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),
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'))]
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(
'1000000000000000000.0599999999999999989'
'9999999999'))],
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 _create_csv_primitive_column_schema_and_lines(
) -> Tuple[CsvSchema, List[str]]:
schema = CsvSchema.builder() \
.add_number_column('tinyint', DataTypes.TINYINT()) \
.add_number_column('smallint', DataTypes.SMALLINT()) \
.add_number_column('int', DataTypes.INT()) \
.add_number_column('bigint', DataTypes.BIGINT()) \
.add_number_column('float', DataTypes.FLOAT()) \
.add_number_column('double', DataTypes.DOUBLE()) \
.add_number_column('decimal', DataTypes.DECIMAL(2, 0)) \
.add_boolean_column('boolean') \
.add_string_column('string') \
.build()
lines = [
'127,'
'-32767,'
'2147483647,'
'-9223372036854775808,'
'3e38,'
'2e-308,'
'1.5,'
'true,'
'string\n',
]
return schema, lines
def test_from_element(self):
t_env = self.t_env
field_names = [
"a", "b", "c", "d", "e", "f", "g", "h", "i", "j", "k", "l", "m",
"n", "o", "p", "q", "r", "s"
]
field_types = [
DataTypes.BIGINT(),
DataTypes.DOUBLE(),
DataTypes.STRING(),
DataTypes.STRING(),
DataTypes.DATE(),
DataTypes.TIME(),
DataTypes.TIMESTAMP(),
DataTypes.TIMESTAMP_WITH_LOCAL_TIME_ZONE(),
DataTypes.INTERVAL(DataTypes.DAY(), DataTypes.SECOND()),
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()
]
schema = DataTypes.ROW(
list(
map(
lambda field_name, field_type: DataTypes.FIELD(
field_name, field_type), field_names, field_types)))
table_sink = source_sink_utils.TestAppendSink(field_names, field_types)
t_env.register_table_sink("Results", table_sink)
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), datetime.datetime(1970, 1, 2, 0, 0),
datetime.timedelta(days=1, microseconds=10), [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
}, bytearray(b'ABCD'), ExamplePoint(
1.0, 2.0), PythonOnlyPoint(3.0, 4.0))], schema)
t.insert_into("Results")
self.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,'
'1970-01-02 00:00:00.0,86400000010,[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_blink_from_element(self):
t_env = BatchTableEnvironment.create(
environment_settings=EnvironmentSettings.new_instance(
).use_blink_planner().in_batch_mode().build())
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(3),
DataTypes.INTERVAL(DataTypes.SECOND(3)),
DataTypes.ARRAY(DataTypes.DOUBLE()),
DataTypes.ARRAY(DataTypes.DOUBLE(False)),
DataTypes.ARRAY(DataTypes.STRING()),
DataTypes.ARRAY(DataTypes.DATE()),
DataTypes.DECIMAL(38, 18),
DataTypes.ROW([
DataTypes.FIELD("a", DataTypes.BIGINT()),
DataTypes.FIELD("b", DataTypes.DOUBLE())
]),
DataTypes.MAP(DataTypes.STRING(), DataTypes.DOUBLE()),
DataTypes.BYTES(),
PythonOnlyUDT()
]
schema = DataTypes.ROW(
list(
map(
lambda field_name, field_type: DataTypes.FIELD(
field_name, field_type), field_names, field_types)))
table_sink = source_sink_utils.TestAppendSink(field_names, field_types)
t_env.register_table_sink("Results", table_sink)
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),
datetime.timedelta(days=1, microseconds=10), [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
}, bytearray(b'ABCD'), PythonOnlyPoint(3.0, 4.0))], schema)
t.insert_into("Results")
t_env.execute("test")
actual = source_sink_utils.results()
expected = [
'1,1.0,hi,hello,1970-01-02,01:00:00,1970-01-02 00:00:00.0,'
'86400000,[1.0, null],[1.0, 2.0],[abc],[1970-01-02],'
'1.000000000000000000,1,2.0,{key=1.0},[65, 66, 67, 68],[3.0, 4.0]'
]
self.assert_equals(actual, expected)
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(),
DataTypes.ROW([
DataTypes.FIELD("a", DataTypes.BIGINT()),
DataTypes.FIELD("b", DataTypes.DOUBLE())
]),
DataTypes.MAP(DataTypes.VARCHAR(), DataTypes.DOUBLE()),
DataTypes.VARBINARY(),
ExamplePointUDT(),
PythonOnlyUDT()
]
t_env.register_table_sink("Results", field_names, field_types,
source_sink_utils.TestAppendSink())
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
@udf(result_type=DataTypes.TINYINT(), func_type="pandas")
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
@udf(result_type=DataTypes.SMALLINT(), func_type="pandas")
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
@udf(result_type=DataTypes.INT(), func_type="pandas")
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
@udf(result_type=DataTypes.BIGINT(), func_type="pandas")
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
@udf(result_type=DataTypes.BOOLEAN(), func_type="pandas")
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
@udf(result_type=DataTypes.FLOAT(), func_type="pandas")
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
@udf(result_type=DataTypes.DOUBLE(), func_type="pandas")
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
@udf(result_type=DataTypes.STRING(), func_type="pandas")
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
@udf(result_type=DataTypes.BYTES(), func_type="pandas")
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
@udf(result_type=DataTypes.DECIMAL(38, 18), func_type="pandas")
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
@udf(result_type=DataTypes.DATE(), func_type="pandas")
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
@udf(result_type=DataTypes.TIME(), func_type="pandas")
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)
@udf(result_type=DataTypes.TIMESTAMP(3), func_type="pandas")
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
array_str_func = udf(array_func,
result_type=DataTypes.ARRAY(DataTypes.STRING()),
func_type="pandas")
array_timestamp_func = udf(array_func,
result_type=DataTypes.ARRAY(
DataTypes.TIMESTAMP(3)),
func_type="pandas")
array_int_func = udf(array_func,
result_type=DataTypes.ARRAY(DataTypes.INT()),
func_type="pandas")
@udf(result_type=DataTypes.ARRAY(DataTypes.STRING()),
func_type="pandas")
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])
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()))
])
@udf(result_type=row_type, func_type="pandas")
def row_func(row_param):
assert isinstance(row_param, pd.DataFrame)
assert isinstance(row_param.f1, pd.Series)
assert isinstance(row_param.f1[0], np.int32), \
'row_param.f1 of wrong type %s !' % type(row_param.f1[0])
assert isinstance(row_param.f2, pd.Series)
assert isinstance(row_param.f2[0], str), \
'row_param.f2 of wrong type %s !' % type(row_param.f2[0])
assert isinstance(row_param.f3, pd.Series)
assert isinstance(row_param.f3[0], datetime.datetime), \
'row_param.f3 of wrong type %s !' % type(row_param.f3[0])
assert isinstance(row_param.f4, pd.Series)
assert isinstance(row_param.f4[0], np.ndarray), \
'row_param.f4 of wrong type %s !' % type(row_param.f4[0])
return row_param
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)
]))
t.select(
tinyint_func(t.a),
smallint_func(t.b),
int_func(t.c),
bigint_func(t.d),
boolean_func(t.e),
boolean_func(t.f),
float_func(t.g),
double_func(t.h),
varchar_func(t.i),
varchar_func(t.j),
varbinary_func(t.k),
decimal_func(t.l),
decimal_func(t.m),
date_func(t.n),
time_func(t.o),
timestamp_func(t.p),
array_str_func(t.q),
array_timestamp_func(t.r),
array_int_func(t.s),
nested_array_func(t.t),
row_func(t.u)) \
.execute_insert("Results").wait()
actual = source_sink_utils.results()
self.assert_equals(actual, [
"+I[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], +I[1, hello, 1970-01-02 00:00:00.123, [1, 2]]]"
])
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.create_temporary_system_function(
"boolean_func", udf(boolean_func, result_type=DataTypes.BOOLEAN()))
self.t_env.create_temporary_system_function(
"tinyint_func", udf(tinyint_func, result_type=DataTypes.TINYINT()))
self.t_env.create_temporary_system_function(
"smallint_func",
udf(smallint_func, result_type=DataTypes.SMALLINT()))
self.t_env.create_temporary_system_function(
"int_func", udf(int_func, result_type=DataTypes.INT()))
self.t_env.create_temporary_system_function(
"bigint_func", udf(bigint_func, result_type=DataTypes.BIGINT()))
self.t_env.create_temporary_system_function(
"bigint_func_none",
udf(bigint_func_none, result_type=DataTypes.BIGINT()))
self.t_env.create_temporary_system_function(
"float_func", udf(float_func, result_type=DataTypes.FLOAT()))
self.t_env.create_temporary_system_function(
"double_func", udf(double_func, result_type=DataTypes.DOUBLE()))
self.t_env.create_temporary_system_function(
"bytes_func", udf(bytes_func, result_type=DataTypes.BYTES()))
self.t_env.create_temporary_system_function(
"str_func", udf(str_func, result_type=DataTypes.STRING()))
self.t_env.create_temporary_system_function(
"date_func", udf(date_func, result_type=DataTypes.DATE()))
self.t_env.create_temporary_system_function(
"time_func", udf(time_func, result_type=DataTypes.TIME()))
self.t_env.create_temporary_system_function(
"timestamp_func",
udf(timestamp_func, result_type=DataTypes.TIMESTAMP(3)))
self.t_env.create_temporary_system_function(
"array_func",
udf(array_func, result_type=DataTypes.ARRAY(DataTypes.BIGINT())))
self.t_env.create_temporary_system_function(
"map_func",
udf(map_func,
result_type=DataTypes.MAP(DataTypes.BIGINT(),
DataTypes.STRING())))
self.t_env.register_function(
"decimal_func",
udf(decimal_func, result_type=DataTypes.DECIMAL(38, 18)))
self.t_env.register_function(
"decimal_cut_func",
udf(decimal_cut_func, result_type=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))
]))
exec_insert_table(
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)"), "Results")
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 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
from pyflink.table import StreamTableEnvironment, DataTypes
from pyflink.table.descriptors import Schema, OldCsv, FileSystem
from pyflink.table.udf import udf
import numpy as np
from sklearn.linear_model import LinearRegression as LR
from sklearn.preprocessing import PolynomialFeatures
from sklearn.model_selection import train_test_split
env = StreamExecutionEnvironment.get_execution_environment()
env.set_parallelism(1)
t_env = StreamTableEnvironment.create(env)
@udf(input_types=[DataTypes.DECIMAL(38, 12, nullable=True)],
result_type=DataTypes.DECIMAL(38, 12, nullable=True))
def myadd(i):
return i * i * 2
# add = udf(myadd, [DataTypes.BIGINT(), DataTypes.BIGINT()], DataTypes.BIGINT())
t_env.register_function("add", myadd)
t_env.connect(FileSystem().path('/tmp/input')) \
.with_format(OldCsv()
.field('b', DataTypes.DECIMAL(38,12,nullable=True))) \
.with_schema(Schema()
.field('b', DataTypes.DECIMAL(38,12,nullable=True))) \
.create_temporary_table('mySource')
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, 1, 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
self.t_env.register_function(
"tinyint_func",
udf(tinyint_func, [DataTypes.TINYINT()],
DataTypes.TINYINT(),
udf_type="pandas"))
self.t_env.register_function(
"smallint_func",
udf(smallint_func, [DataTypes.SMALLINT()],
DataTypes.SMALLINT(),
udf_type="pandas"))
self.t_env.register_function(
"int_func",
udf(int_func, [DataTypes.INT()],
DataTypes.INT(),
udf_type="pandas"))
self.t_env.register_function(
"bigint_func",
udf(bigint_func, [DataTypes.BIGINT()],
DataTypes.BIGINT(),
udf_type="pandas"))
self.t_env.register_function(
"boolean_func",
udf(boolean_func, [DataTypes.BOOLEAN()],
DataTypes.BOOLEAN(),
udf_type="pandas"))
self.t_env.register_function(
"float_func",
udf(float_func, [DataTypes.FLOAT()],
DataTypes.FLOAT(),
udf_type="pandas"))
self.t_env.register_function(
"double_func",
udf(double_func, [DataTypes.DOUBLE()],
DataTypes.DOUBLE(),
udf_type="pandas"))
self.t_env.register_function(
"varchar_func",
udf(varchar_func, [DataTypes.STRING()],
DataTypes.STRING(),
udf_type="pandas"))
self.t_env.register_function(
"varbinary_func",
udf(varbinary_func, [DataTypes.BYTES()],
DataTypes.BYTES(),
udf_type="pandas"))
self.t_env.register_function(
"decimal_func",
udf(decimal_func, [DataTypes.DECIMAL(38, 18)],
DataTypes.DECIMAL(38, 18),
udf_type="pandas"))
self.t_env.register_function(
"date_func",
udf(date_func, [DataTypes.DATE()],
DataTypes.DATE(),
udf_type="pandas"))
self.t_env.register_function(
"time_func",
udf(time_func, [DataTypes.TIME()],
DataTypes.TIME(),
udf_type="pandas"))
self.t_env.register_function(
"timestamp_func",
udf(timestamp_func, [DataTypes.TIMESTAMP(3)],
DataTypes.TIMESTAMP(3),
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'
], [
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)
])
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)],
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))
]))
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)") \
.insert_into("Results")
self.t_env.execute("test")
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,"
"1970-01-01 00:00:00.123"
])
if __name__ == '__main__':
# stream setting
s_env = StreamExecutionEnvironment.get_execution_environment()
s_env.set_parallelism(1)
s_env.set_stream_time_characteristic(TimeCharacteristic.EventTime)
st_env = StreamTableEnvironment.create(s_env)
# csv path for csv sink
result_file = "/tmp/tumble_time_window_streaming.csv"
if os.path.exists(result_file):
os.remove(result_file)
# udf
@udf(input_types=[DataTypes.DECIMAL(38, 12, nullable=True)], result_type=DataTypes.DECIMAL(38, 12, nullable=True))
def myadd(i):
return i * i * 2
st_env.register_function("add", myadd)
# way kafka
st_env \
.connect( # declare the external system to connect to
Kafka()
.version("universal")
.topic("user")
# .start_from_earliest()
# .start_from_earliest()
.start_from_specific_offset(0,496)
.property("zookeeper.connect", "6.86.2.170:2181")
.property("bootstrap.servers", "6.86.2.170:9092")
def test_from_element(self):
t_env = self.t_env
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(3),
DataTypes.INTERVAL(DataTypes.SECOND(3)),
DataTypes.ARRAY(DataTypes.DOUBLE()),
DataTypes.ARRAY(DataTypes.DOUBLE(False)),
DataTypes.ARRAY(DataTypes.STRING()),
DataTypes.ARRAY(DataTypes.DATE()),
DataTypes.DECIMAL(38, 18),
DataTypes.ROW([
DataTypes.FIELD("a", DataTypes.BIGINT()),
DataTypes.FIELD("b", DataTypes.DOUBLE())
]),
DataTypes.MAP(DataTypes.STRING(), DataTypes.DOUBLE()),
DataTypes.BYTES(),
PythonOnlyUDT()
]
schema = DataTypes.ROW(
list(
map(
lambda field_name, field_type: DataTypes.FIELD(
field_name, field_type), field_names, field_types)))
sink_table_ddl = """
CREATE TABLE Results(
a BIGINT,
b DOUBLE,
c STRING,
d STRING,
e DATE,
f TIME,
g TIMESTAMP(3),
h INT,
i ARRAY<DOUBLE>,
j ARRAY<DOUBLE NOT NULL>,
k ARRAY<STRING>,
l ARRAY<DATE>,
m DECIMAL(38, 18),
n ROW<a BIGINT, b DOUBLE>,
o MAP<STRING, DOUBLE>,
p BYTES,
q ARRAY<DOUBLE NOT NULL>)
WITH ('connector'='test-sink')
"""
self.t_env.execute_sql(sink_table_ddl)
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),
datetime.timedelta(days=1, microseconds=10), [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
}, bytearray(b'ABCD'), PythonOnlyPoint(3.0, 4.0))], schema)
t.execute_insert("Results").wait()
actual = source_sink_utils.results()
expected = [
'+I[1, 1.0, hi, hello, 1970-01-02, 01:00, 1970-01-02T00:00, '
'86400, [1.0, null], [1.0, 2.0], [abc], [1970-01-02], '
'1.000000000000000000, +I[1, 2.0], {key=1.0}, [65, 66, 67, 68], [3.0, 4.0]]'
]
self.assert_equals(actual, expected)
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]] or 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.create_temporary_system_function(
"boolean_func", udf(boolean_func, result_type=DataTypes.BOOLEAN()))
self.t_env.create_temporary_system_function(
"tinyint_func", udf(tinyint_func, result_type=DataTypes.TINYINT()))
self.t_env.create_temporary_system_function(
"smallint_func",
udf(smallint_func, result_type=DataTypes.SMALLINT()))
self.t_env.create_temporary_system_function(
"int_func", udf(int_func, result_type=DataTypes.INT()))
self.t_env.create_temporary_system_function(
"bigint_func", udf(bigint_func, result_type=DataTypes.BIGINT()))
self.t_env.create_temporary_system_function(
"bigint_func_none",
udf(bigint_func_none, result_type=DataTypes.BIGINT()))
self.t_env.create_temporary_system_function(
"float_func", udf(float_func, result_type=DataTypes.FLOAT()))
self.t_env.create_temporary_system_function(
"double_func", udf(double_func, result_type=DataTypes.DOUBLE()))
self.t_env.create_temporary_system_function(
"bytes_func", udf(bytes_func, result_type=DataTypes.BYTES()))
self.t_env.create_temporary_system_function(
"str_func", udf(str_func, result_type=DataTypes.STRING()))
self.t_env.create_temporary_system_function(
"date_func", udf(date_func, result_type=DataTypes.DATE()))
self.t_env.create_temporary_system_function(
"time_func", udf(time_func, result_type=DataTypes.TIME()))
self.t_env.create_temporary_system_function(
"timestamp_func",
udf(timestamp_func, result_type=DataTypes.TIMESTAMP(3)))
self.t_env.create_temporary_system_function(
"array_func",
udf(array_func, result_type=DataTypes.ARRAY(DataTypes.BIGINT())))
self.t_env.create_temporary_system_function(
"map_func",
udf(map_func,
result_type=DataTypes.MAP(DataTypes.BIGINT(),
DataTypes.STRING())))
self.t_env.register_function(
"decimal_func",
udf(decimal_func, result_type=DataTypes.DECIMAL(38, 18)))
self.t_env.register_function(
"decimal_cut_func",
udf(decimal_cut_func, result_type=DataTypes.DECIMAL(38, 18)))
sink_table_ddl = """
CREATE TABLE Results(
a BIGINT, b BIGINT, c TINYINT, d BOOLEAN, e SMALLINT, f INT, g FLOAT, h DOUBLE,
i BYTES, j STRING, k DATE, l TIME, m TIMESTAMP(3), n ARRAY<BIGINT>,
o MAP<BIGINT, STRING>, p DECIMAL(38, 18), q DECIMAL(38, 18))
WITH ('connector'='test-sink')
"""
self.t_env.execute_sql(sink_table_ddl)
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(call("bigint_func", t.a), call("bigint_func_none", t.b),
call("tinyint_func", t.c), call("boolean_func", t.d),
call("smallint_func", t.e), call("int_func", t.f),
call("float_func", t.g), call("double_func", t.h),
call("bytes_func", t.i), call("str_func", t.j),
call("date_func", t.k), call("time_func", t.l),
call("timestamp_func", t.m), call("array_func", t.n),
call("map_func", t.o), call("decimal_func", t.p),
call("decimal_cut_func", t.q)) \
.execute_insert("Results").wait()
actual = source_sink_utils.results()
# Currently the sink result precision of DataTypes.TIME(precision) only supports 0.
self.assert_equals(actual, [
"+I[1, null, 1, true, 32767, -2147483648, 1.23, 1.98932, "
"[102, 108, 105, 110, 107], pyflink, 2014-09-13, "
"12:00:00.123, 2018-03-11T03:00:00.123, [1, 2, 3], "
"{1=flink, 2=pyflink}, 1000000000000000000.050000000000000000, "
"1000000000000000000.059999999999999999]"
])
