from pyflink.datastream import StreamExecutionEnvironment
from pyflink.table import StreamTableEnvironment, DataTypes
from pyflink.table.udf import udf
# https://flink.apache.org/2020/04/09/pyflink-udf-support-flink.html
# https://ci.apache.org/projects/flink/flink-docs-release-1.12/dev/python/table-api-users-guide/udfs/python_udfs.html
@udf(input_types=[DataTypes.STRING()],
result_type=DataTypes.MAP(DataTypes.STRING(), DataTypes.STRING()))
def parse(s):
import json
# a dummy parser
res = {}
content = json.loads(s)
if 'item_id' in content:
res['item_id'] = str(
content['item_id']) # REMEMBER to match the result_type
if 'tag' in content:
res['tag'] = content['tag']
return res
env = StreamExecutionEnvironment.get_execution_environment()
t_env = StreamTableEnvironment.create(env)
t_env.register_function("parse", parse)
my_source_ddl = """
create table mySource (
id BIGINT,
def setUp(self):
super(PyFlinkEmbeddedThreadTests, self).setUp()
self.t_env.get_config().set("python.execution-mode", "thread")
class MultiEmit(TableFunction, unittest.TestCase):
def open(self, function_context):
self.counter_sum = 0
def eval(self, x, y):
self.counter_sum += y
for i in range(y):
yield x, i
@udtf(result_types=[DataTypes.BIGINT()])
def identity(x):
if x is not None:
from pyflink.common import Row
return Row(x)
# test specify the input_types
@udtf(input_types=[DataTypes.BIGINT(), DataTypes.BIGINT()],
result_types=DataTypes.BIGINT())
def condition_multi_emit(x, y):
if x == 3:
return range(y, x)
class MultiNum(ScalarFunction):
def register_rides_source(st_env):
st_env \
.connect( # declare the external system to connect to
Kafka()
.version("0.11")
.topic("Rides")
.start_from_earliest()
.property("zookeeper.connect", "zookeeper:2181")
.property("bootstrap.servers", "kafka:9092")) \
.with_format( # declare a format for this system
Json()
.fail_on_missing_field(True)
.schema(DataTypes.ROW([
DataTypes.FIELD("rideId", DataTypes.BIGINT()),
DataTypes.FIELD("isStart", DataTypes.BOOLEAN()),
DataTypes.FIELD("eventTime", DataTypes.TIMESTAMP()),
DataTypes.FIELD("lon", DataTypes.FLOAT()),
DataTypes.FIELD("lat", DataTypes.FLOAT()),
DataTypes.FIELD("psgCnt", DataTypes.INT()),
DataTypes.FIELD("taxiId", DataTypes.BIGINT())]))) \
.with_schema( # declare the schema of the table
Schema()
.field("rideId", DataTypes.BIGINT())
.field("taxiId", DataTypes.BIGINT())
.field("isStart", DataTypes.BOOLEAN())
.field("lon", DataTypes.FLOAT())
.field("lat", DataTypes.FLOAT())
.field("psgCnt", DataTypes.INT())
.field("rideTime", DataTypes.TIMESTAMP())
.rowtime(
Rowtime()
.timestamps_from_field("eventTime")
.watermarks_periodic_bounded(60000))) \
.in_append_mode() \
.register_table_source("source")
]))
t.select("local_zoned_timestamp_func(local_zoned_timestamp_func(a))") \
.insert_into("Results")
self.t_env.execute("test")
actual = source_sink_utils.results()
self.assert_equals(actual, ["1970-01-01T00:00:00.123Z"])
class PyFlinkBlinkBatchUserDefinedFunctionTests(UserDefinedFunctionTests,
PyFlinkBlinkBatchTableTestCase
):
pass
@udf(input_types=[DataTypes.BIGINT(), DataTypes.BIGINT()],
result_type=DataTypes.BIGINT())
def add(i, j):
return i + j
class SubtractOne(ScalarFunction):
def eval(self, i):
return i - 1
class Subtract(ScalarFunction, unittest.TestCase):
def open(self, function_context):
self.subtracted_value = 1
mg = function_context.get_metric_group()
self.counter = mg.add_group("key", "value").counter("my_counter")
def test_scalar_function(self):
# test metric disabled.
self.t_env.get_config().get_configuration().set_string(
'python.metric.enabled', 'false')
# test lambda function
self.t_env.register_function(
"add_one",
udf(lambda i: i + 1, DataTypes.BIGINT(), DataTypes.BIGINT()))
# test Python ScalarFunction
self.t_env.register_function(
"subtract_one",
udf(SubtractOne(), DataTypes.BIGINT(), DataTypes.BIGINT()))
# test Python function
self.t_env.register_function("add", add)
# test callable function
self.t_env.register_function(
"add_one_callable",
udf(CallablePlus(), DataTypes.BIGINT(), DataTypes.BIGINT()))
def partial_func(col, param):
return col + param
# test partial function
import functools
self.t_env.register_function(
"add_one_partial",
udf(functools.partial(partial_func, param=1), DataTypes.BIGINT(),
DataTypes.BIGINT()))
table_sink = source_sink_utils.TestAppendSink(
['a', 'b', 'c', 'd', 'e', 'f'], [
DataTypes.BIGINT(),
DataTypes.BIGINT(),
DataTypes.BIGINT(),
DataTypes.BIGINT(),
DataTypes.BIGINT(),
DataTypes.BIGINT()
])
self.t_env.register_table_sink("Results", table_sink)
t = self.t_env.from_elements([(1, 2, 3), (2, 5, 6), (3, 1, 9)],
['a', 'b', 'c'])
t.where("add_one(b) <= 3") \
.select("add_one(a), subtract_one(b), add(a, c), add_one_callable(a), "
"add_one_partial(a), a") \
.insert_into("Results")
self.t_env.execute("test")
actual = source_sink_utils.results()
self.assert_equals(actual, ["2,1,4,2,2,1", "4,0,12,4,4,3"])
def get_accumulator_type(self):
return DataTypes.ROW([
DataTypes.FIELD(
"f0", DataTypes.MAP(DataTypes.STRING(), DataTypes.STRING())),
DataTypes.FIELD("f1", DataTypes.BIGINT())
])
def get_accumulator_type(self):
return DataTypes.ARRAY(DataTypes.BIGINT())
def test_collect_for_all_data_types(self):
expected_result = [
Row(1, None, 1, True, 32767, -2147483648, 1.23, 1.98932,
bytearray(b'pyflink'), 'pyflink', datetime.date(2014, 9, 13),
datetime.time(12, 0, 0, 123000),
datetime.datetime(2018, 3, 11, 3, 0, 0, 123000),
[Row(['[pyflink]']),
Row(['[pyflink]']),
Row(['[pyflink]'])], {
1: Row(['[flink]']),
2: Row(['[pyflink]'])
}, decimal.Decimal('1000000000000000000.050000000000000000'),
decimal.Decimal('1000000000000000000.059999999999999999'))
]
source = self.t_env.from_elements(
[(1, None, 1, True, 32767, -2147483648, 1.23, 1.98932,
bytearray(b'pyflink'), 'pyflink', datetime.date(2014, 9, 13),
datetime.time(hour=12, minute=0, second=0, microsecond=123000),
datetime.datetime(2018, 3, 11, 3, 0, 0, 123000),
[Row(['pyflink']),
Row(['pyflink']),
Row(['pyflink'])], {
1: Row(['flink']),
2: Row(['pyflink'])
}, decimal.Decimal('1000000000000000000.05'),
decimal.Decimal(
'1000000000000000000.05999999999999999899999999999'))],
DataTypes.ROW([
DataTypes.FIELD("a", DataTypes.BIGINT()),
DataTypes.FIELD("b", DataTypes.BIGINT()),
DataTypes.FIELD("c", DataTypes.TINYINT()),
DataTypes.FIELD("d", DataTypes.BOOLEAN()),
DataTypes.FIELD("e", DataTypes.SMALLINT()),
DataTypes.FIELD("f", DataTypes.INT()),
DataTypes.FIELD("g", DataTypes.FLOAT()),
DataTypes.FIELD("h", DataTypes.DOUBLE()),
DataTypes.FIELD("i", DataTypes.BYTES()),
DataTypes.FIELD("j", DataTypes.STRING()),
DataTypes.FIELD("k", DataTypes.DATE()),
DataTypes.FIELD("l", DataTypes.TIME()),
DataTypes.FIELD("m", DataTypes.TIMESTAMP(3)),
DataTypes.FIELD(
"n",
DataTypes.ARRAY(
DataTypes.ROW(
[DataTypes.FIELD('ss2', DataTypes.STRING())]))),
DataTypes.FIELD(
"o",
DataTypes.MAP(
DataTypes.BIGINT(),
DataTypes.ROW(
[DataTypes.FIELD('ss', DataTypes.STRING())]))),
DataTypes.FIELD("p", DataTypes.DECIMAL(38, 18)),
DataTypes.FIELD("q", DataTypes.DECIMAL(38, 18))
]))
table_result = source.execute()
with table_result.collect() as result:
collected_result = []
for i in result:
collected_result.append(i)
self.assertEqual(expected_result, collected_result)
def test_from_element(self):
t_env = self.t_env
a = array.array('b')
a.fromstring('ABCD')
t = t_env.from_elements([
(1, 1.0, "hi", "hello", datetime.date(1970, 1, 2),
datetime.time(1, 0, 0), datetime.datetime(1970, 1, 2, 0,
0), [1.0, None],
array.array("d",
[1.0, 2.0]), ["abc"], [datetime.date(1970, 1, 2)],
Decimal(1), Row("a", "b")(1, 2.0), {
"key": 1.0
}, a, ExamplePoint(1.0, 2.0), PythonOnlyPoint(3.0, 4.0))
])
field_names = [
"a", "b", "c", "d", "e", "f", "g", "h", "i", "j", "k", "l", "m",
"n", "o", "p", "q"
]
field_types = [
DataTypes.BIGINT(),
DataTypes.DOUBLE(),
DataTypes.STRING(),
DataTypes.STRING(),
DataTypes.DATE(),
DataTypes.TIME(),
DataTypes.TIMESTAMP(),
DataTypes.ARRAY(DataTypes.DOUBLE()),
DataTypes.ARRAY(DataTypes.DOUBLE(False)),
DataTypes.ARRAY(DataTypes.STRING()),
DataTypes.ARRAY(DataTypes.DATE()),
DataTypes.DECIMAL(10, 0),
DataTypes.ROW([
DataTypes.FIELD("a", DataTypes.BIGINT()),
DataTypes.FIELD("b", DataTypes.DOUBLE())
]),
DataTypes.MAP(DataTypes.STRING(), DataTypes.DOUBLE()),
DataTypes.BYTES(),
ExamplePointUDT(),
PythonOnlyUDT()
]
table_sink = source_sink_utils.TestAppendSink(field_names, field_types)
t_env.register_table_sink("Results", table_sink)
t.insert_into("Results")
t_env.exec_env().execute()
actual = source_sink_utils.results()
expected = [
'1,1.0,hi,hello,1970-01-02,01:00:00,1970-01-02 00:00:00.0,[1.0, null],'
'[1.0, 2.0],[abc],[1970-01-02],1,1,2.0,{key=1.0},[65, 66, 67, 68],[1.0, 2.0],'
'[3.0, 4.0]'
]
self.assert_equals(actual, expected)
def test_all_data_types(self):
import pandas as pd
import numpy as np
def tinyint_func(tinyint_param):
assert isinstance(tinyint_param, pd.Series)
assert isinstance(tinyint_param[0], np.int8), \
'tinyint_param of wrong type %s !' % type(tinyint_param[0])
return tinyint_param
def smallint_func(smallint_param):
assert isinstance(smallint_param, pd.Series)
assert isinstance(smallint_param[0], np.int16), \
'smallint_param of wrong type %s !' % type(smallint_param[0])
assert smallint_param[0] == 32767, 'smallint_param of wrong value %s' % smallint_param
return smallint_param
def int_func(int_param):
assert isinstance(int_param, pd.Series)
assert isinstance(int_param[0], np.int32), \
'int_param of wrong type %s !' % type(int_param[0])
assert int_param[0] == -2147483648, 'int_param of wrong value %s' % int_param
return int_param
def bigint_func(bigint_param):
assert isinstance(bigint_param, pd.Series)
assert isinstance(bigint_param[0], np.int64), \
'bigint_param of wrong type %s !' % type(bigint_param[0])
return bigint_param
def boolean_func(boolean_param):
assert isinstance(boolean_param, pd.Series)
assert isinstance(boolean_param[0], np.bool_), \
'boolean_param of wrong type %s !' % type(boolean_param[0])
return boolean_param
def float_func(float_param):
assert isinstance(float_param, pd.Series)
assert isinstance(float_param[0], np.float32), \
'float_param of wrong type %s !' % type(float_param[0])
return float_param
def double_func(double_param):
assert isinstance(double_param, pd.Series)
assert isinstance(double_param[0], np.float64), \
'double_param of wrong type %s !' % type(double_param[0])
return double_param
def varchar_func(varchar_param):
assert isinstance(varchar_param, pd.Series)
assert isinstance(varchar_param[0], str), \
'varchar_param of wrong type %s !' % type(varchar_param[0])
return varchar_param
def varbinary_func(varbinary_param):
assert isinstance(varbinary_param, pd.Series)
assert isinstance(varbinary_param[0], bytes), \
'varbinary_param of wrong type %s !' % type(varbinary_param[0])
return varbinary_param
def decimal_func(decimal_param):
assert isinstance(decimal_param, pd.Series)
assert isinstance(decimal_param[0], decimal.Decimal), \
'decimal_param of wrong type %s !' % type(decimal_param[0])
return decimal_param
def date_func(date_param):
assert isinstance(date_param, pd.Series)
assert isinstance(date_param[0], datetime.date), \
'date_param of wrong type %s !' % type(date_param[0])
return date_param
def time_func(time_param):
assert isinstance(time_param, pd.Series)
assert isinstance(time_param[0], datetime.time), \
'time_param of wrong type %s !' % type(time_param[0])
return time_param
timestamp_value = datetime.datetime(1970, 1, 2, 0, 0, 0, 123000)
def timestamp_func(timestamp_param):
assert isinstance(timestamp_param, pd.Series)
assert isinstance(timestamp_param[0], datetime.datetime), \
'timestamp_param of wrong type %s !' % type(timestamp_param[0])
assert timestamp_param[0] == timestamp_value, \
'timestamp_param is wrong value %s, should be %s!' % (timestamp_param[0],
timestamp_value)
return timestamp_param
def array_func(array_param):
assert isinstance(array_param, pd.Series)
assert isinstance(array_param[0], np.ndarray), \
'array_param of wrong type %s !' % type(array_param[0])
return array_param
def nested_array_func(nested_array_param):
assert isinstance(nested_array_param, pd.Series)
assert isinstance(nested_array_param[0], np.ndarray), \
'nested_array_param of wrong type %s !' % type(nested_array_param[0])
return pd.Series(nested_array_param[0])
def row_func(row_param):
assert isinstance(row_param, pd.Series)
assert isinstance(row_param[0], dict), \
'row_param of wrong type %s !' % type(row_param[0])
return row_param
self.t_env.create_temporary_system_function(
"tinyint_func",
udf(tinyint_func, result_type=DataTypes.TINYINT(), udf_type="pandas"))
self.t_env.create_temporary_system_function(
"smallint_func",
udf(smallint_func, result_type=DataTypes.SMALLINT(), udf_type="pandas"))
self.t_env.create_temporary_system_function(
"int_func",
udf(int_func, result_type=DataTypes.INT(), udf_type="pandas"))
self.t_env.create_temporary_system_function(
"bigint_func",
udf(bigint_func, result_type=DataTypes.BIGINT(), udf_type="pandas"))
self.t_env.create_temporary_system_function(
"boolean_func",
udf(boolean_func, result_type=DataTypes.BOOLEAN(), udf_type="pandas"))
self.t_env.create_temporary_system_function(
"float_func",
udf(float_func, result_type=DataTypes.FLOAT(), udf_type="pandas"))
self.t_env.create_temporary_system_function(
"double_func",
udf(double_func, result_type=DataTypes.DOUBLE(), udf_type="pandas"))
self.t_env.create_temporary_system_function(
"varchar_func",
udf(varchar_func, result_type=DataTypes.STRING(), udf_type="pandas"))
self.t_env.create_temporary_system_function(
"varbinary_func",
udf(varbinary_func, result_type=DataTypes.BYTES(), udf_type="pandas"))
self.t_env.register_function(
"decimal_func",
udf(decimal_func, result_type=DataTypes.DECIMAL(38, 18), udf_type="pandas"))
self.t_env.create_temporary_system_function(
"date_func",
udf(date_func, result_type=DataTypes.DATE(), udf_type="pandas"))
self.t_env.create_temporary_system_function(
"time_func",
udf(time_func, result_type=DataTypes.TIME(), udf_type="pandas"))
self.t_env.create_temporary_system_function(
"timestamp_func",
udf(timestamp_func, result_type=DataTypes.TIMESTAMP(3), udf_type="pandas"))
self.t_env.create_temporary_system_function(
"array_str_func",
udf(array_func, result_type=DataTypes.ARRAY(DataTypes.STRING()), udf_type="pandas"))
self.t_env.create_temporary_system_function(
"array_timestamp_func",
udf(array_func, result_type=DataTypes.ARRAY(DataTypes.TIMESTAMP(3)), udf_type="pandas"))
self.t_env.create_temporary_system_function(
"array_int_func",
udf(array_func, result_type=DataTypes.ARRAY(DataTypes.INT()), udf_type="pandas"))
self.t_env.create_temporary_system_function(
"nested_array_func",
udf(nested_array_func,
result_type=DataTypes.ARRAY(DataTypes.STRING()), udf_type="pandas"))
row_type = DataTypes.ROW(
[DataTypes.FIELD("f1", DataTypes.INT()),
DataTypes.FIELD("f2", DataTypes.STRING()),
DataTypes.FIELD("f3", DataTypes.TIMESTAMP(3)),
DataTypes.FIELD("f4", DataTypes.ARRAY(DataTypes.INT()))])
self.t_env.create_temporary_system_function(
"row_func",
udf(row_func, result_type=row_type, udf_type="pandas"))
table_sink = source_sink_utils.TestAppendSink(
['a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q',
'r', 's', 't', 'u'],
[DataTypes.TINYINT(), DataTypes.SMALLINT(), DataTypes.INT(), DataTypes.BIGINT(),
DataTypes.BOOLEAN(), DataTypes.BOOLEAN(), DataTypes.FLOAT(), DataTypes.DOUBLE(),
DataTypes.STRING(), DataTypes.STRING(), DataTypes.BYTES(), DataTypes.DECIMAL(38, 18),
DataTypes.DECIMAL(38, 18), DataTypes.DATE(), DataTypes.TIME(), DataTypes.TIMESTAMP(3),
DataTypes.ARRAY(DataTypes.STRING()), DataTypes.ARRAY(DataTypes.TIMESTAMP(3)),
DataTypes.ARRAY(DataTypes.INT()),
DataTypes.ARRAY(DataTypes.STRING()), row_type])
self.t_env.register_table_sink("Results", table_sink)
t = self.t_env.from_elements(
[(1, 32767, -2147483648, 1, True, False, 1.0, 1.0, 'hello', '中文',
bytearray(b'flink'), decimal.Decimal('1000000000000000000.05'),
decimal.Decimal('1000000000000000000.05999999999999999899999999999'),
datetime.date(2014, 9, 13), datetime.time(hour=1, minute=0, second=1),
timestamp_value, ['hello', '中文', None], [timestamp_value], [1, 2],
[['hello', '中文', None]], Row(1, 'hello', timestamp_value, [1, 2]))],
DataTypes.ROW(
[DataTypes.FIELD("a", DataTypes.TINYINT()),
DataTypes.FIELD("b", DataTypes.SMALLINT()),
DataTypes.FIELD("c", DataTypes.INT()),
DataTypes.FIELD("d", DataTypes.BIGINT()),
DataTypes.FIELD("e", DataTypes.BOOLEAN()),
DataTypes.FIELD("f", DataTypes.BOOLEAN()),
DataTypes.FIELD("g", DataTypes.FLOAT()),
DataTypes.FIELD("h", DataTypes.DOUBLE()),
DataTypes.FIELD("i", DataTypes.STRING()),
DataTypes.FIELD("j", DataTypes.STRING()),
DataTypes.FIELD("k", DataTypes.BYTES()),
DataTypes.FIELD("l", DataTypes.DECIMAL(38, 18)),
DataTypes.FIELD("m", DataTypes.DECIMAL(38, 18)),
DataTypes.FIELD("n", DataTypes.DATE()),
DataTypes.FIELD("o", DataTypes.TIME()),
DataTypes.FIELD("p", DataTypes.TIMESTAMP(3)),
DataTypes.FIELD("q", DataTypes.ARRAY(DataTypes.STRING())),
DataTypes.FIELD("r", DataTypes.ARRAY(DataTypes.TIMESTAMP(3))),
DataTypes.FIELD("s", DataTypes.ARRAY(DataTypes.INT())),
DataTypes.FIELD("t", DataTypes.ARRAY(DataTypes.ARRAY(DataTypes.STRING()))),
DataTypes.FIELD("u", row_type)]))
exec_insert_table(t.select("tinyint_func(a),"
"smallint_func(b),"
"int_func(c),"
"bigint_func(d),"
"boolean_func(e),"
"boolean_func(f),"
"float_func(g),"
"double_func(h),"
"varchar_func(i),"
"varchar_func(j),"
"varbinary_func(k),"
"decimal_func(l),"
"decimal_func(m),"
"date_func(n),"
"time_func(o),"
"timestamp_func(p),"
"array_str_func(q),"
"array_timestamp_func(r),"
"array_int_func(s),"
"nested_array_func(t),"
"row_func(u)"), "Results")
actual = source_sink_utils.results()
self.assert_equals(actual,
["1,32767,-2147483648,1,true,false,1.0,1.0,hello,中文,"
"[102, 108, 105, 110, 107],1000000000000000000.050000000000000000,"
"1000000000000000000.059999999999999999,2014-09-13,01:00:01,"
"1970-01-02 00:00:00.123,[hello, 中文, null],[1970-01-02 00:00:00.123],"
"[1, 2],[hello, 中文, null],1,hello,1970-01-02 00:00:00.123,[1, 2]"])
from pyflink.table import AggregateFunction, DataTypes
from pyflink.table.udf import udaf
class WeightedAvg(AggregateFunction):
def create_accumulator(self):
# Row(sum, count)
return Row(0, 0)
def get_value(self, accumulator: Row) -> float:
if accumulator[1] == 0:
return 0
else:
return accumulator[0] / accumulator[1]
def accumulate(self, accumulator: Row, value, weight):
accumulator[0] += value * weight
accumulator[1] += weight
def retract(self, accumulator: Row, value, weight):
accumulator[0] -= value * weight
accumulator[1] -= weight
weighted_avg = udaf(f=WeightedAvg(),
result_type=DataTypes.DOUBLE(),
accumulator_type=DataTypes.ROW([
DataTypes.FIELD("f0", DataTypes.BIGINT()),
DataTypes.FIELD("f1", DataTypes.BIGINT())
]))
def test_non_exist_udf_type(self):
with self.assertRaisesRegex(ValueError,
'The udf_type must be one of \'general, pandas\''):
udf(lambda i: i + 1, result_type=DataTypes.BIGINT(), udf_type="non-exist")
result = self.collect(t)
self.assert_equals(result, ["1,3,6,3", "3,2,14,5"])
class BlinkBatchPandasUDFITTests(PandasUDFITTests,
BlinkPandasUDFITTests,
PyFlinkBlinkBatchTableTestCase):
pass
class BlinkStreamPandasUDFITTests(PandasUDFITTests,
BlinkPandasUDFITTests,
PyFlinkBlinkStreamTableTestCase):
pass
@udf(result_type=DataTypes.BIGINT(), udf_type='pandas')
def add(i, j):
return i + j
if __name__ == '__main__':
import unittest
try:
import xmlrunner
testRunner = xmlrunner.XMLTestRunner(output='target/test-reports')
except ImportError:
testRunner = None
unittest.main(testRunner=testRunner, verbosity=2)
def pv_uv_demo():
s_env = StreamExecutionEnvironment.get_execution_environment()
s_env.set_stream_time_characteristic(TimeCharacteristic.EventTime)
s_env.set_parallelism(1)
# use blink table planner
st_env = StreamTableEnvironment.create(
s_env,
environment_settings=EnvironmentSettings.new_instance(
).in_streaming_mode().use_blink_planner().build())
# use flink table planner
# st_env = StreamTableEnvironment.create(s_env)
st_env \
.connect( # declare the external system to connect to
Kafka()
.version("0.11")
.topic("user_behavior")
.start_from_earliest()
.property("zookeeper.connect", "localhost:2181")
.property("bootstrap.servers", "localhost:9092")
) \
.with_format( # declare a format for this system
Json()
.fail_on_missing_field(True)
.json_schema(
"{"
" type: 'object',"
" properties: {"
" user_id: {"
" type: 'string'"
" },"
" item_id: {"
" type: 'string'"
" },"
" category_id: {"
" type: 'string'"
" },"
" behavior: {"
" type: 'string'"
" },"
" ts: {"
" type: 'string',"
" format: 'date-time'"
" }"
" }"
"}"
)
) \
.with_schema( # declare the schema of the table
Schema()
.field("user_id", DataTypes.STRING())
.field("item_id", DataTypes.STRING())
.field("category_id", DataTypes.STRING())
.field("behavior", DataTypes.STRING())
.field("rowtime", DataTypes.TIMESTAMP())
.rowtime(
Rowtime()
.timestamps_from_field("ts")
.watermarks_periodic_bounded(60000))
) \
.in_append_mode() \
.register_table_source("source")
# use custom retract sink connector
custom_connector = CustomConnectorDescriptor('jdbc', 1, False) \
.property("connector.driver", "org.apache.derby.jdbc.ClientDriver") \
.property("connector.url", "jdbc:derby://localhost:1527/firstdb") \
.property("connector.table", "pv_uv_table") \
.property("connector.write.flush.max-rows", "1")
st_env.connect(custom_connector) \
.with_schema(
Schema()
.field("startTime", DataTypes.TIMESTAMP())
.field("endTime", DataTypes.TIMESTAMP())
.field("pv", DataTypes.BIGINT())
.field("uv", DataTypes.BIGINT())
).register_table_sink("sink")
st_env.scan("source").window(Tumble.over("1.hours").on("rowtime").alias("w")) \
.group_by("w") \
.select("w.start as startTime, w.end as endTime, COUNT(1) as pv, user_id.count.distinct as uv").insert_into("sink")
st_env.execute("table pv uv")
def test_expression(self):
expr1 = col('a')
expr2 = col('b')
expr3 = col('c')
expr4 = col('d')
expr5 = lit(10)
# comparison functions
self.assertEqual('equals(a, b)', str(expr1 == expr2))
self.assertEqual('mod(2, b)', str(2 % expr2))
self.assertEqual('notEquals(a, b)', str(expr1 != expr2))
self.assertEqual('lessThan(a, b)', str(expr1 < expr2))
self.assertEqual('lessThanOrEqual(a, b)', str(expr1 <= expr2))
self.assertEqual('greaterThan(a, b)', str(expr1 > expr2))
self.assertEqual('greaterThanOrEqual(a, b)', str(expr1 >= expr2))
# logic functions
self.assertEqual('and(a, b)', str(expr1 & expr2))
self.assertEqual('or(a, b)', str(expr1 | expr2))
self.assertEqual('isNotTrue(a)', str(expr1.is_not_true))
self.assertEqual('isNotTrue(a)', str(~expr1))
# arithmetic functions
self.assertEqual('plus(a, b)', str(expr1 + expr2))
self.assertEqual('plus(2, b)', str(2 + expr2))
self.assertEqual('plus(cast(b, DATE), 2)', str(expr2.to_date + 2))
self.assertEqual('minus(a, b)', str(expr1 - expr2))
self.assertEqual('minus(cast(b, DATE), 2)', str(expr2.to_date - 2))
self.assertEqual('times(a, b)', str(expr1 * expr2))
self.assertEqual('divide(a, b)', str(expr1 / expr2))
self.assertEqual('mod(a, b)', str(expr1 % expr2))
self.assertEqual('power(a, b)', str(expr1**expr2))
self.assertEqual('minusPrefix(a)', str(-expr1))
self.assertEqual('exp(a)', str(expr1.exp))
self.assertEqual('log10(a)', str(expr1.log10))
self.assertEqual('log2(a)', str(expr1.log2))
self.assertEqual('ln(a)', str(expr1.ln))
self.assertEqual('log(a)', str(expr1.log()))
self.assertEqual('cosh(a)', str(expr1.cosh))
self.assertEqual('sinh(a)', str(expr1.sinh))
self.assertEqual('sin(a)', str(expr1.sin))
self.assertEqual('cos(a)', str(expr1.cos))
self.assertEqual('tan(a)', str(expr1.tan))
self.assertEqual('cot(a)', str(expr1.cot))
self.assertEqual('asin(a)', str(expr1.asin))
self.assertEqual('acos(a)', str(expr1.acos))
self.assertEqual('atan(a)', str(expr1.atan))
self.assertEqual('tanh(a)', str(expr1.tanh))
self.assertEqual('degrees(a)', str(expr1.degrees))
self.assertEqual('radians(a)', str(expr1.radians))
self.assertEqual('sqrt(a)', str(expr1.sqrt))
self.assertEqual('abs(a)', str(expr1.abs))
self.assertEqual('abs(a)', str(abs(expr1)))
self.assertEqual('sign(a)', str(expr1.sign))
self.assertEqual('round(a, b)', str(expr1.round(expr2)))
self.assertEqual('between(a, b, c)', str(expr1.between(expr2, expr3)))
self.assertEqual('notBetween(a, b, c)',
str(expr1.not_between(expr2, expr3)))
self.assertEqual('ifThenElse(a, b, c)', str(expr1.then(expr2, expr3)))
self.assertEqual('isNull(a)', str(expr1.is_null))
self.assertEqual('isNotNull(a)', str(expr1.is_not_null))
self.assertEqual('isTrue(a)', str(expr1.is_true))
self.assertEqual('isFalse(a)', str(expr1.is_false))
self.assertEqual('isNotTrue(a)', str(expr1.is_not_true))
self.assertEqual('isNotFalse(a)', str(expr1.is_not_false))
self.assertEqual('distinct(a)', str(expr1.distinct))
self.assertEqual('sum(a)', str(expr1.sum))
self.assertEqual('sum0(a)', str(expr1.sum0))
self.assertEqual('min(a)', str(expr1.min))
self.assertEqual('max(a)', str(expr1.max))
self.assertEqual('count(a)', str(expr1.count))
self.assertEqual('avg(a)', str(expr1.avg))
self.assertEqual('stddevPop(a)', str(expr1.stddev_pop))
self.assertEqual('stddevSamp(a)', str(expr1.stddev_samp))
self.assertEqual('varPop(a)', str(expr1.var_pop))
self.assertEqual('varSamp(a)', str(expr1.var_samp))
self.assertEqual('collect(a)', str(expr1.collect))
self.assertEqual("as(a, 'a', 'b', 'c')",
str(expr1.alias('a', 'b', 'c')))
self.assertEqual('cast(a, INT)', str(expr1.cast(DataTypes.INT())))
self.assertEqual('asc(a)', str(expr1.asc))
self.assertEqual('desc(a)', str(expr1.desc))
self.assertEqual('in(a, b, c, d)', str(expr1.in_(expr2, expr3, expr4)))
self.assertEqual('start(a)', str(expr1.start))
self.assertEqual('end(a)', str(expr1.end))
self.assertEqual('bin(a)', str(expr1.bin))
self.assertEqual('hex(a)', str(expr1.hex))
self.assertEqual('truncate(a, 3)', str(expr1.truncate(3)))
# string functions
self.assertEqual('substring(a, b, 3)', str(expr1.substring(expr2, 3)))
self.assertEqual("trim(true, false, ' ', a)",
str(expr1.trim_leading()))
self.assertEqual("trim(false, true, ' ', a)",
str(expr1.trim_trailing()))
self.assertEqual("trim(true, true, ' ', a)", str(expr1.trim()))
self.assertEqual('replace(a, b, c)', str(expr1.replace(expr2, expr3)))
self.assertEqual('charLength(a)', str(expr1.char_length))
self.assertEqual('upper(a)', str(expr1.upper_case))
self.assertEqual('lower(a)', str(expr1.lower_case))
self.assertEqual('initCap(a)', str(expr1.init_cap))
self.assertEqual("like(a, 'Jo_n%')", str(expr1.like('Jo_n%')))
self.assertEqual("similar(a, 'A+')", str(expr1.similar('A+')))
self.assertEqual('position(a, b)', str(expr1.position(expr2)))
self.assertEqual('lpad(a, 4, b)', str(expr1.lpad(4, expr2)))
self.assertEqual('rpad(a, 4, b)', str(expr1.rpad(4, expr2)))
self.assertEqual('overlay(a, b, 6, 2)', str(expr1.overlay(expr2, 6,
2)))
self.assertEqual("regexpReplace(a, b, 'abc')",
str(expr1.regexp_replace(expr2, 'abc')))
self.assertEqual('regexpExtract(a, b, 3)',
str(expr1.regexp_extract(expr2, 3)))
self.assertEqual('fromBase64(a)', str(expr1.from_base64))
self.assertEqual('toBase64(a)', str(expr1.to_base64))
self.assertEqual('ltrim(a)', str(expr1.ltrim))
self.assertEqual('rtrim(a)', str(expr1.rtrim))
self.assertEqual('repeat(a, 3)', str(expr1.repeat(3)))
self.assertEqual("over(a, 'w')", str(expr1.over('w')))
# temporal functions
self.assertEqual('cast(a, DATE)', str(expr1.to_date))
self.assertEqual('cast(a, TIME(0))', str(expr1.to_time))
self.assertEqual('cast(a, TIMESTAMP(3))', str(expr1.to_timestamp))
self.assertEqual('extract(YEAR, a)',
str(expr1.extract(TimeIntervalUnit.YEAR)))
self.assertEqual('floor(a, YEAR)',
str(expr1.floor(TimeIntervalUnit.YEAR)))
self.assertEqual('ceil(a)', str(expr1.ceil()))
# advanced type helper functions
self.assertEqual("get(a, 'col')", str(expr1.get('col')))
self.assertEqual('flatten(a)', str(expr1.flatten))
self.assertEqual('at(a, 0)', str(expr1.at(0)))
self.assertEqual('cardinality(a)', str(expr1.cardinality))
self.assertEqual('element(a)', str(expr1.element))
# time definition functions
self.assertEqual('rowtime(a)', str(expr1.rowtime))
self.assertEqual('proctime(a)', str(expr1.proctime))
self.assertEqual('120', str(expr5.year))
self.assertEqual('120', str(expr5.years))
self.assertEqual('30', str(expr5.quarter))
self.assertEqual('30', str(expr5.quarters))
self.assertEqual('10', str(expr5.month))
self.assertEqual('10', str(expr5.months))
self.assertEqual('6048000000', str(expr5.week))
self.assertEqual('6048000000', str(expr5.weeks))
self.assertEqual('864000000', str(expr5.day))
self.assertEqual('864000000', str(expr5.days))
self.assertEqual('36000000', str(expr5.hour))
self.assertEqual('36000000', str(expr5.hours))
self.assertEqual('600000', str(expr5.minute))
self.assertEqual('600000', str(expr5.minutes))
self.assertEqual('10000', str(expr5.second))
self.assertEqual('10000', str(expr5.seconds))
self.assertEqual('10', str(expr5.milli))
self.assertEqual('10', str(expr5.millis))
# hash functions
self.assertEqual('md5(a)', str(expr1.md5))
self.assertEqual('sha1(a)', str(expr1.sha1))
self.assertEqual('sha224(a)', str(expr1.sha224))
self.assertEqual('sha256(a)', str(expr1.sha256))
self.assertEqual('sha384(a)', str(expr1.sha384))
self.assertEqual('sha512(a)', str(expr1.sha512))
self.assertEqual('sha2(a, 224)', str(expr1.sha2(224)))
# json functions
self.assertEqual("IS_JSON('42')", str(lit('42').is_json()))
self.assertEqual("IS_JSON('42', SCALAR)",
str(lit('42').is_json(JsonType.SCALAR)))
self.assertEqual("JSON_EXISTS('{}', '$.x')",
str(lit('{}').json_exists('$.x')))
self.assertEqual(
"JSON_EXISTS('{}', '$.x', FALSE)",
str(lit('{}').json_exists('$.x', JsonExistsOnError.FALSE)))
self.assertEqual(
"JSON_VALUE('{}', '$.x', STRING, NULL, null, NULL, null)",
str(lit('{}').json_value('$.x')))
self.assertEqual(
"JSON_VALUE('{}', '$.x', INT, DEFAULT, 42, ERROR, null)",
str(
lit('{}').json_value('$.x', DataTypes.INT(),
JsonValueOnEmptyOrError.DEFAULT, 42,
JsonValueOnEmptyOrError.ERROR, None)))
self.assertEqual(
"JSON_QUERY('{}', '$.x', WITHOUT_ARRAY, NULL, EMPTY_ARRAY)",
str(
lit('{}').json_query('$.x', JsonQueryWrapper.WITHOUT_ARRAY,
JsonQueryOnEmptyOrError.NULL,
JsonQueryOnEmptyOrError.EMPTY_ARRAY)))
from pyflink.dataset import ExecutionEnvironment
from pyflink.table import TableConfig, DataTypes, BatchTableEnvironment
from pyflink.table.descriptors import Schema, OldCsv, FileSystem
exec_env = ExecutionEnvironment.get_execution_environment()
exec_env.set_parallelism(2)
t_config = TableConfig()
t_env = BatchTableEnvironment.create(exec_env, t_config)
t_env.connect(FileSystem().path('input')) \
.with_format(OldCsv()
.line_delimiter(' ')
.field('word', DataTypes.STRING())) \
.with_schema(Schema()
.field('word', DataTypes.STRING())) \
.register_table_source("inputSource")
t_env.connect(FileSystem().path('output')) \
.with_format(OldCsv().field_delimiter(',').field('word', DataTypes.STRING()).field('count', DataTypes.BIGINT()))\
.with_schema(Schema().field('word', DataTypes.STRING()).field('count', DataTypes.BIGINT()))\
.register_table_sink('sink')
t_env.scan('inputSource').group_by('word').select('word, count(1)').insert_into('sink')
t_env.execute('my first job')
def get_result_type(self):
return DataTypes.STRING()
def get_accumulator_type(self):
return DataTypes.ROW([DataTypes.FIELD("f0", DataTypes.LIST_VIEW(DataTypes.STRING()))])
def test_sliding_group_window_over_time(self):
# create source file path
tmp_dir = self.tempdir
data = [
'1,1,2,2018-03-11 03:10:00',
'3,3,2,2018-03-11 03:10:00',
'2,2,1,2018-03-11 03:10:00',
'2,2,1,2018-03-11 03:30:00',
'1,1,3,2018-03-11 03:40:00',
'1,1,8,2018-03-11 04:20:00',
]
source_path = tmp_dir + '/test_sliding_group_window_over_time.csv'
with open(source_path, 'w') as fd:
for ele in data:
fd.write(ele + '\n')
self.t_env.create_temporary_system_function("my_sum",
SumAggregateFunction())
source_table = """
create table source_table(
a TINYINT,
b SMALLINT,
c INT,
rowtime TIMESTAMP(3),
WATERMARK FOR rowtime AS rowtime - INTERVAL '60' MINUTE
) with(
'connector.type' = 'filesystem',
'format.type' = 'csv',
'connector.path' = '%s',
'format.ignore-first-line' = 'false',
'format.field-delimiter' = ','
)
""" % source_path
self.t_env.execute_sql(source_table)
t = self.t_env.from_path("source_table")
from pyflink.testing import source_sink_utils
table_sink = source_sink_utils.TestAppendSink(['a', 'b', 'c', 'd'], [
DataTypes.TINYINT(),
DataTypes.TIMESTAMP(3),
DataTypes.TIMESTAMP(3),
DataTypes.BIGINT()
])
self.t_env.register_table_sink("Results", table_sink)
t.window(Slide.over(lit(1).hours)
.every(lit(30).minutes)
.on(t.rowtime)
.alias("w")) \
.group_by(t.a, col("w")) \
.select(t.a, col("w").start, col("w").end, call("my_sum", t.c).alias("c")) \
.execute_insert("Results") \
.wait()
actual = source_sink_utils.results()
self.assert_equals(actual, [
"+I[1, 2018-03-11 02:30:00.0, 2018-03-11 03:30:00.0, 2]",
"+I[2, 2018-03-11 02:30:00.0, 2018-03-11 03:30:00.0, 1]",
"+I[3, 2018-03-11 02:30:00.0, 2018-03-11 03:30:00.0, 2]",
"+I[1, 2018-03-11 03:00:00.0, 2018-03-11 04:00:00.0, 5]",
"+I[3, 2018-03-11 03:00:00.0, 2018-03-11 04:00:00.0, 2]",
"+I[2, 2018-03-11 03:00:00.0, 2018-03-11 04:00:00.0, 2]",
"+I[2, 2018-03-11 03:30:00.0, 2018-03-11 04:30:00.0, 1]",
"+I[1, 2018-03-11 03:30:00.0, 2018-03-11 04:30:00.0, 11]",
"+I[1, 2018-03-11 04:00:00.0, 2018-03-11 05:00:00.0, 8]"
])
def create_another_table_schema():
return TableSchema(["first2", "second", "third"],
[DataTypes.STRING(), DataTypes.STRING(), DataTypes.STRING()])
def get_result_type(self):
return DataTypes.BIGINT()
def test_scalar_function(self):
# test metric disabled.
self.t_env.get_config().get_configuration().set_string(
'python.metric.enabled', 'false')
# test lambda function
add_one = udf(lambda i: i + 1, result_type=DataTypes.BIGINT())
# test Python ScalarFunction
subtract_one = udf(SubtractOne(), result_type=DataTypes.BIGINT())
# test callable function
add_one_callable = udf(CallablePlus(), result_type=DataTypes.BIGINT())
def partial_func(col, param):
return col + param
# test partial function
import functools
add_one_partial = udf(functools.partial(partial_func, param=1),
result_type=DataTypes.BIGINT())
# check memory limit is set
@udf(result_type=DataTypes.BIGINT())
def check_memory_limit(exec_mode):
if exec_mode == "process":
assert os.environ['_PYTHON_WORKER_MEMORY_LIMIT'] is not None
return 1
table_sink = source_sink_utils.TestAppendSink(
['a', 'b', 'c', 'd', 'e', 'f', 'g'], [
DataTypes.BIGINT(),
DataTypes.BIGINT(),
DataTypes.BIGINT(),
DataTypes.BIGINT(),
DataTypes.BIGINT(),
DataTypes.BIGINT(),
DataTypes.BIGINT()
])
self.t_env.register_table_sink("Results", table_sink)
execution_mode = self.t_env.get_config().get_configuration(
).get_string("python.execution-mode", "process")
t = self.t_env.from_elements([(1, 2, 3), (2, 5, 6), (3, 1, 9)],
['a', 'b', 'c'])
t.where(add_one(t.b) <= 3).select(
add_one(t.a), subtract_one(t.b), add(t.a, t.c), add_one_callable(t.a),
add_one_partial(t.a), check_memory_limit(execution_mode), t.a) \
.execute_insert("Results").wait()
actual = source_sink_utils.results()
self.assert_equals(
actual, ["+I[2, 1, 4, 2, 2, 1, 1]", "+I[4, 0, 12, 4, 4, 1, 3]"])
def test_udf_with_constant_params(self):
def udf_with_constant_params(p, null_param, tinyint_param,
smallint_param, int_param, bigint_param,
decimal_param, float_param, double_param,
boolean_param, str_param, date_param,
time_param, timestamp_param):
from decimal import Decimal
import datetime
assert null_param is None, 'null_param is wrong value %s' % null_param
assert isinstance(tinyint_param, int), 'tinyint_param of wrong type %s !' \
% type(tinyint_param)
p += tinyint_param
assert isinstance(smallint_param, int), 'smallint_param of wrong type %s !' \
% type(smallint_param)
p += smallint_param
assert isinstance(int_param, int), 'int_param of wrong type %s !' \
% type(int_param)
p += int_param
assert isinstance(bigint_param, int), 'bigint_param of wrong type %s !' \
% type(bigint_param)
p += bigint_param
assert decimal_param == Decimal('1.05'), \
'decimal_param is wrong value %s ' % decimal_param
p += int(decimal_param)
assert isinstance(float_param, float) and float_equal(float_param, 1.23, 1e-06), \
'float_param is wrong value %s ' % float_param
p += int(float_param)
assert isinstance(double_param, float) and float_equal(double_param, 1.98932, 1e-07), \
'double_param is wrong value %s ' % double_param
p += int(double_param)
assert boolean_param is True, 'boolean_param is wrong value %s' % boolean_param
assert str_param == 'flink', 'str_param is wrong value %s' % str_param
assert date_param == datetime.date(year=2014, month=9, day=13), \
'date_param is wrong value %s' % date_param
assert time_param == datetime.time(hour=12, minute=0, second=0), \
'time_param is wrong value %s' % time_param
assert timestamp_param == datetime.datetime(1999, 9, 10, 5, 20, 10), \
'timestamp_param is wrong value %s' % timestamp_param
return p
self.t_env.register_function(
"udf_with_constant_params",
udf(udf_with_constant_params,
input_types=[
DataTypes.BIGINT(),
DataTypes.BIGINT(),
DataTypes.TINYINT(),
DataTypes.SMALLINT(),
DataTypes.INT(),
DataTypes.BIGINT(),
DataTypes.DECIMAL(38, 18),
DataTypes.FLOAT(),
DataTypes.DOUBLE(),
DataTypes.BOOLEAN(),
DataTypes.STRING(),
DataTypes.DATE(),
DataTypes.TIME(),
DataTypes.TIMESTAMP(3)
],
result_type=DataTypes.BIGINT()))
self.t_env.register_function(
"udf_with_all_constant_params",
udf(lambda i, j: i + j,
[DataTypes.BIGINT(), DataTypes.BIGINT()], DataTypes.BIGINT()))
table_sink = source_sink_utils.TestAppendSink(
['a', 'b'],
[DataTypes.BIGINT(), DataTypes.BIGINT()])
self.t_env.register_table_sink("Results", table_sink)
t = self.t_env.from_elements([(1, 2, 3), (2, 5, 6), (3, 1, 9)],
['a', 'b', 'c'])
self.t_env.register_table("test_table", t)
self.t_env.sql_query("select udf_with_all_constant_params("
"cast (1 as BIGINT),"
"cast (2 as BIGINT)), "
"udf_with_constant_params(a, "
"cast (null as BIGINT),"
"cast (1 as TINYINT),"
"cast (1 as SMALLINT),"
"cast (1 as INT),"
"cast (1 as BIGINT),"
"cast (1.05 as DECIMAL),"
"cast (1.23 as FLOAT),"
"cast (1.98932 as DOUBLE),"
"true,"
"'flink',"
"cast ('2014-09-13' as DATE),"
"cast ('12:00:00' as TIME),"
"cast ('1999-9-10 05:20:10' as TIMESTAMP))"
" from test_table").insert_into("Results")
self.t_env.execute("test")
actual = source_sink_utils.results()
self.assert_equals(actual, ["3,8", "3,9", "3,10"])
from pyflink.datastream import StreamExecutionEnvironment, TimeCharacteristic
from pyflink.table import StreamTableEnvironment, DataTypes, EnvironmentSettings
from pyflink.table.udf import udf
provinces = ("Beijing", "Shanghai", "Hangzhou", "Shenzhen", "Jiangxi",
"Chongqing", "Xizang")
@udf(input_types=[DataTypes.STRING()], result_type=DataTypes.STRING())
def province_id_to_name(id):
return provinces[id]
def log_processing():
env = StreamExecutionEnvironment.get_execution_environment()
env_settings = EnvironmentSettings.Builder().use_blink_planner().build()
t_env = StreamTableEnvironment.create(stream_execution_environment=env,
environment_settings=env_settings)
t_env.get_config().get_configuration().set_boolean(
"python.fn-execution.memory.managed", True)
source_ddl = """
CREATE TABLE payment_msg(
createTime VARCHAR,
orderId BIGINT,
payAmount DOUBLE,
payPlatform INT,
provinceId INT
) WITH (
'connector.type' = 'kafka',
'connector.version' = 'universal',
def test_all_data_types(self):
def boolean_func(bool_param):
assert isinstance(bool_param, bool), 'bool_param of wrong type %s !' \
% type(bool_param)
return bool_param
def tinyint_func(tinyint_param):
assert isinstance(tinyint_param, int), 'tinyint_param of wrong type %s !' \
% type(tinyint_param)
return tinyint_param
def smallint_func(smallint_param):
assert isinstance(smallint_param, int), 'smallint_param of wrong type %s !' \
% type(smallint_param)
assert smallint_param == 32767, 'smallint_param of wrong value %s' % smallint_param
return smallint_param
def int_func(int_param):
assert isinstance(int_param, int), 'int_param of wrong type %s !' \
% type(int_param)
assert int_param == -2147483648, 'int_param of wrong value %s' % int_param
return int_param
def bigint_func(bigint_param):
assert isinstance(bigint_param, int), 'bigint_param of wrong type %s !' \
% type(bigint_param)
return bigint_param
def bigint_func_none(bigint_param):
assert bigint_param is None, 'bigint_param %s should be None!' % bigint_param
return bigint_param
def float_func(float_param):
assert isinstance(float_param, float) and float_equal(float_param, 1.23, 1e-6), \
'float_param is wrong value %s !' % float_param
return float_param
def double_func(double_param):
assert isinstance(double_param, float) and float_equal(double_param, 1.98932, 1e-7), \
'double_param is wrong value %s !' % double_param
return double_param
def bytes_func(bytes_param):
assert bytes_param == b'flink', \
'bytes_param is wrong value %s !' % bytes_param
return bytes_param
def str_func(str_param):
assert str_param == 'pyflink', \
'str_param is wrong value %s !' % str_param
return str_param
def date_func(date_param):
from datetime import date
assert date_param == date(year=2014, month=9, day=13), \
'date_param is wrong value %s !' % date_param
return date_param
def time_func(time_param):
from datetime import time
assert time_param == time(hour=12, minute=0, second=0, microsecond=123000), \
'time_param is wrong value %s !' % time_param
return time_param
def timestamp_func(timestamp_param):
from datetime import datetime
assert timestamp_param == datetime(2018, 3, 11, 3, 0, 0, 123000), \
'timestamp_param is wrong value %s !' % timestamp_param
return timestamp_param
def array_func(array_param):
assert array_param == [[1, 2, 3]], \
'array_param is wrong value %s !' % array_param
return array_param[0]
def map_func(map_param):
assert map_param == {1: 'flink', 2: 'pyflink'}, \
'map_param is wrong value %s !' % map_param
return map_param
def decimal_func(decimal_param):
from decimal import Decimal
assert decimal_param == Decimal('1000000000000000000.050000000000000000'), \
'decimal_param is wrong value %s !' % decimal_param
return decimal_param
def decimal_cut_func(decimal_param):
from decimal import Decimal
assert decimal_param == Decimal('1000000000000000000.059999999999999999'), \
'decimal_param is wrong value %s !' % decimal_param
return decimal_param
self.t_env.register_function(
"boolean_func",
udf(boolean_func, [DataTypes.BOOLEAN()], DataTypes.BOOLEAN()))
self.t_env.register_function(
"tinyint_func",
udf(tinyint_func, [DataTypes.TINYINT()], DataTypes.TINYINT()))
self.t_env.register_function(
"smallint_func",
udf(smallint_func, [DataTypes.SMALLINT()], DataTypes.SMALLINT()))
self.t_env.register_function(
"int_func", udf(int_func, [DataTypes.INT()], DataTypes.INT()))
self.t_env.register_function(
"bigint_func",
udf(bigint_func, [DataTypes.BIGINT()], DataTypes.BIGINT()))
self.t_env.register_function(
"bigint_func_none",
udf(bigint_func_none, [DataTypes.BIGINT()], DataTypes.BIGINT()))
self.t_env.register_function(
"float_func",
udf(float_func, [DataTypes.FLOAT()], DataTypes.FLOAT()))
self.t_env.register_function(
"double_func",
udf(double_func, [DataTypes.DOUBLE()], DataTypes.DOUBLE()))
self.t_env.register_function(
"bytes_func",
udf(bytes_func, [DataTypes.BYTES()], DataTypes.BYTES()))
self.t_env.register_function(
"str_func", udf(str_func, [DataTypes.STRING()],
DataTypes.STRING()))
self.t_env.register_function(
"date_func", udf(date_func, [DataTypes.DATE()], DataTypes.DATE()))
self.t_env.register_function(
"time_func", udf(time_func, [DataTypes.TIME()], DataTypes.TIME()))
self.t_env.register_function(
"timestamp_func",
udf(timestamp_func, [DataTypes.TIMESTAMP(3)],
DataTypes.TIMESTAMP(3)))
self.t_env.register_function(
"array_func",
udf(array_func,
[DataTypes.ARRAY(DataTypes.ARRAY(DataTypes.BIGINT()))],
DataTypes.ARRAY(DataTypes.BIGINT())))
self.t_env.register_function(
"map_func",
udf(map_func,
[DataTypes.MAP(DataTypes.BIGINT(), DataTypes.STRING())],
DataTypes.MAP(DataTypes.BIGINT(), DataTypes.STRING())))
self.t_env.register_function(
"decimal_func",
udf(decimal_func, [DataTypes.DECIMAL(38, 18)],
DataTypes.DECIMAL(38, 18)))
self.t_env.register_function(
"decimal_cut_func",
udf(decimal_cut_func, [DataTypes.DECIMAL(38, 18)],
DataTypes.DECIMAL(38, 18)))
table_sink = source_sink_utils.TestAppendSink([
'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm',
'n', 'o', 'p', 'q'
], [
DataTypes.BIGINT(),
DataTypes.BIGINT(),
DataTypes.TINYINT(),
DataTypes.BOOLEAN(),
DataTypes.SMALLINT(),
DataTypes.INT(),
DataTypes.FLOAT(),
DataTypes.DOUBLE(),
DataTypes.BYTES(),
DataTypes.STRING(),
DataTypes.DATE(),
DataTypes.TIME(),
DataTypes.TIMESTAMP(3),
DataTypes.ARRAY(DataTypes.BIGINT()),
DataTypes.MAP(DataTypes.BIGINT(), DataTypes.STRING()),
DataTypes.DECIMAL(38, 18),
DataTypes.DECIMAL(38, 18)
])
self.t_env.register_table_sink("Results", table_sink)
import datetime
import decimal
t = self.t_env.from_elements(
[(1, None, 1, True, 32767, -2147483648, 1.23, 1.98932,
bytearray(b'flink'), 'pyflink', datetime.date(2014, 9, 13),
datetime.time(hour=12, minute=0, second=0, microsecond=123000),
datetime.datetime(2018, 3, 11, 3, 0, 0, 123000), [[1, 2, 3]], {
1: 'flink',
2: 'pyflink'
}, decimal.Decimal('1000000000000000000.05'),
decimal.Decimal(
'1000000000000000000.05999999999999999899999999999'))],
DataTypes.ROW([
DataTypes.FIELD("a", DataTypes.BIGINT()),
DataTypes.FIELD("b", DataTypes.BIGINT()),
DataTypes.FIELD("c", DataTypes.TINYINT()),
DataTypes.FIELD("d", DataTypes.BOOLEAN()),
DataTypes.FIELD("e", DataTypes.SMALLINT()),
DataTypes.FIELD("f", DataTypes.INT()),
DataTypes.FIELD("g", DataTypes.FLOAT()),
DataTypes.FIELD("h", DataTypes.DOUBLE()),
DataTypes.FIELD("i", DataTypes.BYTES()),
DataTypes.FIELD("j", DataTypes.STRING()),
DataTypes.FIELD("k", DataTypes.DATE()),
DataTypes.FIELD("l", DataTypes.TIME()),
DataTypes.FIELD("m", DataTypes.TIMESTAMP(3)),
DataTypes.FIELD(
"n", DataTypes.ARRAY(DataTypes.ARRAY(DataTypes.BIGINT()))),
DataTypes.FIELD(
"o", DataTypes.MAP(DataTypes.BIGINT(),
DataTypes.STRING())),
DataTypes.FIELD("p", DataTypes.DECIMAL(38, 18)),
DataTypes.FIELD("q", DataTypes.DECIMAL(38, 18))
]))
t.select("bigint_func(a), bigint_func_none(b),"
"tinyint_func(c), boolean_func(d),"
"smallint_func(e),int_func(f),"
"float_func(g),double_func(h),"
"bytes_func(i),str_func(j),"
"date_func(k),time_func(l),"
"timestamp_func(m),array_func(n),"
"map_func(o),decimal_func(p),"
"decimal_cut_func(q)") \
.insert_into("Results")
self.t_env.execute("test")
actual = source_sink_utils.results()
# Currently the sink result precision of DataTypes.TIME(precision) only supports 0.
self.assert_equals(actual, [
"1,null,1,true,32767,-2147483648,1.23,1.98932,"
"[102, 108, 105, 110, 107],pyflink,2014-09-13,"
"12:00:00,2018-03-11 03:00:00.123,[1, 2, 3],"
"{1=flink, 2=pyflink},1000000000000000000.050000000000000000,"
"1000000000000000000.059999999999999999"
])
def custom_kafka_source_demo():
custom_connector = CustomConnectorDescriptor('kafka', 1, True) \
.property('connector.topic', 'user') \
.property('connector.properties.0.key', 'zookeeper.connect') \
.property('connector.properties.0.value', 'localhost:2181') \
.property('connector.properties.1.key', 'bootstrap.servers') \
.property('connector.properties.1.value', 'localhost:9092') \
.properties({'connector.version': '0.11', 'connector.startup-mode': 'earliest-offset'})
# the key is 'format.json-schema'
custom_format = CustomFormatDescriptor('json', 1) \
.property('format.json-schema',
"{"
" type: 'object',"
" properties: {"
" a: {"
" type: 'string'"
" },"
" b: {"
" type: 'string'"
" },"
" c: {"
" type: 'string'"
" },"
" time: {"
" type: 'string',"
" format: 'date-time'"
" }"
" }"
"}") \
.properties({'format.fail-on-missing-field': 'true'})
s_env = StreamExecutionEnvironment.get_execution_environment()
s_env.set_parallelism(1)
s_env.set_stream_time_characteristic(TimeCharacteristic.ProcessingTime)
st_env = StreamTableEnvironment.create(s_env)
result_file = "/tmp/custom_kafka_source_demo.csv"
if os.path.exists(result_file):
os.remove(result_file)
st_env \
.connect(custom_connector) \
.with_format(
custom_format
) \
.with_schema( # declare the schema of the table
Schema()
.field("proctime", DataTypes.TIMESTAMP())
.proctime()
.field("a", DataTypes.STRING())
.field("b", DataTypes.STRING())
.field("c", DataTypes.STRING())
) \
.in_append_mode() \
.register_table_source("source")
st_env.register_table_sink(
"result",
CsvTableSink(
["a", "b"],
[DataTypes.STRING(), DataTypes.STRING()], result_file))
st_env.scan("source").window(Tumble.over("2.rows").on("proctime").alias("w")) \
.group_by("w, a") \
.select("a, max(b)").insert_into("result")
st_env.execute("custom kafka source demo")
from pyflink.datastream import StreamExecutionEnvironment
from pyflink.table import StreamTableEnvironment, DataTypes
from pyflink.table.udf import udf
# https://flink.apache.org/2020/04/09/pyflink-udf-support-flink.html
# https://ci.apache.org/projects/flink/flink-docs-release-1.12/dev/python/table-api-users-guide/udfs/python_udfs.html
@udf(input_types=[DataTypes.INT(), DataTypes.INT()], result_type=DataTypes.BIGINT(), func_type="pandas")
def add(i, j):
return i + j
env = StreamExecutionEnvironment.get_execution_environment()
t_env = StreamTableEnvironment.create(env)
t_env.create_temporary_function("add", add)
my_source_ddl = """
create table mySource (
a INT,
b INT
) with (
'connector' = 'datagen',
'rows-per-second' = '5'
)
"""
my_sink_ddl = """
create table mySink (
c BIGINT
) with (
'connector' = 'print'
def test_expressions(self):
expr1 = col('a')
expr2 = col('b')
expr3 = col('c')
self.assertEqual('10', str(lit(10, DataTypes.INT(False))))
self.assertEqual('rangeTo(1, 2)', str(range_(1, 2)))
self.assertEqual('and(a, b, c)', str(and_(expr1, expr2, expr3)))
self.assertEqual('or(a, b, c)', str(or_(expr1, expr2, expr3)))
from pyflink.table.expressions import UNBOUNDED_ROW, UNBOUNDED_RANGE, CURRENT_ROW, \
CURRENT_RANGE
self.assertEqual('unboundedRow()', str(UNBOUNDED_ROW))
self.assertEqual('unboundedRange()', str(UNBOUNDED_RANGE))
self.assertEqual('currentRow()', str(CURRENT_ROW))
self.assertEqual('currentRange()', str(CURRENT_RANGE))
self.assertEqual('currentDate()', str(current_date()))
self.assertEqual('currentTime()', str(current_time()))
self.assertEqual('currentTimestamp()', str(current_timestamp()))
self.assertEqual('localTime()', str(local_time()))
self.assertEqual('localTimestamp()', str(local_timestamp()))
self.assertEquals('toTimestampLtz(123, 0)',
str(to_timestamp_ltz(123, 0)))
self.assertEqual(
"temporalOverlaps(cast('2:55:00', TIME(0)), 3600000, "
"cast('3:30:00', TIME(0)), 7200000)",
str(
temporal_overlaps(
lit("2:55:00").to_time,
lit(1).hours,
lit("3:30:00").to_time,
lit(2).hours)))
self.assertEqual("dateFormat(time, '%Y, %d %M')",
str(date_format(col("time"), "%Y, %d %M")))
self.assertEqual(
"timestampDiff(DAY, cast('2016-06-15', DATE), cast('2016-06-18', DATE))",
str(
timestamp_diff(TimePointUnit.DAY,
lit("2016-06-15").to_date,
lit("2016-06-18").to_date)))
self.assertEqual('array(1, 2, 3)', str(array(1, 2, 3)))
self.assertEqual("row('key1', 1)", str(row("key1", 1)))
self.assertEqual("map('key1', 1, 'key2', 2, 'key3', 3)",
str(map_("key1", 1, "key2", 2, "key3", 3)))
self.assertEqual('4', str(row_interval(4)))
self.assertEqual('pi()', str(pi()))
self.assertEqual('e()', str(e()))
self.assertEqual('rand(4)', str(rand(4)))
self.assertEqual('randInteger(4)', str(rand_integer(4)))
self.assertEqual('atan2(1, 2)', str(atan2(1, 2)))
self.assertEqual('minusPrefix(a)', str(negative(expr1)))
self.assertEqual('concat(a, b, c)', str(concat(expr1, expr2, expr3)))
self.assertEqual("concat_ws(', ', b, c)",
str(concat_ws(', ', expr2, expr3)))
self.assertEqual('uuid()', str(uuid()))
self.assertEqual('null', str(null_of(DataTypes.BIGINT())))
self.assertEqual('log(a)', str(log(expr1)))
self.assertEqual('ifThenElse(a, b, c)',
str(if_then_else(expr1, expr2, expr3)))
self.assertEqual('withColumns(a, b, c)',
str(with_columns(expr1, expr2, expr3)))
self.assertEqual('a.b.c(a)', str(call('a.b.c', expr1)))
def register_rides_sink(st_env):
st_env \
.connect( # declare the external system to connect to
Kafka()
.version("0.11")
.topic("TempResults")
.property("zookeeper.connect", "zookeeper:2181")
.property("bootstrap.servers", "kafka:9092")) \
.with_format( # declare a format for this system
Json()
.fail_on_missing_field(True)
.schema(DataTypes.ROW([
DataTypes.FIELD("rideId", DataTypes.BIGINT()),
DataTypes.FIELD("taxiId", DataTypes.BIGINT()),
DataTypes.FIELD("isStart", DataTypes.BOOLEAN()),
DataTypes.FIELD("lon", DataTypes.FLOAT()),
DataTypes.FIELD("lat", DataTypes.FLOAT()),
DataTypes.FIELD("psgCnt", DataTypes.INT()),
DataTypes.FIELD("rideTime", DataTypes.TIMESTAMP())
]))) \
.with_schema( # declare the schema of the table
Schema()
.field("rideId", DataTypes.BIGINT())
.field("taxiId", DataTypes.BIGINT())
.field("isStart", DataTypes.BOOLEAN())
.field("lon", DataTypes.FLOAT())
.field("lat", DataTypes.FLOAT())
.field("psgCnt", DataTypes.INT())
.field("rideTime", DataTypes.TIMESTAMP())) \
.in_append_mode() \
.register_table_sink("sink")
# https://ci.apache.org/projects/flink/flink-docs-master/getting-started/walkthroughs/python_table_api.html
from pyflink.dataset import ExecutionEnvironment
from pyflink.table import TableConfig, DataTypes, BatchTableEnvironment
from pyflink.table.descriptors import Schema, OldCsv, FileSystem
exec_env = ExecutionEnvironment.get_execution_environment()
exec_env.set_parallelism(1)
t_config = TableConfig()
t_env = BatchTableEnvironment.create(exec_env, t_config)
t_env.connect(FileSystem().path('/tmp/input')) \
.with_format(OldCsv()
.field('word', DataTypes.STRING())) \
.with_schema(Schema()
.field('word', DataTypes.STRING())) \
.create_temporary_table('mySource')
t_env.connect(FileSystem().path('/tmp/output')) \
.with_format(OldCsv()
.field_delimiter('\t')
.field('word', DataTypes.STRING())
.field('count', DataTypes.BIGINT())) \
.with_schema(Schema()
.field('word', DataTypes.STRING())
.field('count', DataTypes.BIGINT())) \
.create_temporary_table('mySink')
t_env.from_path('mySource') \
.group_by('word') \
.select('word, count(1)') \
