Accessing Airflow Params

Parent: Apache Airflow

---

You can acces airflow parameters either by Jinja Templating or kwargs. Either way works but there are important considerations. Remember that without render_template_as_native_obj=True jinja templating returns your params as strings, even if they’re set as integers.

This is the comparison

Kwargs:

	@task()
	def function(**context):
		params = context['params']
		
		return business_logic_function(params['district'])
		
	function()

At first glance this looks fine and even the Airflow documentation mentions this as the first method of passing DAG params to a function. But at a glance there are a few weaknesses here:

1. **context pulls all the dag params in every function, and it happens every time the dag gets parsed by Airflow
2. Its not visible at a glance which param gets passed to this function.
3. Less readable than the next alternative

Jinja Templating:

	@task()
	def function(district):
		params = context['params']
		
		return business_logic_function(district)
		
	function('{{params.district}}')
 

Now this is significantly better because:

1. Only the params used in this function is accessed
2. Its explicitly visible which params are required, both in the function definition and implementation
3. More readable

Specific to jinja

You need to pass render_template_as_native_obj=True in the @dag decorator, otherwise all params will be rendered as string

Related:

Writing clean code Refactoring

Resources: