[BUGFIX] resolve issues with RecordArray sync for peekAll

[runspired]

This PR resolves issues with ensuring that RecordArray is correctly synced for peekAll.

Resolves #5271
Resolves #5167
Resolves #5175
Resolves #5111 / #5157

It also includes tests and a fix for #5350 and #5025 / #5095 wherein non-materialized records would remain in live-record-arrays and filtered record arrays upon unload.

Resolves #5025

cc @tylerturdenpants @igorT @hjdivad @workmanw

[tylerturdenpants]

@runspired, it works so far! I'll check back in when I get more coverage of my app

[HeroicEric]

I tested this with our app that has been stuck at Ember Data 2.13.2 for a while and this lets us go all the way to 3.x 😄

[sly7-7]

I really have to dig and try to narrow my problem then, I still encounter a

ember-console.js:29 TypeError: Cannot read property 'eachAttribute' of null
    at new Snapshot (-private.js:5377)
    at InternalModel.createSnapshot (-private.js:7173)
    at _findHasMany (-private.js:9104)
    at Class.findHasMany (-private.js:11343)
    at ManyRelationship.fetchLink (-private.js:4871)
    at ManyRelationship.findLink (-private.js:4132)
    at ManyRelationship.getRecords (-private.js:4913)
    at Class.get (-private.js:13357)
    at ComputedPropertyPrototype.get (ember-metal.js:4064)
    at get (ember-metal.js:3436)

In my scenario, there are several unloadAll calls and a route.refresh call, whereas models are still displayed on the screen.

[runspired]

@sly7-7 hrm :/ Makes me suspicious that #5376 has not caught everything it needed to.

[tylerturdenpants]

@sly7-7 I had the same issue way back in 2.13. I think I resolved it by routing away from where the models were in use and then performed the unload.

As an aside, @runspired, what will it take to get this merged? This has been the most promising fix in the last few months. I owe you big time. You’re a huge value to this community.

[btecu]

Will this make it into 3.1?

[sly7-7]

@tylerturdenpants
@sly7-7 I had the same issue way back in 2.13. I think I resolved it by routing away from where the models were in use and then performed the unload.

Thank you for the tip, I will see if I can try that as a workaround, but for sure there is 'something' wrong 'somewhere'

[runspired]

@sly7-7 I have some thoughts on what your bug is, will try to write a test to see if it fails.

[Kilowhisky]

Found a related issue: #5395 #bummer

[tylerturdenpants]

@bmac In what version is this slated to land?

[sly7-7]

@tylerturdenpants It should be in the just released 3.1. You will encounter a warning though: #5372 (comment) but as @bmac said patch will be soon released too 😄

[tylerturdenpants]

Oh nice @sly7-7. Wish you luck on your issues. I’m still encountering the “Id already used” but is much easier to patch.

[bmac]

Since this was just merged I didn't cherry pick it into 3.1 just yet. Currently its only on 3.2.0-beta.1 but I'd be willing to make a patch release to 3.1 in a few days if nothing major gets reported on the beta release.

[omairvaiyani]

Unfortunately I'm still receiving the error:
Ember Data 3.2.0-beta.1
"Assertion Failed: Cannot create a new chain watcher for `<.. model .. >` after it has been destroyed."