_operation,
                                                             "rule": test_input_rule}}})
        test_aggregation_processor = AggregationProcessor(config, input_data_structure)

        spark = SparkSession.builder.getOrCreate()
        sc = spark.sparkContext
        test_rdd = sc.parallelize([(4, 2, 1), (7, 1, 1), (1, 0, 1), (2, 5, 1), (1, 1, 1)])
        test_aggregation_lambda = test_aggregation_processor.get_aggregation_lambda()

        self.assertIsInstance(test_aggregation_lambda, types.LambdaType, "get_aggregation_lambda should return "
                                                                         "lambda function")

        test_result = test_aggregation_lambda(test_rdd)
        self.assertTupleEqual(test_result, (7, 0, 5), "Error in aggregation operation. Tuple should be equal") 

def test_separate_key_from_center(self):
        spark = SparkSession.builder.getOrCreate()
        sc = spark.sparkContext

        rdd = sc.parallelize([
            (100, "217.69.143.60", 4000),
            (100, "217.69.143.60", 4000),
            (1500, "192.168.30.2", 54000),
            (200, "192.168.30.2", 3000),
            (200, "192.168.30.2", 3000)
        ])

        config = Config(CONFIG_PATH)
        aggregation_processor = AggregationProcessor(config, StructType([packet_size, src_ip, traffic]))

        separate_key = aggregation_processor._get_separate_key_lambda()
        result = separate_key(rdd)
        self.assertListEqual(result.collect(),
                             [(('217.69.143.60',), (100, 4000)), (('217.69.143.60',), (100, 4000)),
                              (('192.168.30.2',), (1500, 54000)), (('192.168.30.2',), (200, 3000)),
                              (('192.168.30.2',), (200, 3000))], "Lists should be equal") 

def create_other_df(spark_session):
    try:
        # Building a simple dataframe:
        schema = StructType([
            StructField("city", StringType(), True),
            StructField("dates", StringType(), True),
            StructField("population", IntegerType(), True)])

        dates = ['1991/02/25', '1998/05/10', '1993/03/15', '1992/07/17']
        cities = ['Caracas', 'Ccs', '   São Paulo   ', '~Madrid']
        population = [37800000, 19795791, 12341418, 6489162]

        # Dataframe:
        df = spark_session.createDataFrame(list(zip(cities, dates, population)), schema=schema)
        assert_spark_df(df)
        return df
    except RuntimeError:
        sys.exit(1) 

def create_another_df(spark_session):
    try:
        # Building a simple dataframe:
        schema = StructType([
            StructField("city", StringType(), True),
            StructField("dates", StringType(), True),
            StructField("population", IntegerType(), True)])

        dates = ['1991/02/25', '1998/05/10', '1993/03/15', '1992/07/17']
        cities = ['Caracas', 'Caracas', '   Maracaibo   ', 'Madrid']
        population = [37800000, 19795791, 12341418, 6489162]

        # Dataframe:
        df = spark_session.createDataFrame(list(zip(cities, dates, population)), schema=schema)
        assert_spark_df(df)
        return df
    except RuntimeError:
        sys.exit(1) 

def local_process(session, sc, input_directory, context_size):
    from os import listdir
    from os.path import isfile, join
    filenames = [filename for filename in listdir(input_directory) if isfile(join(input_directory, filename))]
    file_paths = ["input/CC-MAIN-20170423031158-00000-ip-10-145-167-34.ec2.internal.warc.wet.gz"]
    for file_path in file_paths:
        filename = file_path.split("/")[-1]
        corpus = WARCFile(fileobj=gzip.open(file_path))
        list_of_records = sc.parallelize(
            [re.sub(r'[^\x00-\x7F]+', ' ', record.payload.read().decode("UTF-8")) for i, record in enumerate(corpus)], 10)
        coor = list_of_records.flatMap(partial(map_coocurence, context_size))
        coor_count = coor.reduceByKey(lambda x, y: x + y)
        result = coor_count.map(lambda x: [x[0][0], x[0][1], x[1]])
        tempFile = NamedTemporaryFile(delete=True)
        tempFile.close()
        schema = StructType([
            StructField("Target Word", StringType(), True),
            StructField("Context Word", StringType(), True),
            StructField("Coocurrence Prob", DoubleType(), True)
        ])
        DF = session.createDataFrame(result, schema)

        #DF.show()
        DF.write.csv(os.path.join('output', filename)) 

def test_unionAll_with_udt(self):
        from pyspark.sql.tests import ExamplePoint, ExamplePointUDT
        row1 = (1.0, ExamplePoint(1.0, 2.0))
        row2 = (2.0, ExamplePoint(3.0, 4.0))
        schema = StructType([StructField("label", DoubleType(), False),
                             StructField("point", ExamplePointUDT(), False)])
        df1 = self.sqlCtx.createDataFrame([row1], schema)
        df2 = self.sqlCtx.createDataFrame([row2], schema)

        result = df1.unionAll(df2).orderBy("label").collect()
        self.assertEqual(
            result,
            [
                Row(label=1.0, point=ExamplePoint(1.0, 2.0)),
                Row(label=2.0, point=ExamplePoint(3.0, 4.0))
            ]
        ) 

def test_structs_nested_subset(self):
        schema_has(
            T.StructType([
                T.StructField(
                    'f1',
                    T.ArrayType(T.StructType([
                        T.StructField('f11', T.IntegerType()),
                        T.StructField('f12', T.StringType()),
                    ])),
                ),
            ]),
            T.StructType([
                T.StructField(
                    'f1',
                    T.ArrayType(T.StructType([T.StructField('f11', T.IntegerType())])),
                ),
            ]),
        ) 

def test_maps_nested_subset(self):
        schema_has(
            T.MapType(
                T.StringType(),
                T.MapType(
                    T.StringType(),
                    T.StructType([
                        T.StructField('f1', T.MapType(T.StringType(), T.LongType())),
                        T.StructField('f2', T.MapType(T.StringType(), T.IntegerType())),
                    ]),
                ),
            ),
            T.MapType(
                T.StringType(),
                T.MapType(
                    T.StringType(),
                    T.StructType([
                        T.StructField('f1', T.MapType(T.StringType(), T.LongType())),
                    ]),
                ),
            ),
        ) 

def test_inner_join(self):
        first_df = self.spark.createDataFrame(
            data=[(1, ), (2, ), (3, )],
            schema=T.StructType([T.StructField('id', T.IntegerType())]),
        )
        second_df = self.spark.createDataFrame(
            data=[(2, ), (3, ), (4, )],
            schema=T.StructType([T.StructField('id', T.IntegerType())]),
        )
        third_df = self.spark.createDataFrame(
            data=[(3, ), (4, ), (5, )],
            schema=T.StructType([T.StructField('id', T.IntegerType())]),
        )

        joined_df = SF.multijoin([first_df, second_df, third_df], on='id', how='inner')

        self.assertDataFrameEqual(joined_df, [{'id': 3}]) 

def test_outer_join(self):
        first_df = self.spark.createDataFrame(
            data=[(1, ), (2, ), (3, )],
            schema=T.StructType([T.StructField('id', T.IntegerType())]),
        )
        second_df = self.spark.createDataFrame(
            data=[(2, ), (3, ), (4, )],
            schema=T.StructType([T.StructField('id', T.IntegerType())]),
        )
        third_df = self.spark.createDataFrame(
            data=[(3, ), (4, ), (5, )],
            schema=T.StructType([T.StructField('id', T.IntegerType())]),
        )

        joined_df = SF.multijoin([first_df, second_df, third_df], on='id', how='outer')

        self.assertDataFrameEqual(joined_df, [{'id': i} for i in [1, 2, 3, 4, 5]]) 

def test_coalescing(self):
        first_df = self.spark.createDataFrame(
            data=[(1, None), (2, 'hi'), (3, None), (4, 'may')],
            schema=T.StructType([
                T.StructField('id', T.IntegerType()),
                T.StructField('value', T.StringType()),
            ]),
        )
        second_df = self.spark.createDataFrame(
            data=[(2, 'hey'), (3, 'you'), (4, None)],
            schema=T.StructType([
                T.StructField('id', T.IntegerType()),
                T.StructField('value', T.StringType()),
            ]),
        )

        joined_df = SF.multijoin([first_df, second_df], on='id', how='inner', coalesce=['value'])

        self.assertDataFrameEqual(
            joined_df,
            [{'id': 2, 'value': 'hi'}, {'id': 3, 'value': 'you'}, {'id': 4, 'value': 'may'}],
        ) 

def test_coalescing_light_type_mismatch(self):
        first_df = self.spark.createDataFrame(
            data=[(1, None), (2, 'hi'), (3, None), (4, 'may')],
            schema=T.StructType([
                T.StructField('id', T.IntegerType()),
                T.StructField('value', T.StringType()),
            ]),
        )
        second_df = self.spark.createDataFrame(
            data=[(2, 2), (3, 3), (4, None)],
            schema=T.StructType([
                T.StructField('id', T.IntegerType()),
                T.StructField('value', T.IntegerType()),
            ]),
        )

        joined_df = SF.multijoin([first_df, second_df], on='id', how='inner', coalesce=['value'])

        self.assertDataFrameEqual(
            joined_df,
            [{'id': 2, 'value': 'hi'}, {'id': 3, 'value': '3'}, {'id': 4, 'value': 'may'}],
        ) 

def test_coalescing_heavy_type_mismatch(self):
        first_df = self.spark.createDataFrame(
            data=[(1, None), (2, 'hi'), (3, None), (4, 'may')],
            schema=T.StructType([
                T.StructField('id', T.IntegerType()),
                T.StructField('value', T.StringType()),
            ]),
        )
        second_df = self.spark.createDataFrame(
            data=[(2, [2, ]), (3, [3, ]), (4, None)],
            schema=T.StructType([
                T.StructField('id', T.IntegerType()),
                T.StructField('value', T.ArrayType(T.IntegerType())),
            ]),
        )

        with self.assertRaises(U.AnalysisException):
            SF.multijoin([first_df, second_df], on='id', how='inner', coalesce=['value']) 

def test_default_as_a_lit(self):
        df = self.spark.createDataFrame(
            data=[('one', ), ('two', ), ('three', ), ('hi', )],
            schema=T.StructType([T.StructField('name', T.StringType())]),
        )

        df = df.withColumn('value', SF.switch_case('name', default=0))

        self.assertDataFrameEqual(
            df,
            [
                {'name': 'one', 'value': 0},
                {'name': 'two', 'value': 0},
                {'name': 'three', 'value': 0},
                {'name': 'hi', 'value': 0},
            ],
        ) 

def test_default_as_a_column(self):
        df = self.spark.createDataFrame(
            data=[('one', ), ('two', ), ('three', ), ('hi', )],
            schema=T.StructType([T.StructField('name', T.StringType())]),
        )

        df = df.withColumn('value', SF.switch_case('name', default=F.col('name')))

        self.assertDataFrameEqual(
            df,
            [
                {'name': 'one', 'value': 'one'},
                {'name': 'two', 'value': 'two'},
                {'name': 'three', 'value': 'three'},
                {'name': 'hi', 'value': 'hi'},
            ],
        ) 

def test_switch_as_a_string_cases_as_kwargs(self):
        df = self.spark.createDataFrame(
            data=[('one', ), ('two', ), ('three', ), ('hi', )],
            schema=T.StructType([T.StructField('name', T.StringType())]),
        )

        df = df.withColumn('value', SF.switch_case('name', one=1, two=2, three=3, default=0))

        self.assertDataFrameEqual(
            df,
            [
                {'name': 'one', 'value': 1},
                {'name': 'two', 'value': 2},
                {'name': 'three', 'value': 3},
                {'name': 'hi', 'value': 0},
            ],
        ) 

def test_switch_as_a_column_cases_as_kwargs(self):
        df = self.spark.createDataFrame(
            data=[('one', ), ('two', ), ('three', ), ('hi', )],
            schema=T.StructType([T.StructField('name', T.StringType())]),
        )

        df = df.withColumn(
            'value',
            SF.switch_case(F.col('name'), one=1, two=2, three=3, default=0),
        )

        self.assertDataFrameEqual(
            df,
            [
                {'name': 'one', 'value': 1},
                {'name': 'two', 'value': 2},
                {'name': 'three', 'value': 3},
                {'name': 'hi', 'value': 0},
            ],
        ) 

def test_dict_cases_override_kwarg_cases(self):
        df = self.spark.createDataFrame(
            data=[('one', ), ('two', ), ('three', ), ('hi', )],
            schema=T.StructType([T.StructField('name', T.StringType())]),
        )

        df = df.withColumn(
            'value',
            SF.switch_case('name', {'one': 11, 'three': 33}, one=1, two=2, three=3, default=0),
        )

        self.assertDataFrameEqual(
            df,
            [
                {'name': 'one', 'value': 11},
                {'name': 'two', 'value': 2},
                {'name': 'three', 'value': 33},
                {'name': 'hi', 'value': 0},
            ],
        ) 

def test_redis_client_init(self):
        df = self.spark.createDataFrame(
            data=[
                ('k1', 'k14', [1, 14, 141]),
            ],
            schema=T.StructType([
                T.StructField('key_1', T.StringType()),
                T.StructField('key_2', T.StringType()),
                T.StructField('aux_data', T.ArrayType(T.IntegerType())),
            ])
        )

        df.write_ext.redis(
            key_by=['key_2'],
            max_pipeline_size=3,
            redis_client_init=partial(redis.StrictRedis, 'redis.docker'),
        )

        redis_client = redis.StrictRedis('redis.docker')

        self.assertEqual(redis_client.keys(), [b'k14'])

        written_data = json.loads(redis_client.get('k14'))
        expected = {'key_1': 'k1', 'key_2': 'k14', 'aux_data': [1, 14, 141]}
        self.assertEqual(written_data, expected) 

def test_db(self):
        df = self.spark.createDataFrame(
            data=[
                ('k1', 'k14', [1, 14, 141]),
            ],
            schema=T.StructType([
                T.StructField('key_1', T.StringType()),
                T.StructField('key_2', T.StringType()),
                T.StructField('aux_data', T.ArrayType(T.IntegerType())),
            ])
        )

        df.write_ext.by_url('redis://redis.docker/1?keyBy=key_2&maxPipelineSize=3')

        redis_client = redis.StrictRedis('redis.docker', db=1)

        self.assertEqual(redis_client.keys(), [b'k14'])

        written_data = json.loads(redis_client.get('k14'))
        expected = {'key_1': 'k1', 'key_2': 'k14', 'aux_data': [1, 14, 141]}
        self.assertEqual(written_data, expected) 

def _get_record_store_df_schema():
        """get instance usage schema."""

        columns = ["event_timestamp_string",
                   "event_type", "event_quantity_name",
                   "event_status", "event_version",
                   "record_type", "resource_uuid", "tenant_id",
                   "user_id", "region", "zone",
                   "host", "project_id", "service_group", "service_id",
                   "event_date", "event_hour", "event_minute",
                   "event_second", "metric_group", "metric_id",
                   "namespace", "pod_name", "app", "container_name",
                   "interface", "deployment", "daemon_set"]

        columns_struct_fields = [StructField(field_name, StringType(), True)
                                 for field_name in columns]

        # Add a column for a non-string fields
        columns_struct_fields.insert(0,
                                     StructField("event_timestamp_unix",
                                                 DoubleType(), True))
        columns_struct_fields.insert(0,
                                     StructField("event_quantity",
                                                 DoubleType(), True))

        schema = StructType(columns_struct_fields)

        return schema 

def _get_pre_transform_specs_df_schema():
        """get pre_transform_specs df schema."""

        # FIXME: change when pre_transform_specs df is finalized

        event_type = StructField("event_type", StringType(), True)

        metric_id_list = StructField("metric_id_list",
                                     ArrayType(StringType(),
                                               containsNull=False),
                                     True)
        required_raw_fields_list = StructField("required_raw_fields_list",
                                               ArrayType(StringType(),
                                                         containsNull=False),
                                               True)
        service_id = StructField("service_id", StringType(), True)

        event_processing_params = \
            StructField("event_processing_params",
                        StructType([StructField("set_default_zone_to",
                                                StringType(), True),
                                    StructField("set_default_geolocation_to",
                                                StringType(), True),
                                    StructField("set_default_region_to",
                                                StringType(), True),
                                    ]), True)

        schema = StructType([event_processing_params, event_type,
                             metric_id_list, required_raw_fields_list,
                             service_id])

        return schema 

def _get_grouping_results_df_schema(group_by_column_list):
        """get grouping results schema."""

        group_by_field_list = [StructField(field_name, StringType(), True)
                               for field_name in group_by_column_list]

        # Initialize columns for string fields
        columns = ["firstrecord_timestamp_string",
                   "lastrecord_timestamp_string"]

        columns_struct_fields = [StructField(field_name, StringType(), True)
                                 for field_name in columns]

        # Add columns for non-string fields
        columns_struct_fields.append(StructField("firstrecord_timestamp_unix",
                                                 DoubleType(), True))
        columns_struct_fields.append(StructField("lastrecord_timestamp_unix",
                                                 DoubleType(), True))
        columns_struct_fields.append(StructField("firstrecord_quantity",
                                                 DoubleType(), True))
        columns_struct_fields.append(StructField("lastrecord_quantity",
                                                 DoubleType(), True))
        columns_struct_fields.append(StructField("record_count",
                                                 DoubleType(), True))

        instance_usage_schema_part = StructType(columns_struct_fields)

        grouping_results = \
            StructType([StructField("grouping_key",
                                    StringType(), True),
                        StructField("results",
                                    instance_usage_schema_part,
                                    True),
                        StructField("grouping_key_dict",
                                    StructType(group_by_field_list))])

        # schema = \
        #     StructType([StructField("GroupingResults", grouping_results)])
        return grouping_results 

def test_df_schema(self):
        """Test creation of a data-frame schema"""

        # create reference schema
        ref_sub_schema = StructType([StructField(n, t) for n, t in [('long1', LongType()), ('double1', DoubleType())]])
        ref_cols = [('long', LongType()), ('double', DoubleType()), ('string', StringType()),
                    ('timestamp', TimestampType()), ('struct', ref_sub_schema)]
        ref_schema = StructType([StructField(*c) for c in ref_cols])

        # specification with Python types
        specs = {}
        sub_spec = odict([('long1', int), ('double1', float)])
        specs['python'] = odict([('long', int), ('double', float), ('string', str), ('timestamp', datetime.datetime),
                                 ('struct', sub_spec)])

        # specification with Spark types
        sub_spec = odict([('long1', LongType()), ('double1', DoubleType())])
        specs['spark'] = odict([('long', LongType()), ('double', DoubleType()), ('string', StringType()),
                                ('timestamp', TimestampType()), ('struct', sub_spec)])

        # specification with Spark-type classes
        sub_spec = odict([('long1', LongType), ('double1', DoubleType)])
        specs['spark-class'] = odict([('long', LongType), ('double', DoubleType), ('string', StringType),
                                      ('timestamp', TimestampType), ('struct', sub_spec)])

        # specification mixed types
        sub_spec = odict([('long1', LongType), ('double1', float)])
        specs['mixed'] = odict([('long', int), ('double', DoubleType), ('string', str),
                                ('timestamp', TimestampType()), ('struct', sub_spec)])

        # test with different specifications
        for descr, spec in specs.items():
            schema = df_schema(spec)
            self.assertIs(type(schema), type(ref_schema), 'unexpected schema type for "{}"'.format(descr))
            self.assertTupleEqual(tuple(schema), tuple(ref_schema), 'unexpected schema for "{}"'.format(descr))

        # test with incorrect types
        spec = odict([('long', 'long'), ('double', 'double'), ('string', 'string')])
        with self.assertRaises(TypeError):
            df_schema(spec) 

def test_apply_schema_to_dict_and_rows(self):
        schema = StructType().add("b", StringType()).add("a", IntegerType())
        input = [{"a": 1}, {"b": "coffee"}]
        rdd = self.sc.parallelize(input)
        for verify in [False, True]:
            df = self.spark.createDataFrame(input, schema, verifySchema=verify)
            df2 = self.spark.createDataFrame(rdd, schema, verifySchema=verify)
            self.assertEqual(df.schema, df2.schema)

            rdd = self.sc.parallelize(range(10)).map(lambda x: Row(a=x, b=None))
            df3 = self.spark.createDataFrame(rdd, schema, verifySchema=verify)
            self.assertEqual(10, df3.count())
            input = [Row(a=x, b=str(x)) for x in range(10)]
            df4 = self.spark.createDataFrame(input, schema, verifySchema=verify)
            self.assertEqual(10, df4.count()) 

def test_create_dataframe_schema_mismatch(self):
        input = [Row(a=1)]
        rdd = self.sc.parallelize(range(3)).map(lambda i: Row(a=i))
        schema = StructType([StructField("a", IntegerType()), StructField("b", StringType())])
        df = self.spark.createDataFrame(rdd, schema)
        self.assertRaises(Exception, lambda: df.show()) 

def test_apply_schema(self):
        from datetime import date, datetime
        rdd = self.sc.parallelize([(127, -128, -32768, 32767, 2147483647, 1.0,
                                    date(2010, 1, 1), datetime(2010, 1, 1, 1, 1, 1),
                                    {"a": 1}, (2,), [1, 2, 3], None)])
        schema = StructType([
            StructField("byte1", ByteType(), False),
            StructField("byte2", ByteType(), False),
            StructField("short1", ShortType(), False),
            StructField("short2", ShortType(), False),
            StructField("int1", IntegerType(), False),
            StructField("float1", FloatType(), False),
            StructField("date1", DateType(), False),
            StructField("time1", TimestampType(), False),
            StructField("map1", MapType(StringType(), IntegerType(), False), False),
            StructField("struct1", StructType([StructField("b", ShortType(), False)]), False),
            StructField("list1", ArrayType(ByteType(), False), False),
            StructField("null1", DoubleType(), True)])
        df = self.spark.createDataFrame(rdd, schema)
        results = df.rdd.map(lambda x: (x.byte1, x.byte2, x.short1, x.short2, x.int1, x.float1,
                             x.date1, x.time1, x.map1["a"], x.struct1.b, x.list1, x.null1))
        r = (127, -128, -32768, 32767, 2147483647, 1.0, date(2010, 1, 1),
             datetime(2010, 1, 1, 1, 1, 1), 1, 2, [1, 2, 3], None)
        self.assertEqual(r, results.first())

        df.createOrReplaceTempView("table2")
        r = self.spark.sql("SELECT byte1 - 1 AS byte1, byte2 + 1 AS byte2, " +
                           "short1 + 1 AS short1, short2 - 1 AS short2, int1 - 1 AS int1, " +
                           "float1 + 1.5 as float1 FROM table2").first()

        self.assertEqual((126, -127, -32767, 32766, 2147483646, 2.5), tuple(r))

        from pyspark.sql.types import _parse_schema_abstract, _infer_schema_type
        rdd = self.sc.parallelize([(127, -32768, 1.0, datetime(2010, 1, 1, 1, 1, 1),
                                    {"a": 1}, (2,), [1, 2, 3])])
        abstract = "byte1 short1 float1 time1 map1{} struct1(b) list1[]"
        schema = _parse_schema_abstract(abstract)
        typedSchema = _infer_schema_type(rdd.first(), schema)
        df = self.spark.createDataFrame(rdd, typedSchema)
        r = (127, -32768, 1.0, datetime(2010, 1, 1, 1, 1, 1), {"a": 1}, Row(b=2), [1, 2, 3])
        self.assertEqual(r, tuple(df.first())) 

def test_udt(self):
        from pyspark.sql.types import _parse_datatype_json_string, _infer_type, _verify_type
        from pyspark.sql.tests import ExamplePointUDT, ExamplePoint

        def check_datatype(datatype):
            pickled = pickle.loads(pickle.dumps(datatype))
            assert datatype == pickled
            scala_datatype = self.spark._jsparkSession.parseDataType(datatype.json())
            python_datatype = _parse_datatype_json_string(scala_datatype.json())
            assert datatype == python_datatype

        check_datatype(ExamplePointUDT())
        structtype_with_udt = StructType([StructField("label", DoubleType(), False),
                                          StructField("point", ExamplePointUDT(), False)])
        check_datatype(structtype_with_udt)
        p = ExamplePoint(1.0, 2.0)
        self.assertEqual(_infer_type(p), ExamplePointUDT())
        _verify_type(ExamplePoint(1.0, 2.0), ExamplePointUDT())
        self.assertRaises(ValueError, lambda: _verify_type([1.0, 2.0], ExamplePointUDT()))

        check_datatype(PythonOnlyUDT())
        structtype_with_udt = StructType([StructField("label", DoubleType(), False),
                                          StructField("point", PythonOnlyUDT(), False)])
        check_datatype(structtype_with_udt)
        p = PythonOnlyPoint(1.0, 2.0)
        self.assertEqual(_infer_type(p), PythonOnlyUDT())
        _verify_type(PythonOnlyPoint(1.0, 2.0), PythonOnlyUDT())
        self.assertRaises(ValueError, lambda: _verify_type([1.0, 2.0], PythonOnlyUDT())) 

def test_simple_udt_in_df(self):
        schema = StructType().add("key", LongType()).add("val", PythonOnlyUDT())
        df = self.spark.createDataFrame(
            [(i % 3, PythonOnlyPoint(float(i), float(i))) for i in range(10)],
            schema=schema)
        df.show() 

def test_complex_nested_udt_in_df(self):
        from pyspark.sql.functions import udf

        schema = StructType().add("key", LongType()).add("val", PythonOnlyUDT())
        df = self.spark.createDataFrame(
            [(i % 3, PythonOnlyPoint(float(i), float(i))) for i in range(10)],
            schema=schema)
        df.collect()

        gd = df.groupby("key").agg({"val": "collect_list"})
        gd.collect()
        udf = udf(lambda k, v: [(k, v[0])], ArrayType(df.schema))
        gd.select(udf(*gd)).collect() 

def test_apply_schema_with_udt(self):
        from pyspark.sql.tests import ExamplePoint, ExamplePointUDT
        row = (1.0, ExamplePoint(1.0, 2.0))
        schema = StructType([StructField("label", DoubleType(), False),
                             StructField("point", ExamplePointUDT(), False)])
        df = self.spark.createDataFrame([row], schema)
        point = df.head().point
        self.assertEqual(point, ExamplePoint(1.0, 2.0))

        row = (1.0, PythonOnlyPoint(1.0, 2.0))
        schema = StructType([StructField("label", DoubleType(), False),
                             StructField("point", PythonOnlyUDT(), False)])
        df = self.spark.createDataFrame([row], schema)
        point = df.head().point
        self.assertEqual(point, PythonOnlyPoint(1.0, 2.0)) 

def test_metadata_null(self):
        from pyspark.sql.types import StructType, StringType, StructField
        schema = StructType([StructField("f1", StringType(), True, None),
                             StructField("f2", StringType(), True, {'a': None})])
        rdd = self.sc.parallelize([["a", "b"], ["c", "d"]])
        self.spark.createDataFrame(rdd, schema) 

def test_save_and_load_builder(self):
        df = self.df
        tmpPath = tempfile.mkdtemp()
        shutil.rmtree(tmpPath)
        df.write.json(tmpPath)
        actual = self.spark.read.json(tmpPath)
        self.assertEqual(sorted(df.collect()), sorted(actual.collect()))

        schema = StructType([StructField("value", StringType(), True)])
        actual = self.spark.read.json(tmpPath, schema)
        self.assertEqual(sorted(df.select("value").collect()), sorted(actual.collect()))

        df.write.mode("overwrite").json(tmpPath)
        actual = self.spark.read.json(tmpPath)
        self.assertEqual(sorted(df.collect()), sorted(actual.collect()))

        df.write.mode("overwrite").options(noUse="this options will not be used in save.")\
                .option("noUse", "this option will not be used in save.")\
                .format("json").save(path=tmpPath)
        actual =\
            self.spark.read.format("json")\
                           .load(path=tmpPath, noUse="this options will not be used in load.")
        self.assertEqual(sorted(df.collect()), sorted(actual.collect()))

        defaultDataSourceName = self.spark.conf.get("spark.sql.sources.default",
                                                    "org.apache.spark.sql.parquet")
        self.spark.sql("SET spark.sql.sources.default=org.apache.spark.sql.json")
        actual = self.spark.read.load(path=tmpPath)
        self.assertEqual(sorted(df.collect()), sorted(actual.collect()))
        self.spark.sql("SET spark.sql.sources.default=" + defaultDataSourceName)

        shutil.rmtree(tmpPath) 

def test_stream_read_options(self):
        schema = StructType([StructField("data", StringType(), False)])
        df = self.spark.readStream\
            .format('text')\
            .option('path', 'python/test_support/sql/streaming')\
            .schema(schema)\
            .load()
        self.assertTrue(df.isStreaming)
        self.assertEqual(df.schema.simpleString(), "struct<data:string>") 

def test_stream_read_options_overwrite(self):
        bad_schema = StructType([StructField("test", IntegerType(), False)])
        schema = StructType([StructField("data", StringType(), False)])
        df = self.spark.readStream.format('csv').option('path', 'python/test_support/sql/fake') \
            .schema(bad_schema)\
            .load(path='python/test_support/sql/streaming', schema=schema, format='text')
        self.assertTrue(df.isStreaming)
        self.assertEqual(df.schema.simpleString(), "struct<data:string>") 

def test_column_name_with_non_ascii(self):
        if sys.version >= '3':
            columnName = "??"
            self.assertTrue(isinstance(columnName, str))
        else:
            columnName = unicode("??", "utf-8")
            self.assertTrue(isinstance(columnName, unicode))
        schema = StructType([StructField(columnName, LongType(), True)])
        df = self.spark.createDataFrame([(1,)], schema)
        self.assertEqual(schema, df.schema)
        self.assertEqual("DataFrame[??: bigint]", str(df))
        self.assertEqual([("??", 'bigint')], df.dtypes)
        self.assertEqual(1, df.select("??").first()[0])
        self.assertEqual(1, df.select(df["??"]).first()[0]) 

def test_decimal(self):
        from decimal import Decimal
        schema = StructType([StructField("decimal", DecimalType(10, 5))])
        df = self.spark.createDataFrame([(Decimal("3.14159"),)], schema)
        row = df.select(df.decimal + 1).first()
        self.assertEqual(row[0], Decimal("4.14159"))
        tmpPath = tempfile.mkdtemp()
        shutil.rmtree(tmpPath)
        df.write.parquet(tmpPath)
        df2 = self.spark.read.parquet(tmpPath)
        row = df2.first()
        self.assertEqual(row[0], Decimal("3.14159")) 

def test_fillna(self):
        schema = StructType([
            StructField("name", StringType(), True),
            StructField("age", IntegerType(), True),
            StructField("height", DoubleType(), True)])

        # fillna shouldn't change non-null values
        row = self.spark.createDataFrame([(u'Alice', 10, 80.1)], schema).fillna(50).first()
        self.assertEqual(row.age, 10)

        # fillna with int
        row = self.spark.createDataFrame([(u'Alice', None, None)], schema).fillna(50).first()
        self.assertEqual(row.age, 50)
        self.assertEqual(row.height, 50.0)

        # fillna with double
        row = self.spark.createDataFrame([(u'Alice', None, None)], schema).fillna(50.1).first()
        self.assertEqual(row.age, 50)
        self.assertEqual(row.height, 50.1)

        # fillna with string
        row = self.spark.createDataFrame([(None, None, None)], schema).fillna("hello").first()
        self.assertEqual(row.name, u"hello")
        self.assertEqual(row.age, None)

        # fillna with subset specified for numeric cols
        row = self.spark.createDataFrame(
            [(None, None, None)], schema).fillna(50, subset=['name', 'age']).first()
        self.assertEqual(row.name, None)
        self.assertEqual(row.age, 50)
        self.assertEqual(row.height, None)

        # fillna with subset specified for numeric cols
        row = self.spark.createDataFrame(
            [(None, None, None)], schema).fillna("haha", subset=['name', 'age']).first()
        self.assertEqual(row.name, "haha")
        self.assertEqual(row.age, None)
        self.assertEqual(row.height, None) 

def test_BinaryType_serialization(self):
        # Pyrolite version <= 4.9 could not serialize BinaryType with Python3 SPARK-17808
        schema = StructType([StructField('mybytes', BinaryType())])
        data = [[bytearray(b'here is my data')],
                [bytearray(b'and here is some more')]]
        df = self.spark.createDataFrame(data, schema=schema)
        df.collect() 

def test_save_and_load_table(self):
        df = self.df
        tmpPath = tempfile.mkdtemp()
        shutil.rmtree(tmpPath)
        df.write.saveAsTable("savedJsonTable", "json", "append", path=tmpPath)
        actual = self.spark.createExternalTable("externalJsonTable", tmpPath, "json")
        self.assertEqual(sorted(df.collect()),
                         sorted(self.spark.sql("SELECT * FROM savedJsonTable").collect()))
        self.assertEqual(sorted(df.collect()),
                         sorted(self.spark.sql("SELECT * FROM externalJsonTable").collect()))
        self.assertEqual(sorted(df.collect()), sorted(actual.collect()))
        self.spark.sql("DROP TABLE externalJsonTable")

        df.write.saveAsTable("savedJsonTable", "json", "overwrite", path=tmpPath)
        schema = StructType([StructField("value", StringType(), True)])
        actual = self.spark.createExternalTable("externalJsonTable", source="json",
                                                schema=schema, path=tmpPath,
                                                noUse="this options will not be used")
        self.assertEqual(sorted(df.collect()),
                         sorted(self.spark.sql("SELECT * FROM savedJsonTable").collect()))
        self.assertEqual(sorted(df.select("value").collect()),
                         sorted(self.spark.sql("SELECT * FROM externalJsonTable").collect()))
        self.assertEqual(sorted(df.select("value").collect()), sorted(actual.collect()))
        self.spark.sql("DROP TABLE savedJsonTable")
        self.spark.sql("DROP TABLE externalJsonTable")

        defaultDataSourceName = self.spark.getConf("spark.sql.sources.default",
                                                   "org.apache.spark.sql.parquet")
        self.spark.sql("SET spark.sql.sources.default=org.apache.spark.sql.json")
        df.write.saveAsTable("savedJsonTable", path=tmpPath, mode="overwrite")
        actual = self.spark.createExternalTable("externalJsonTable", path=tmpPath)
        self.assertEqual(sorted(df.collect()),
                         sorted(self.spark.sql("SELECT * FROM savedJsonTable").collect()))
        self.assertEqual(sorted(df.collect()),
                         sorted(self.spark.sql("SELECT * FROM externalJsonTable").collect()))
        self.assertEqual(sorted(df.collect()), sorted(actual.collect()))
        self.spark.sql("DROP TABLE savedJsonTable")
        self.spark.sql("DROP TABLE externalJsonTable")
        self.spark.sql("SET spark.sql.sources.default=" + defaultDataSourceName)

        shutil.rmtree(tmpPath) 

def createExternalTable(self, tableName, path=None, source=None, schema=None, **options):
        """Creates an external table based on the dataset in a data source.

        It returns the DataFrame associated with the external table.

        The data source is specified by the ``source`` and a set of ``options``.
        If ``source`` is not specified, the default data source configured by
        ``spark.sql.sources.default`` will be used.

        Optionally, a schema can be provided as the schema of the returned :class:`DataFrame` and
        created external table.

        :return: :class:`DataFrame`
        """
        if path is not None:
            options["path"] = path
        if source is None:
            source = self._sparkSession.conf.get(
                "spark.sql.sources.default", "org.apache.spark.sql.parquet")
        if schema is None:
            df = self._jcatalog.createExternalTable(tableName, source, options)
        else:
            if not isinstance(schema, StructType):
                raise TypeError("schema should be StructType")
            scala_datatype = self._jsparkSession.parseDataType(schema.json())
            df = self._jcatalog.createExternalTable(tableName, source, scala_datatype, options)
        return DataFrame(df, self._sparkSession._wrapped) 

def __init__(self, predictionAndObservations):
        sc = predictionAndObservations.ctx
        sql_ctx = SQLContext.getOrCreate(sc)
        df = sql_ctx.createDataFrame(predictionAndObservations, schema=StructType([
            StructField("prediction", DoubleType(), nullable=False),
            StructField("observation", DoubleType(), nullable=False)]))
        java_class = sc._jvm.org.apache.spark.mllib.evaluation.RegressionMetrics
        java_model = java_class(df._jdf)
        super(RegressionMetrics, self).__init__(java_model) 

def __init__(self, predictionAndLabels):
        sc = predictionAndLabels.ctx
        sql_ctx = SQLContext.getOrCreate(sc)
        df = sql_ctx.createDataFrame(predictionAndLabels, schema=StructType([
            StructField("prediction", DoubleType(), nullable=False),
            StructField("label", DoubleType(), nullable=False)]))
        java_class = sc._jvm.org.apache.spark.mllib.evaluation.MulticlassMetrics
        java_model = java_class(df._jdf)
        super(MulticlassMetrics, self).__init__(java_model) 

def validate(self, transformations):
        new_fields = []
        for transformation in transformations:
            if isinstance(transformation, FieldTransformation):  # it's transformed name
                if isinstance(transformation.operation, str):  # it's rename
                    renamed_field = self.__get_field(transformation.operation)
                    new_fields.append(types.StructField(transformation.field_name, renamed_field.dataType))
                else:  # is Syntaxtree
                    field_type = self._validate_syntax_tree(transformation.operation)
                    new_fields.append(types.StructField(transformation.field_name, field_type))
            else:  # no transforms
                new_fields.append(self.__get_field(transformation))
        return types.StructType(new_fields)