Skip to main content

Machine Learning Orchestration

Project description

Dbnd Airflow Operator

This plugin was written to provide an explicit way of declaratively passing messages between two airflow operators.

This plugin was inspired by AIP-31. Essentially, this plugin connects between dbnd's implementation of tasks and pipelines to airflow operators.

This implementation uses XCom communication and XCom templates to transfer said messages. This plugin is fully functional, however as soon as AIP-31 is implemented it will support all edge-cases.

Fully tested on airflow 1.10.X.

Code Example

Here is an example of how we achieve our goal:

import logging
from typing import Tuple
from datetime import timedelta, datetime
from airflow import DAG
from airflow.utils.dates import days_ago
from airflow.operators.python_operator import PythonOperator
from dbnd import task

# Define arguments that we will pass to our DAG
default_args = {
    "owner": "airflow",
    "depends_on_past": False,
    "start_date": days_ago(2),
    "retries": 1,
    "retry_delay": timedelta(seconds=10),
}
@task
def my_task(p_int=3, p_str="check", p_int_with_default=0) -> str:
    logging.info("I am running")
    return "success"


@task
def my_multiple_outputs(p_str="some_string") -> Tuple[int, str]:
    return (1, p_str + "_extra_postfix")


def some_python_function(input_path, output_path):
    logging.error("I am running")
    input_value = open(input_path, "r").read()
    with open(output_path, "w") as output_file:
        output_file.write(input_value)
        output_file.write("\n\n")
        output_file.write(str(datetime.now().strftime("%Y-%m-%dT%H:%M:%S")))
    return "success"

# Define DAG context
with DAG(dag_id="dbnd_operators", default_args=default_args) as dag_operators:
    # t1, t2 and t3 are examples of tasks created by instantiating operators
    # All tasks and operators created under this DAG context will be collected as a part of this DAG
    t1 = my_task(2)
    t2, t3 = my_multiple_outputs(t1)
    python_op = PythonOperator(
        task_id="some_python_function",
        python_callable=some_python_function,
        op_kwargs={"input_path": t3, "output_path": "/tmp/output.txt"},
    )
    """
    t3.op describes the operator used to execute my_multiple_outputs
    This call defines the some_python_function task's operator as dependent upon t3's operator
    """
    python_op.set_upstream(t3.op)

As you can see, messages are passed explicitly between all three tasks:

  • t1, the result of the first task is passed to the next task my_multiple_outputs
  • t2 and t3 represent the results of my_multiple_outputs
  • some_python_function is wrapped with an operator
  • The new python operator is defined as dependent upon t3's execution (downstream) - explicitly.

Note: If you run a function marked with the @task decorator without a DAG context, and without using the dbnd library to run it - it will execute absolutely normally!

Using this method to pass arguments between tasks not only improves developer user-experience, but also allows for pipeline execution support for many use-cases. It does not break currently existing DAGs.

Using dbnd_config

Let's look at the example again, but change the default_args defined at the very top:

default_args = {
    "owner": "airflow",
    "depends_on_past": False,
    "start_date": days_ago(2),
    "retries": 1,
    "retry_delay": timedelta(minutes=5),
    'dbnd_config': {
        "my_task.p_int_with_default": 4
    }
}

Added a new key-value pair to the arguments called dbnd_config

dbnd_config is expected to define a dictionary of configuration settings that you can pass to your tasks. For example, the dbnd_config in this code section defines that the int parameter p_int_with_default passed to my_task will be overridden and changed to 4 from the default value 0.

To see further possibilities of changing configuration settings, see our documentation

Project details


Release history Release notifications | RSS feed

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

dbnd-airflow-operator-0.25.6.tar.gz (16.6 kB view details)

Uploaded Source

Built Distribution

dbnd_airflow_operator-0.25.6-py2.py3-none-any.whl (21.3 kB view details)

Uploaded Python 2 Python 3

File details

Details for the file dbnd-airflow-operator-0.25.6.tar.gz.

File metadata

  • Download URL: dbnd-airflow-operator-0.25.6.tar.gz
  • Upload date:
  • Size: 16.6 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/3.1.1 pkginfo/1.5.0.1 requests/2.22.0 setuptools/42.0.2 requests-toolbelt/0.9.1 tqdm/4.41.1 CPython/3.6.3

File hashes

Hashes for dbnd-airflow-operator-0.25.6.tar.gz
Algorithm Hash digest
SHA256 1ddef69219641c9016ad85d1635195a2da2e321cd74603249274cd92a0b43744
MD5 00ae1b165ff6e93449f67cbbb45b321f
BLAKE2b-256 06b974273c8550529873b7d9d088b9e2adb1bf584de230dd11aa28269810aaca

See more details on using hashes here.

File details

Details for the file dbnd_airflow_operator-0.25.6-py2.py3-none-any.whl.

File metadata

  • Download URL: dbnd_airflow_operator-0.25.6-py2.py3-none-any.whl
  • Upload date:
  • Size: 21.3 kB
  • Tags: Python 2, Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/3.1.1 pkginfo/1.5.0.1 requests/2.22.0 setuptools/42.0.2 requests-toolbelt/0.9.1 tqdm/4.41.1 CPython/3.6.3

File hashes

Hashes for dbnd_airflow_operator-0.25.6-py2.py3-none-any.whl
Algorithm Hash digest
SHA256 1b146935aca9dd93deccb2b4fb6e91e2946c73daab5a13ceefed8f769642f2fa
MD5 edec8295fc9c29679173c54b5f009263
BLAKE2b-256 bda77054674f27150414d36fa5e3ab15e2fae6e5a7aa760b998c5780246f1bbb

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page