what's the role of index here? it doesn't seem to be tracking the end of the array, but it looks like it could intermix the ids returned in order from different descendants.

we don't want to push these to the end of the array, right? if we insert 5 descendant ids, do we want to add the next descendants inside of those, or do we want to increment index by 5 or 6 instead?

We want to also search all the newly added clientIds for descendants, so we shouldn't increment. index just keeps increasing until we looped over all descendants and nested descendants.

so it doesn't matter in which order we add the descendent ids to the list?
did you consider using ids.push( ...order )?

It does matter :) I used push at first, but the tests were expecting the ids in a certain order (descendants right after parent ID vs after all the parent siblings

okay I'm confused then, because it looks like we only get descendants immediately after their parents if there are only one descendant per parent

It's deeply adding all descendants, we keep expanding the array as we find more, so the loop will also include those newly added descendants

Not strictly related to your changes, but I've always thought it's awkward when writing application code that we have both of these functions. They're named similarly and I can never remember which is which. Could we deprecate getClientIdsWithDescendants and make it so that getClientIdsOfDescendants defaults to rootIds = ''?

Yes, it's really weird that we have both of these functions, I can deprecate it in a follow-up

I think using a stack (while ( id = stack.pop() )) for this would be easier to grok. But I suppose you'd have to push items onto the stack in reverse order which is probably slow.

Carry on 👍