Open rprechelt opened 8 months ago
@johannkm Do you know how feasible this is?
We tag runs with dagster/partition: <dim 1 value>|<dim 2 value>
, dagster/partition/<dim 1 name>: dim 1 value
, and dagster/partition/<dim 2 name>: dim 2 value
. The latter two should be enough to power this from the backend using the existing tags filter.
I think if we decide on how the new filter option should look, we can add this with only JS. A challenge with design might be that we can't really know if partition is a multi partition (e.g. there might be multiple partitions defs which share the same name and some are, some aren't).
@braunjj could we put this in the design queue? Happy to share more details as needed
What's the use case?
On the "Runs" page in the Dagster UI, there is the ability to filter the displayed runs by "Partition" 👍. However, in the case of a
MultiPartitionDefinition
, the UI does not allow for filtering by the value of only one dimension: the UI will not "select" unless I actually select a fullkey1|key2
partition key.It would be extremely to be able to filter on only
key1
and have all possible values ofdimension2
in the run view, and vice versa.For example, I have the following definition:
In the Dagster filter UI, I can only filter by
USA|20231212
(and see all historical runs for that country on that date), whereas it would be incredibly useful to be able to filter by justUSA
and see all runs for all dates with that country. Similarly, being able to filter on20231212
and see runs relating to "USA", "Australia", and "Canada" on that date.Ideas of implementation
The most generic solution would be to allow "fuzzy filtering" for partitions in the "Runs" filter UI (which covers all possible use cases). If this is not an option, the MVP would be adding the 1D partition keys to the list of available choices in the filter menu, e.g. in the example above, make "USA", "Australia", and "Canada" selectable options in the filter UI.
Additional information
No response
Message from the maintainers
Impacted by this issue? Give it a 👍! We factor engagement into prioritization.