Airflow Programming Model

From NovaOrdis Knowledge Base
Revision as of 01:46, 18 July 2022 by Ovidiu (talk | contribs) (→‎Combining @task and Operators)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

External

Internal

Overview

Airflow DAGs are programmed in Python.

Airflow DAG Examples

https://github.com/apache/airflow/tree/main/airflow/example_dags

requirements.txt

apache-airflow == 2.3.3

Declaring a DAG

With @dag Decorator

https://airflow.apache.org/docs/apache-airflow/stable/concepts/dags.html#the-dag-decorator

The @dag decorator turns a Python function into a DAG generator function. @dag is only available in Airflow 2 and newer. The decorator can be configured with the a number of pre-defined parameters. The function includes task declaration and task relationship declaration, as follows:

from airflow.decorators import dag, task
from datetime import datetime


@dag(
    schedule_interval=None,
    start_date=datetime(2022, 7, 13, 0),
    catchup=False
)
def some_dag():
    @task
    def task_a():
        print(">>> A")

    @task
    def task_b():
        print(">>> B")

    @task
    def task_c():
        print(">>> C")

    task_a() >> task_b() >> task_c()


dag = some_dag()

Also see:

Airflow Concepts | Declaring a DAG

With DAG() Constructor

Also see:

Airflow Concepts | Declaring a DAG

import time
from datetime import datetime, timedelta
from airflow import DAG
from airflow.operators.empty import EmptyOperator
from airflow.operators.python import PythonOperator

default_args = {
    'owner': 'somebody',
    "depends_on_past": False,
    "email": ["somebody@example.com"],
    "email_on_failure": True,
    "email_on_retry": False
}

dag = DAG('some_dag',
          max_active_runs=1,
          catchup=True,
          start_date=datetime(2022, 7, 13, 0),
          schedule_interval="* * * * *",
          dagrun_timeout=timedelta(minutes=2),
          default_args=default_args
          )


def some_function():
    print("something")


with dag:
    start = EmptyOperator(task_id='start', dag=dag)
    end = EmptyOperator(task_id='end', dag=dag)
    some_function = PythonOperator(
        task_id='some_function',
        python_callable=some_function,
        dag=dag
    )

    start >> some_function >> end

Declaring a Task

Using the PythonOperator

from airflow import DAG
from airflow.operators.python import PythonOperator

dag = DAG('some_dag', [...])

def some_function():
    print("something")

with dag:
    some_function = PythonOperator(
        task_id='some_function',
        python_callable=some_function,
        dag=dag
    )

    [...]

Using the @task Decorator

Airflow TaskFlow | Programming Model

Combining @task and Operators

from airflow.models.dag import DAG
from airflow.decorators import task
from airflow.operators.empty import EmptyOperator
from datetime import datetime

@task
def task_a():
    print(">>> task_a")

with DAG(
        dag_id='some_dag',
        start_date=datetime(2022, 7, 13, 0)
) as dag:
    start = EmptyOperator(task_id='start', dag=dag)
    end = EmptyOperator(task_id='end', dag=dag)

    @task
    def task_b():
        print(">>> task_b")

    start >> task_a() >> task_b() >> end

XComs Programming Model

Airflow XComs | Programming Model

Module Management

TO PROCESS: https://airflow.apache.org/docs/apache-airflow/2.3.2/modules_management.html?highlight=import