Retain field names on broken references + restorable analysis version history

Q&A
, ,
1

The problem (real scenario):

I recreated a Redshift view (prod.waterfall) that feeds a QuickSight dataset. Because changing one column’s type required a DROP + CREATE rather than CREATE OR REPLACE, the recreate cascaded into several QuickSight issues — but the most painful one was this:

Every visual and filter that referenced the view lost its field bindings and showed only --Missing field--, with no indication of what the original field was. A pivot table showed blank “missing” slots; multiple filters silently broke. With no field name shown, I had no way from the UI to know what each well or filter was supposed to point at.

The only way I could recover the mapping was to pull the analysis definition via CLI (describe-analysis-definition) and read the Column.ColumnName values out of the raw JSON — manually tracing visual IDs and filter IDs to reconstruct what every broken reference originally was. It took hours for me to fix this issue.

Can you escalate this issue to the Dev team :

Request 1 - Show the field name on a broken reference.

Request 2 - Restorable analysis version history that survives dataset changes

2

Hi @snarayanan,

Nice seeing you back in the community! Thanks so much for those suggestions, as they would definitely be useful in mitigating ramifications of data reconstruction. I will make sure to mark this post as a Feature Request and archive it so that someone from the AWS Team may address it in the future.

Thank you!