Is there a way to return a

RunRequest

that does nothing? It seems that a sensor's cursor is advanced only if a RunRequest is returned. However, we have a few use cases where we'd like to run a function (slack notification) in response to a multi-asset sensor, but have no associated jobs or assets to run.

Does

context.update_cursor

not work if you don't yield a

RunRequest

? I was under the impression that it does, regardless of if you yield a SkipReason or a RunRequest

been a while since i’ve been deep in sensor code, but i think

context.update_cursor

should do the trick

hmm i've been calling

advance_all_cursors

, is that different from

context.update_cursor

?

i think that should be the same taking a step back - what are you trying to do with the sensor? would a run status sensor do the trick so that you dont need to manually modify the context yourself?

the goal is to provide an interface for users to subscribe to events within the data pipeline, where an event is a combination of several assets being materialized at least once. I've implemented this using the

multi_asset_sensor

decorator to create a sensor that listens on the configured set of assets. Each sensor tick, if all of the assets have materialized at least once since the last time the sensor fired, it will execute a a user-provided function (with the expectation that the function is lightweight, disposable, stateless, etc... primarily will be notification-related), plus optionally returning a

RunRequest

to initiate runs of the configured jobs or request assets for that multi-asset sensor. It actually works fine if someone configures a job or request asset, but I noticed that when neither is provided, the user-provided function executes every tick if all of the monitored assets have materialized at least once, and none of the monitored assets have materialized since the last time that all of them were materialized.

ah interesting. Does returning a

SkipReason

in that case not work?

ah, it does! so semantically a SkipReason is sort of a "do nothing but commit cursor state" signal...

yeah i think so. probably good to do a double check on cursor state - but i’m like 99% sure that’s what it does

Just checked, the cursor does get updated, i think the main feedback is that "Skip" is not really the behavior more like... "Pass"?

yeah, i suppose the idea is like “skip launching any runs”