Unifying ID datatypes and values

[jmdelfa]

There are three different data types to represent activity ids:
SimulatedActivityId
ActivityDirectiveId
SchedulingActivityDirectiveId

They all are wrappers of the long basic type
These values might be different.

I propose this discussion to talk about potential re-engineering to unify/simplify the datatypes and the id values

[jmdelfa]

Matthew L Dailis
A couple topics covered:
SimulatedActivityId, DirectiveId, and SchedulingActivityDirectiveId are difficult to work with - especially when we need to keep multiple maps between them in sync
Juan to use Apache BidiMap to avoid needing to maintain bidirectional maps by hand
Juan suggests we collapse these ids - three is too many
Anchors are asymmetric - one activity points at another
It may be desirable to add a UI feature that allows moving an entire group of activities by dragging any of its constituent parts. This would likely need to be in addition to the status quo - which states that moving an anchored activity changes its start offset without affecting other activities

Theresa Kamerman
:raccoon-1: 15 minutes ago
Re: the first topic: #734. And, as discussed, collapsing SimulatedActivityId and DirectiveId is not possible as there is not a 1-1 mapping (there may be directives that weren’t simulated due to simulation bounds, and decomposed child spans will have SimulatedActivityIds without a DirectiveId). DirectiveId and SchedulingDirectiveId may potentially be collapsible, though that would require other refactors (edited)

Adrien H Maillard
9 minutes ago
ActivityDirectiveId originate from the DB auto increment. But the scheduler is creating new activities. Every time you would want to create an activity, you’d have to go the whole way up to worker/server, make a DB request, then get back down in the scheduler. And then maybe this activity is finally removed from the plan (because of backtrack let’s say) so you have to make another request to cancel the previous one.

Adrien H Maillard
5 minutes ago
Or we provide a very restricted PlanService to the scheduler to do this kind of things.

Theresa Kamerman
:raccoon-1: 3 minutes ago
TBF we do have a 1:1 mapping for DB directive ids to scheduling directive ids, plus we already create the rest of the mapping upon insertion to the DB so we can update spans

Juan Delfa
< 1 minute ago
https://github.com/NASA-AMMOS/aerie-team/discussions/24

Adrien H Maillard
1 minute ago
Yes, this happens after scheduling and allows the scheduler to be self-contained. Analogy would be having simulation creating db spans while simulating to get rid of id maps

[jmdelfa]

btw, I think the discussion covers two topics: one is to explore the possibility of unifying datatypes and the second to explore the possibility of having less different id values for any given activity