Hey, @schrockn, sorry I just realized I wont be back until later in the day on Friday. Anytime between 1-5pm should work.

np

updated to 3 pm

lol well that’s one way to do it

oh its gets better than that

if you have a function that returns an array of dags

there is no way to let the system know that it’s an array of dags

so you have to loop over them and inject them into machine-generated variable names in global scope

__globals[str(uuid.uuid4()))] = dag

or whatever

its amazing

there's some real gems in the docs

wow… that’s wild

Ya I know the founder

Had lunch with him a few months back

Once we have confidence in our airflow story will reach out

from the Airflow meetup -- this is an ING guy's example of a "common pattern" that they tried to abstract out of their existing airflow DAG codebase

this is that team juryrigging some very familiar machinery on top of airflow

set the channel topic: Project announcements and general conversation

state of airflow testing: https://medium.com/wbaa/datas-inferno-7-circles-of-data-testing-hell-with-airflow-cef4adff58d8, https://blog.usejournal.com/testing-in-airflow-part-1-dag-validation-tests-dag-definition-tests-and-unit-tests-2aa94970570c

i feel like we should meet this ING crowd

^^ i am thinking about the above in the context of an "Excel+" style of GUI-supported pipeline & solid authoring, where the solid inputs/outputs could be defined in a form and the body could be defined using a controlled, sandboxed subset of python like the above

this weekend an analyst was showing me anaplan