Airflow operators

Introduction to Apache Airflow in Python

Mike Metzger

Data Engineer

Operators

Represent a single task in a workflow
Run independently (usually)
Generally do not share information
Various operators to perform different tasks

@dag(
  dag_id="Example_Dag"
)
def example_dag():
  @task
  def task1():
    return "The result from task1"

  task1()

example_dag()

@task (PythonOperator)

Executes a Python function
Any Python function can be decorated with @task
Can pass data with other functions / tasks

from airflow.sdk import task

@task
def printme():
    print("This goes in the logs!")

printme()

@task arguments

Can pass arguments to task / function like an ordinary Python function

@task
def printme(name: str):
    print(f"Hi {name} - This goes in the logs!")


printme(name='DataCamp')


# Adds:
#   Hi DataCamp - This goes in the logs!
# to the Airflow logs

@task.bash (BashOperator)

@task.bash
def bash_example():
  return "echo 'Example!'"

bash_example()

@task.bash
def run_cleanup():
  return "runcleanup.sh"

run_cleanup()

Executes a given Bash command or script
Runs the command in a temporary directory
Can specify environment variables for the command

Task dependencies

Each Dag has a set of tasks to complete
Task dependencies specify running order
Different methods to specify dependencies

Airflow Dag of connected tasks showing dependency order between them

Bitshift syntax

>> and <<

task1() >> task2()

# task1 completes before task2 starts


task1() >> task2() >> task3()

# task1 completes before task2 and task2 completes before task3


task1() >> task3()
task2() >> task3()

# task1 and task2 can run together but both must complete before task3 runs

Bitshift syntax example

# Download sales data before reconciling
download_sales_data() >> reconcile()

# Download inventory data before reconciling
download_inventory_data() >> reconcile()

Both must complete before reconciling, but order isn't specified

Let's practice!

Introduction to Apache Airflow in Python