up

up

think i got it

up

up

up

try

ya i think those last few naming bits are the only issues i can see

ill dig in on this

I didnt read close enough - communicate does wait for close. As seen by test failures - the mystery continues.

p sure this was me: i in fact can not spell

ah ok cool thanks

where is that in the diff i dont see any ref to dev null

how does this fix the catalina thing?

cool beans

but its not that much code were shuffling around so ill let you make the call

less is more

looks reasonable so far - think its worth updating this diff til it functions as a proper run launcher

In all of these dagit cases, we're fetching the pipeline in a sub-component where the container component has already fetched the pipeline.
We may hit an error case in which dagit gets into a bad state in between requests, or we hit a framework bug while resolving some downstream piece of data, but these should be extremely rare.

There is a deeper issue with the compute_log_manager that I decided to punt on debugging.

up

dont add new class

met to discuss this - @schrockn was able to help outline how using the dotted notation for all composite config (subselected or not) instead of the hierarchal structure or a hybrid would make sense. The example of vscode settings helps us believe this as likely a better end user experience.

fix dagoit

up

up

codepointer is a little awkward - cant think of anything better though. Maybe code_pointer?

https://dagster.phacility.com/D2709

I definitely think this is better than trying to re-derive the pipeline from the function args

up

Did you guys see the picture i drew in an earlier comment? I have a hard time reconciling a mental model that makes sense for subsetting in to a composite where maintaining the original config structure makes sense. I agree that this is a gnarly user experience - but I think its because subsetting in to a composite is fundamentally complex and that complexity is accurately exposed to the user. I think maintaining the config structure will mask the complexity causing greater confusion.

woah nice work on the comments - thats really nice. I would like @prha / @schrockn to also take a pass, mostly for opinion on the design choices highlighted in summary.

should be fine but would be nice to have the motivation written down in the diff summary

can we do this at ReconstructableRepo creation instead? Lets take the time to figure out how to have a test in place for this