data source:
class Sum0(AggregateFunction):
def get_value(self, accumulator):
return accumulator[0]
def create_accumulator(self):
return Row(0)
def accumulate(self, accumulator, *args):
if args[0] is not None:
accumulator[0] += args[0]
def retract(self, accumulator, *args):
if args[0] is not None:
accumulator[0] -= args[0]
def merge(self, accumulator, accumulators):
for acc in accumulators:
accumulator[0] += acc[0]
def get_result_type(self):
return "BIGINT"
def get_accumulator_type(self):
return 'ROW<f0 BIGINT>'
ds = env.from_collection(
collection=[(1, 2, "Lee", datetime.now() - timedelta(hours=4)),
(2, 3, "Lee", datetime.now() - timedelta(hours=4)),
(3, 4, "Jay", datetime.now() - timedelta(hours=4)),
(5, 6, "Jay", datetime.now() - timedelta(hours=2)),
(7, 8, "Lee", datetime.now())],
type_info=Types.ROW([Types.INT(),
Types.INT(),
Types.STRING(),
Types.SQL_TIMESTAMP()]))
table_schema = Schema.new_builder() \
.column("f0", "INT") \
.column("f1", "INT") \
.column("f2", "STRING") \
.column_by_expression("rowtime", "CAST(f3 AS TIMESTAMP(3))") \
.watermark("rowtime", "rowtime - INTERVAL '1' SECOND") \
.build()
ts = table_env.from_data_stream(ds, table_schema) \
.alias("value", "count", "name", "rowtime")
this main sql
insert into kafka_sink
select name,sum_udf_agg(value) as agg_data from
TABLE(TUMBLE(TABLE source, DESCRIPTOR(rowtime),INTERVAL '1' HOURS ))
group by window_start, window_end, name
error_detail: pyflink.util.exceptions.TableException: org.apache.flink.table.api.TableException: Table sink 'default_catalog.default_database.kafka_test' doesn't support consuming update changes which is produced by node PythonGroupAggregate(groupBy=[window_start, window_end, name], select=[window_start, window_end, name, sum_udf(value) AS agg_data])
if i use sum in flink system, it will be ok ,I don't understand. Why?
