Demonstrate dirty change tracking failure

[NullVoxPopuli]

Editing one value should not dirty other values.

This can be seen by editing any of the fields in the paused test, and seeing two logs print.
When this bug is fixed, only the log for the changed field should print.

[SergeAstapov]

Seem like in order to fix this, we should use smth like TrackedObject as underlying data structure…
however as it’s not automatically deeply autotracked, I was thinking we may need to switch to smth like graph for the underlying data structure where each graph edge would be of type Change and it wpuld allow us to get rid of using spread when changes gets applied/reverted

[NullVoxPopuli]

https://github.com/NullVoxPopuli/ember-deep-tracked/blob/main/ember-deep-tracked already solved this via Proxies :D

I think in order for ember-changeset to be performant and correct in change-tracking, we may need to lose the dependency on validated-changeset

[sukima]

I realized that dirty changes on a global standpoint is a big issue for asynchronous validations! If you have ten async validators and you make a change to one field all ten will spawn network requests!

[sukima]

As an example on how best to approach the issue of deep tracking Immer.js is a prime example of tracking changes via on-demand proxies.

[NullVoxPopuli]

As an example on how best to approach the issue of deep tracking Immer.js is a prime example of tracking changes via on-demand proxies.

It's there any advantage to this for our purposes over deep-tracked? 🤔

[sukima]

It's there any advantage to this for our purposes over deep-tracked? 🤔

A super quick scan of your ember-deep-tracked and it looks like it is doing as I described above. It looks like it is generating a proxy for each level on demand (when the getter/setter is called). Pretty sure Immur does this also.