danielballan
commented
1 year ago An unrelated recently-merged PR, https://github.com/bluesky/databroker/pull/638, broke main. That PR was so old that that the CI determination got stale. A lesson there is to re-run CI on old PRs.
The last 3 commits here clean that up.
Description
Update usage of
ArrayAdapter.Motivation and Context
Tiled made backward-compatible changes in https://github.com/bluesky/tiled/pull/537. The databroker adapter was not fully updated to account for this. The issue went un-caught because it only occurs on a codepath where a databroker handler returns a Dask Array. This is not tested.
How Has This Been Tested?
I tested this interactively by looking at two pieces of real beamline data:
...using current
main(broken) and this PR branch (currently deployed, works).There should be a unit test but that would actually be a pretty big addition to the current Databroker test harness, and we are in a rush to land this before the resumption of operations.