If we export the update tags that are used internally it'll be easier to document what they're for and people will be less likely to make a typo which is hard to debug since it's used as an extensible Set<string> and there is no way a dev mode would be able to detect 'unregistered' use since there is no registration process like commands.
Some examples in the monorepo that I'm aware of:
historic
history-push
history-merge
paste
collaboration
skip-collab
skip-scroll-into-view
It might also be a good time to think about how and whether updates with different tags should really be merged, since you could have two updates in the same tick with very different semantics. It might be the case that for correct behavior they should be reconciled separately at least if they were not part of the same top-level update. It seems reasonable that cascading transforms and any other sort of explicit nested update should inherit tags but if the updates were completely separate it might not make sense. I think this sort of bug would be very hard to catch and reproduce, and tags are probably used rarely enough that it wouldn't be a performance impact to reconcile more than once in a tick (iff updates with different tag sets are to be combined) to avoid this kind of incoherence.
If we export the update tags that are used internally it'll be easier to document what they're for and people will be less likely to make a typo which is hard to debug since it's used as an extensible
Set<string>and there is no way a dev mode would be able to detect 'unregistered' use since there is no registration process like commands.Some examples in the monorepo that I'm aware of:
It might also be a good time to think about how and whether updates with different tags should really be merged, since you could have two updates in the same tick with very different semantics. It might be the case that for correct behavior they should be reconciled separately at least if they were not part of the same top-level update. It seems reasonable that cascading transforms and any other sort of explicit nested update should inherit tags but if the updates were completely separate it might not make sense. I think this sort of bug would be very hard to catch and reproduce, and tags are probably used rarely enough that it wouldn't be a performance impact to reconcile more than once in a tick (iff updates with different tag sets are to be combined) to avoid this kind of incoherence.