Airflow - override()

 

"""
Example DAG demonstrating a workflow with nested branching. The join tasks are created with
``none_failed_min_one_success`` trigger rule such that they are skipped whenever their corresponding
branching tasks are skipped.
"""from __future__ import annotationsimport pendulumfrom airflow.providers.standard.operators.empty import EmptyOperator
from airflow.sdk import DAG, TriggerRule, taskwith DAG(dag_id="example_nested_branch_dag",start_date=pendulum.datetime(2021, 1, 1, tz="UTC"),catchup=False,schedule="@daily",tags=["example"],
) as dag:@task.branch()def branch(task_id_to_return: str) -> str:return task_id_to_returnbranch_1 = branch.override(task_id="branch_1")(task_id_to_return="true_1")join_1 = EmptyOperator(task_id="join_1", trigger_rule=TriggerRule.NONE_FAILED_MIN_ONE_SUCCESS)true_1 = EmptyOperator(task_id="true_1")false_1 = EmptyOperator(task_id="false_1")branch_2 = branch.override(task_id="branch_2")(task_id_to_return="true_2")join_2 = EmptyOperator(task_id="join_2", trigger_rule=TriggerRule.NONE_FAILED_MIN_ONE_SUCCESS)true_2 = EmptyOperator(task_id="true_2")false_2 = EmptyOperator(task_id="false_2")false_3 = EmptyOperator(task_id="false_3")branch_1 >> true_1 >> join_1branch_1 >> false_1 >> branch_2 >> [true_2, false_2] >> join_2 >> false_3 >> join_1

 

 

1️⃣ What override() does

• branch is a TaskFlow task (Python function decorated with @task.branch()).

• .override() is a method on TaskFlow tasks that lets you override some parameters of the task without changing the original function.

• Common things you can override:

  • task_id (unique identifier of the task in the DAG)

  • retries

  • retry_delay

  • execution_timeout

  • doc_md, etc.

Essentially, it creates a new task object with the same Python callable but updated configuration.

---

2️⃣ Why it's used here

branch.override(task_id='branch_1')

• The original function is branch(task_id_to_return: str).

• We want to instantiate it as a DAG task with a specific task_id called 'branch_1'.

• Without override(), the task would take its default function name as the task_id (here, 'branch'), which may conflict if you create multiple instances.

---

3️⃣ Putting it together

branch_1 = branch.override(task_id='branch_1')(task_id_to_return='true_1')

Step by step:

1. branch.override(task_id='branch_1') → creates a TaskFlow task object with task_id='branch_1'.

2. (task_id_to_return='true_1') → calls the task with the argument 'true_1' and returns a TaskInstance that Airflow can schedule.

So branch_1 is a DAG task instance with a custom task_id, not just a Python function call.

---

✅ Analogy

Think of override() as:

“I like this function, but for this DAG task, I want it to have a different name or different execution settings.”

• Original function: branch

• DAG task: branch_1 (with overridden task_id)