我有一个Python模块，其结构如下：
my_module/
...
tests/
__init__.py
my_test.py
...
其中my_test.py的定义如下：
from __future__ import print_function, unicode_literals
import os
import unittest
from datetime import timedelta, date
from airflow import configuration
from airflow.models import TaskInstance as TI, DAG, DagRun
from airflow.operators.python_operator import PythonOperator
from airflow.settings import Session
from airflow.utils import timezone
from airflow.utils.state import State
DEFAULT_DATE = timezone.datetime(2016, 1, 1)
END_DATE = timezone.datetime(2016, 1, 2)
INTERVAL = timedelta(hours=12)
FROZEN_NOW = timezone.datetime(2016, 1, 2, 12, 1, 1)
TI_CONTEXT_ENV_VARS = ['AIRFLOW_CTX_DAG_ID',
'AIRFLOW_CTX_TASK_ID',
'AIRFLOW_CTX_EXECUTION_DATE',
'AIRFLOW_CTX_DAG_RUN_ID']
class Call:
def __init__(self, *args, **kwargs):
self.args = args
self.kwargs = kwargs
def build_recording_function(calls_collection):
"""
We can not use a Mock instance as a PythonOperator callable function or some tests fail with a
TypeError: Object of type Mock is not JSON serializable
Then using this custom function recording custom Call objects for further testing
(replacing Mock.assert_called_with assertion method)
"""
def recording_function(*args, **kwargs):
calls_collection.append(Call(*args, **kwargs))
return recording_function
class PythonOperatorTest(unittest.TestCase):
@classmethod
def setUpClass(cls):
super(PythonOperatorTest, cls).setUpClass()
session = Session()
session.query(DagRun).delete()
session.query(TI).delete()
session.commit()
session.close()
def setUp(self):
super(PythonOperatorTest, self).setUp()
configuration.load_test_config()
self.dag = DAG(
'test_dag',
default_args={
'owner': 'airflow',
'start_date': DEFAULT_DATE},
schedule_interval=INTERVAL)
self.addCleanup(self.dag.clear)
self.clear_run()
self.addCleanup(self.clear_run)
def tearDown(self):
super(PythonOperatorTest, self).tearDown()
session = Session()
session.query(DagRun).delete()
session.query(TI).delete()
print(len(session.query(DagRun).all()))
session.commit()
session.close()
for var in TI_CONTEXT_ENV_VARS:
if var in os.environ:
del os.environ[var]
def do_run(self):
self.run = True
def clear_run(self):
self.run = False
def is_run(self):
return self.run
def _assertCallsEqual(self, first, second):
self.assertIsInstance(first, Call)
self.assertIsInstance(second, Call)
self.assertTupleEqual(first.args, second.args)
self.assertDictEqual(first.kwargs, second.kwargs)
def test_python_callable_arguments_are_templatized(self):
"""Test PythonOperator op_args are templatized"""
recorded_calls = []
task = PythonOperator(
task_id='python_operator',
# a Mock instance cannot be used as a callable function or test fails with a
# TypeError: Object of type Mock is not JSON serializable
python_callable=(build_recording_function(recorded_calls)),
op_args=[
4,
date(2019, 1, 1),
"dag {{dag.dag_id}} ran on {{ds}}."
],
dag=self.dag)
self.dag.create_dagrun(
run_id='manual__' + DEFAULT_DATE.isoformat(),
execution_date=DEFAULT_DATE,
start_date=DEFAULT_DATE,
state=State.RUNNING
)
task.run(start_date=DEFAULT_DATE, end_date=DEFAULT_DATE)
self.assertEqual(1, len(recorded_calls))
self._assertCallsEqual(
recorded_calls[0],
Call(4,
date(2019, 1, 1),
"dag {} ran on {}.".format(self.dag.dag_id, DEFAULT_DATE.date().isoformat()))
)
在终端中，当我运行nosetests test/my_test.py时，测试失败，因为Jinja模板未正确呈现。完整的日志如下。
======================================================================
FAIL: Test PythonOperator op_args are templatized
----------------------------------------------------------------------
Traceback (most recent call last):
File "/home/user/my_module/tests/my_test.py", line 120, in test_python_callable_arguments_are_templatized
"dag {} ran on {}.".format(self.dag.dag_id, DEFAULT_DATE.date().isoformat()))
File "/home/user/my_module/tests/my_test.py", line 88, in _assertCallsEqual
self.assertTupleEqual(first.args, second.args)
AssertionError: Tuples differ: (4, datetime.date(2019, 1, 1), 'dag {{dag.dag_id}} ran on {{ds}}.') != (4, datetime.date(2019, 1, 1), 'dag test_dag ran on 2016-01-01.')
First differing element 2:
'dag {{dag.dag_id}} ran on {{ds}}.'
'dag test_dag ran on 2016-01-01.'
- (4, datetime.date(2019, 1, 1), 'dag {{dag.dag_id}} ran on {{ds}}.')
? ^^ --------- ^^^^^^
+ (4, datetime.date(2019, 1, 1), 'dag test_dag ran on 2016-01-01.')
? ^^^^^ ^^^^^^^^^^
-------------------- >> begin captured logging << --------------------
airflow.utils.log.logging_mixin.LoggingMixin: INFO: Reading the config from /home/user/airflow/airflow.cfg
airflow.settings: INFO: Configured default timezone <Timezone [UTC]>
airflow.logging_config: DEBUG: Unable to load custom logging, using default config instead
--------------------- >> end captured logging << ---------------------
但是，my_test.py中的代码没有任何问题，因为它只是test_python_operator.py（v1-10-stable分支）中文件Airflow github repository的子集。天真地，我希望此测试可以正常运行，但不能运行。
我想念什么？
编辑：我正在使用apache-airflow 1.10.2，Python 3.6.8和机头1.3.7。
答案 0 :(得分：2)
这是因为，在{strong> Airflow 1.10.2 中，'op_args'，'op_kwargs'字段不是模板字段。问题中的链接是Airflow信息库的master分支。
PythonOperator，'op_args'。
将这些字段包括到'op_kwargs'的提交（这仍然在主版本中，没有包含在任何发行版本中）：https://github.com/apache/airflow/commit/7ab245b296efc73db3ce4ce0edbae473e357698c
对于Airflow 1.10.2：检查此文件-https://github.com/apache/airflow/blob/1.10.2/tests/operators/test_python_operator.py
此外，请勿使用template_fields分支，因为它包含即将发布的版本 1.10.3 的代码。您应该改用 1.10.2 标记：https://github.com/apache/airflow/tree/1.10.2
PythonOperator （1.10.2）： https://github.com/apache/airflow/blob/1.10.2/airflow/operators/python_operator.py#L65
v1-10-stablePythonOperator （母版-开发分支）： https://github.com/apache/airflow/blob/master/airflow/operators/python_operator.py#L72
class PythonOperator(BaseOperator):
template_fields = ('templates_dict',)
template_ext = tuple()
ui_color = '#ffefeb'
@apply_defaults
def __init__(
self,
python_callable,
op_args=None,
op_kwargs=None,
provide_context=False,
templates_dict=None,
templates_exts=None,
*args, **kwargs):
...
答案 1 :(得分：0)
您很可能使用1.10或更早的版本来运行测试。在该版本中，op_args中的PythonOperator没有被模板化。但是在master中，您很可能正在使用op_args的测试模板，并对其进行了相应的测试。如果您确实要使用气流测试作为示例，则应从与您安装的版本相对应的任何分支中获取它们。