Hello guys, With airflow, we can chain different tasks/actions without, mandatorily, having outputs from one inputing the others. On Dagster however, all examples on the doc show pipelines with

op

functions that mandatorily follow this input/output paradigm. Am I required to generate outputs and inputs in my functions to create a graph/job in Dagster? Currently I want to orchestrate functions that don't generate output. Thank you all !

Hi Davi, I’m new here and to Dagster (for a question of my own) so I’ll let them correct me, but saw your question on login and I’m pretty sure with ops you can just have a function that doesn’t return anything and use that as

ins

to another

op

or use the functional dependency chaining of

@job

definitions if it’s dependent e.g.

op3(op2(op1()))

Example of what I mean by functional chaining https://docs.dagster.io/tutorial/ops-jobs/connecting-ops#a-more-complex-dag I’m using an

@asset

pipeline and for that type there’s a specific `non_argument_deps`param to

@asset

which tells it to depend on another

@asset

without requiring that asset’s output if it generates any - as per what I’ve read

@asset

s are built on top of

@op

so I imagine this is the way `@op`s can be used. Example of

non_argument_deps

https://docs.dagster.io/tutorial/assets/non-argument-deps#an-unzipped-csv-of-cereal-ratings Hope this helps and is potentially correct :)

@Mat Brady Thank you so much! Does it bother you to send here a snapcode of how you use the functional chaining on your project ? Thank you so much !

🤔 So you have something like this in code:

from dagster import asset, OpExecutionContext


@asset
def asset1():
    # do some stuff and don't return anything
    pass

@asset(non_argument_deps={'asset1'}, required_resource_keys={'api'})
def asset2(context: OpExecutionContext):
    # do some stuff and don't return anything
    pass

@asset(non_argument_deps={'asset2'}, required_resource_keys={'api'})
def asset3(context: OpExecutionContext):
    # do some stuff and don't return anything
    pass

if you're using ops, not assets, here's how to define dependencies without passing data: https://docs.dagster.io/concepts/ops-jobs-graphs/graphs#defining-nothing-dependencies

Ah! I was so close and yet so far 😄

@Mat Brady To avoid chaining to many parentheses on the job, you can do it like this as well(I prefer):

@op
def op_1():
    # Do something
    return Nothing


@op(ins={"depends": In(Nothing)})
def op_2():
    # Do something
    return Nothing


@op(ins={"depends": In(Nothing)})
def op_3():
    # Do something

@job
def job():
    dependency_link = op_1()
    dependency_link = op_2(dependency_link)
    op_3(dependency_link)