Yes it seems to. Req'ing changes so we can chat before landing

One last question on this one: does it totally eliminate free text search?

cool. some minor feedback

We need to consider what this means from a product perspective. This diff leaves the world with overlapping notions of step subset selection that leaves the world in a more confused state.

got it! that all makes sense

up

bit concerned about bundling up behavior change here

up

upmessage

just bouncing to you for queue management while you investigate the weird isort issue per convo in slack

Cool. Can you include a test case that demonstrates when this is hit and then assert against the error message?

up

try optional

Abandoning this for now to let to run state machine play out

Ok this looks good then. Do we have a test to ensure that this works? e.g. that the notebook is actually somewhere where a downstream solid can access it? Also the materialization should still exist even if the output isn't specifed.

This is a definite improvement. The one thing we need to think about here is the the FileManager stuff as current written is meant for ephemeral files, meaning that the files will be deleted after the pipeline is done running. That is probably not what we want here (and we should test for that). (see local_file_manager)

lgtm

Definitely think we need to be flexible on data migrations. We can't *require* upgrading all rows for all cases.

Also we definitely need to do a pass on improving error messages in general

most excellent

upmessage

we should also talk to dwall about his dbt integration at some point to see if we can upstream

up

up

up

up

up

up

haha

will do refactor in followup

up

up

upmessage

rebase

rebase

up

👍🏻

upmessage

I am going to combine this with https://dagster.phacility.com/D2393 because I think it is actually easier to review together

up

up

So here is the version that just uses the value as the hash. The thing I liked about the previous scheme is that I could cache the hash value and it doesn't consume that much memory. But I don't feel too strongly

upmessage

up

upmessage

up

up

up

👍🏻

up

👍🏻

up

upmessage

up

up

Current state of things the error in the list case is a bit rough but I don't really think we should fix it. We may want to add a specialized error message for this case in the config validation code path (which would apply to non-default-value cases too).

tests

ya go for it