Don't recalculate derived properties until next access about ampersand-state HOT 4 OPEN

The only issue I can think of is change listeners - any ideas on this? Should we be checking for the presence of a change or change:${derivedKey} listener, then revert to the current behavior if present?

from ampersand-state.

The only issue I can think of is change listeners - any ideas on this? Should we be checking for the presence of a change or change:${derivedKey} listener, then revert to the current behavior if present?

Are there many situations where you would have a derived property on a model, but would have no change/change:${derivedKey} listeners on that model at all?

from ampersand-state.

I'm not going to suggest that this has as nice an API as a derived property does - I've not tried to abstract it in any way - but an alternative way of doing this if you don't need the change events, and just want a cached getter, is to just create a cached getter, using Object.defineProperty on the prototype, and something like lodash.memoize for the caching.

If you do need the change events, then I'm not sure that you can avoid the current behaviour.

Here's an example:

http://esnextb.in/?gist=acf4e2cae3bab5eae5f4

from ampersand-state.

Yes - I don't use change listeners at all, I'm using ampersand-state with a Flux pattern so I know exactly when changes have been made.

I often use derived properties to mask and cache expensive operations that are not used often - until now I had assumed that recalculation was only done at access, not if any dep had changed.

It should be relatively simple to record when change handlers have been attached. Can you think of any issues with this approach?

from ampersand-state.