Closed sgratiy closed 2 years ago
@sgratiy FYI, there was a repo configuration setting that was not running the CI for external first-time contributors. I've been told that that issue is now fixed so if you do a force push or something like that it should now properly run the CI.
Hey @berl, @neuromusic could you please review this bugfix.
@sgratiy sorry this needs more careful checking of downstream impacts. Can you verify that switching the dimensions in the DecodedIntensityTable
doesn't impact anything? Even if there's no test coverage, this would likely introduce a breaking change in the output that we use later.
@berl I have added the notebook demonstrating the problem in the main description.
thanks @sgratiy.
For the record, this was indeed a bug in how starfish/core/spots/DetectPixels/combine_adjacent_features.py
created a DecodedIntensityTable
. This was previously missed because users (and the starfish codebase) never did anything with the IntensityTable
part of a DecodedIntensityTable
created in this way. I'll revert the revert and we can pretend like this never happened.
This PR fixes a bug in with the
CombineAdjustFeatures.run
that was flipping round and channel coordinates in the resultingDecodedIntensityTable
compare to the input intensities.The fix: Use explicit parameter calls when building table coordinates Correct shape definition for the IntensityTable docstring
This notebook shows demonstrates the intensity table coordinates before and after the fix.