access input metadata added by add_input_metadata #17908
Unanswered
Replies: 1 comment 2 replies
-
I think this is a feature we will need to add. I made an issue for it here #17923 |
Beta Was this translation helpful? Give feedback.
2 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
I can add some metadata to the asset UI using the
OutputContext.add_output_metadata
method in thehandle_output
method of a custom IO manager.Is it possible to do the same in the
load_input
method (I would guess to the downstream tasks)? I have tried adding that metadata using theInputContext.add_input_metadata
method thinking that would expose the metadata to the UI but no luck. Where does the input metadata go when you useadd_input_metadata
and where can you access this from?While i’ve been digging I saw some discussion that said there were issues or problems with this in partitioned assets, though I am not using partitioned assets for this particular case. There seemed to be a handful workarounds, but I’d like to understand what the
add_input_metadata
is doing and how that “should” be used to see if that would work for me before opting to create a workaround.The question was originally asked in Dagster Slack.
Beta Was this translation helpful? Give feedback.
All reactions