Replacement between (identical) datasets leads to errors in analysis

I have a dataset A with a set of columns CS coming from a set of joined datasets DS1.
Then I create a dataset B with the same CS coming from a DS2. Replacing A with B in an analysis leads to:

  1. “Fields no longer in analysis” on all visuals
  2. Broken filters. All of them
  3. Calculated fields in the analysis refer to the datasets from DS1, i.e., they are prefixed with IDs of the datasets in DS1

I attach the screenshot of an issue that shows that all fields on a visual are valid after the replacement, but it still shows the error described in “1)”.

I’ve had a similar experience at times in the past and there are a few workarounds that have solved it for me but it requires a bit of rework. How many visuals are impacted? Start with the first visual and drag time, machine_id, measure_value_varchar, and running_count_of_errors_per_machine from your dataset on the left into the proper field wells. That basically replaces the current field well values with the updated fields from your replacement dataset.

Once you’ve done that for the first visual do the same for the filters that you have on that visual.

I assume all fields that are in your replacement dataset are exactly the same and in the same order as your A dataset. That should lead to this issue not occuring so it may require a support ticket to be submitted but I’ve had success with the above workaround in the past.

Thanks for your help. All visuals and filters are broken, in total around 25 elements. After replacing the fields with their identities on the left, I now see “The dataset changed too much”. Going further to filters on this visual,
seems there is no way to edit them, as the column to which a filter is applied can’t be changed (“Edit” on the screenshot below reopens the same view).
Deleting the filters returns the visual back, but the amount of “hacks” and time spent on this is not appropriate. Ideally, replacing with an identical dataset should work out-of-the-box…

Delete the filter and then add it again.

Yes, it fixes the visual (only this one), but that’s almost the same amount of work as doing the visual from scratch. Apart from positioning it on a sheet.

I recommend that you submit a support ticket to flag this issue as a bug because you shouldn’t need to recreate everything. That’s what I’ve done in the past because you shouldn’t need to do all this manual work when replacing a dataset.