Details
-
Type:
Improvement
-
Status: Won't Fix
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: afw
-
Labels:None
-
Team:Data Release Production
Description
Currently, if there is a general failure reported in the flags of a measurement algorithm there is no convenient was to quickly scan all of the flags to determine what is causing the error. This is a request for a python convenience function that returns the names (or at least the enum/numbers) of all the set flags in a given schema.
Attachments
Activity
I thought this was a good idea. We have something similar for masks, and I find it extremely useful to reduce a value (or a pixel location) into a list of mask plane names. I think it would be helpful to ask for a source, "tell me everything that went wrong".
Paul Price
added a comment - I thought this was a good idea. We have something similar for masks, and I find it extremely useful to reduce a value (or a pixel location) into a list of mask plane names. I think it would be helpful to ask for a source, "tell me everything that went wrong". Assigned to JDS as a placeholder for now. We'll prioritize this on the backlog then see where we can fit it into the schedule.
Just thinking... I think it is would be nice to have three options for display.
1) a reduced bit pattern like output. This is how I found the problem in the centroid algorithm. The output looked like this.
1 0 0 1 0 0 0 0 0 0
1 0 0 1 0 0 0 0 0 0
1 0 0 1 0 0 0 0 0 0
1 0 0 1 0 0 0 0 0 0
1 0 0 1 0 0 0 0 0 0
1 0 0 1 0 0 0 0 0 0
1 0 0 1 0 0 0 0 0 0
...
2) A list of the names of flags turned on from the schema.
One entry might look like:
measBase.SdssShapeAlgorithm.SHIFT
3) A list of the names of the flags turns on with their description
An entry might look like:
measBase.SdssShapeAlgorithm.SHIFT - "centroid shifted by more than the maximum allowed amount"
Chris Walter
added a comment - Just thinking... I think it is would be nice to have three options for display.
1) a reduced bit pattern like output. This is how I found the problem in the centroid algorithm. The output looked like this.
1 0 0 1 0 0 0 0 0 0
1 0 0 1 0 0 0 0 0 0
1 0 0 1 0 0 0 0 0 0
1 0 0 1 0 0 0 0 0 0
1 0 0 1 0 0 0 0 0 0
1 0 0 1 0 0 0 0 0 0
1 0 0 1 0 0 0 0 0 0
...
2) A list of the names of flags turned on from the schema.
One entry might look like:
measBase.SdssShapeAlgorithm.SHIFT
3) A list of the names of the flags turns on with their description
An entry might look like:
measBase.SdssShapeAlgorithm.SHIFT - "centroid shifted by more than the maximum allowed amount" The DM-CCB discussed this issue on 2019-04-03. We agreed that this remains a feature request that we should maintain on the backlog, but not an immediate priority for implementation.
I'm calling this Won't Fix because we've moved to Parquet-based datasets for all user-facing tables, and I don't see us investing much future effort in convenience-code in afw.table as a result.
This problem may well arise there, too, but I suspect it will take a sufficiently different form that the current ticket isn't worth tracking.
This isn't exactly what you want, as it doesn't filter on whether a flag is set, but if e.g. base_SdssShape_flag is set, then you can print all the flags that might contribute to it via print record.extract("base_SdssShape_flag*"). That will return a name-value dict of all fields that match the given glob, and by strong convention any flag that could contribute to "base_SdssShape_flag" itself will be have a name that starts with that prefix.