Wow that is an incredible relief. I was pretty sure when that ticket closed that you would go the other way and assert on all retractions, including the ones you missed

Many thanks, this makes our lives much easier

There was a lot of discussion around this. Please let us know after you have used it in anger your thoughts.

Will do

I will tell you that having had to hammer existing values to fit into a reasonable predicate has not been fun

Sorry about that. Hopefully this works much better.

attribute predicates were applied on retractions too? or what was the issue here?

Yes, they were applied on retractions too, mostly

what problems did it cause?

If you have existing data that violates a predicate, you can’t add the predicate you want and then get rid of bad data. You have to either compromise the predicate to fit the bad data you have (possibly allowing new bad data); or you have to get rid of the bad data first, install the predicate, then check your data again (retractions of bad data may fail in the meantime), and possibly remove the predicate and repeat.

got it, thx

I wonder why they didn't took the approach of "all data must satisfy the predicate before the attribute predicate can be added".

That would be a (potentially long) blocking transaction

If you don't mind, besides HA, does clubhouse runs more than one transactor to serve customers?

only one, for now. We’re working on sharding.