I've started using derived-atom
as a way to aggregate stuff locally so that that has a reaction before deciding to re-render a component - see https://github.com/CmdrDats/rumperftest/blob/master/src/perftest.cljs#L36-L40 - I'm a bit worried about memory leak with this pattern - will the derived atom watchers get released when the component is no longer in scope? or is there a different way I should be doing this?
@cmdrdats The pattern you describe is known (part of optimization of a component) and the issue you describe is also known as @roman01la said above.
Personally I use a hook that does the cleanup
Something along the lines: https://github.com/g7s/rum/blob/hooks/src/rum/core.cljs#L572
@g7s great, thanks for the feedback - use-derived-atom
looks interesting - is this in a fork?
Yes it is in a fork that I had made long ago.
ah, ok - I see the current rum code has:
(defn use-reducer [reducer-fn initial-value]
[initial-value reducer-fn])
(defn use-effect!
([setup-fn])
([setup-fn deps]))
(defn use-callback
([callback] callback)
([callback deps] callback))
probably in prep for moving away from React?
I guess a similar effect might be achieved with a mixin if there's a registry of derived atoms against a component?
Yeah exactly you can do that with a mixin as you describe
there's a related issue for derived atom and watchers https://github.com/tonsky/rum/issues/94
Thanks :) i did see this, but my question is a bit broader - would the pattern I'm using be considered as intended, or should I be doing something entirely different? I find this particularly when dealing with collections where I need to iterate through the collection to embed their components, but then if I just react on the collection itself, any change (obviously) rerenders the entire parent component