dstandish commented on a change in pull request #20530: URL: https://github.com/apache/airflow/pull/20530#discussion_r776416984
########## File path: airflow/decorators/sensor.py ########## @@ -0,0 +1,93 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +from inspect import signature +from typing import Any, Callable, Dict, Optional, Tuple + +from airflow.decorators.base import get_unique_task_id, task_decorator_factory +from airflow.sensors.base import BaseSensorOperator + + +class DecoratedSensorOperator(BaseSensorOperator): + """ + Wraps a Python callable and captures args/kwargs when called for execution. + + :param python_callable: A reference to an object that is callable + :type python_callable: python callable + :param task_id: task Id + :type task_id: str + :param op_kwargs: a dictionary of keyword arguments that will get unpacked + in your function (templated) + :type op_kwargs: dict + :param op_args: a list of positional arguments that will get unpacked when + calling your callable (templated) + :type op_args: list + :param kwargs_to_upstream: For certain operators, we might need to upstream certain arguments + that would otherwise be absorbed by the DecoratedOperator (for example python_callable for the + PythonOperator). This gives a user the option to upstream kwargs as needed. + :type kwargs_to_upstream: dict + """ + + template_fields = ('op_args', 'op_kwargs') + template_fields_renderers = {"op_args": "py", "op_kwargs": "py"} + + # since we won't mutate the arguments, we should just do the shallow copy + # there are some cases we can't deepcopy the objects (e.g protobuf). + shallow_copy_attrs = ('python_callable',) + + def __init__( + self, + *, + python_callable: Callable, + task_id: str, + op_args: Tuple[Any], + op_kwargs: Dict[str, Any], + **kwargs, + ) -> None: + kwargs.pop('multiple_outputs') + kwargs['task_id'] = get_unique_task_id(task_id, kwargs.get('dag'), kwargs.get('task_group')) + self.python_callable = python_callable + # Check that arguments can be binded + signature(python_callable).bind(*op_args, **op_kwargs) + self.op_args = op_args + self.op_kwargs = op_kwargs + super().__init__(**kwargs) + + def poke(self, context: Dict) -> bool: + return self.python_callable(*self.op_args, **self.op_kwargs) + + +def sensor(python_callable: Optional[Callable] = None, multiple_outputs: Optional[bool] = None, **kwargs): + """ + Wraps a function into an Airflow operator. + + Accepts kwargs for operator kwarg. Can be reused in a single DAG. + + :param python_callable: Function to decorate + :type python_callable: Optional[Callable] + :param multiple_outputs: if set, function return value will be + unrolled to multiple XCom values. List/Tuples will unroll to xcom values + with index as key. Dict will unroll to xcom values with keys as XCom keys. + Defaults to False. Review comment: > As discussed in [#20546 (comment)](https://github.com/apache/airflow/pull/20546#issuecomment-1002476070) - I believe it's not going to work if we return Tuple (because the tuple is always "truthy" and it will work very unexpectedly if the sensors are used with old Airflow version - they will never wait) Yeah I agree that that PR as is would be problematic and would need changes before merge. I look forward to collaborating on figuring out the best way to do this. I think we should make it happen before we introduce this taskflow sensor because depending on our choices we could be (and might have to be) stricter with the contract re return types in taskflow sensor and don't want to have to change it after release. I did have a concern with your truthy-based return namely pandas dataframes don't behave, and commented on that pr. There's a third way demonstrated in https://github.com/apache/airflow/pull/20547 which introduces a PokeReturnValue class which is an optional return value which can resolve this ambiguity problem. We could have the taskflow sensor behave in such a way that tuple or list is converted to a PokeReturnValue instance and then the base sensor can know what to do with it. Lemme know your thoughts. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
