WIP client-side-delete semantics `unloadRecord` #5273

hjdivad · 2017-12-06T18:30:18Z

hjdivad · 2017-12-06T18:31:36Z

Follow-on for #5136 #5137

hjdivad · 2017-12-14T01:32:20Z

rwjblue · 2017-12-18T23:50:35Z

tests/integration/records/unload-test.js

@@ -1488,3 +1488,8 @@ test('many:many async unload', function (assert) {
    assert.equal(person1Friends.isDestroyed, true, 'previous ManyArray is destroyed in the runloop after refetching');
  });
 });
+
+test('1 sync : 1 async');


Fwiw I do t think QUnit supports this syntax (assuming you are intending to basically create a skip?)...

@rwjblue creating a todo, not a skip.

igorT · 2018-01-09T23:50:35Z

addon/-private/system/relationships/state/has-many.js

    this.__loadingPromise = null;
  }

+  _inverseIsAsync() {


This could live on the parent class

igorT · 2018-01-09T23:52:30Z

addon/-private/system/relationships/state/has-many.js

@@ -44,6 +52,10 @@ export default class ManyRelationship extends Relationship {
        isPolymorphic: this.isPolymorphic
      });
    }
+    if (this._retainedManyArray !== null) {


Should this ever happen if manyArray is not null?

lets add an assert

igorT · 2018-01-09T23:53:48Z

addon/-private/system/relationships/state/has-many.js

+      }
+      this._removeInternalModelFromManyArray(this._retainedManyArray, inverseInternalModel);
+    } else {
+      this.removeInternalModelFromOwn(inverseInternalModel);


This seems to be common for both hasMany and belongsTo?

igorT · 2018-01-09T23:54:05Z

addon/-private/system/relationships/state/has-many.js

@@ -143,7 +167,26 @@ export default class ManyRelationship extends Relationship {
      return;
    }
    super.removeInternalModelFromOwn(internalModel, idx);
-    let manyArray = this.manyArray;
+    // TODO: explain this madness


igorT · 2018-01-09T23:54:34Z

addon/-private/system/relationships/state/relationship.js

-    return this.internalModel.modelName;
+  _inverseIsAsync() {
+    // TODO: clean this up; it's for implicit relationships
+    return this.isAsync;


this seems incorrect?

igorT · 2018-01-10T01:02:56Z

addon/-private/system/relationships/state/has-many.js

-    let manyArray = this.manyArray;
+    // TODO: explain this madness
+    this._removeInternalModelFromManyArray(this.manyArray, internalModel, idx);
+    this._removeInternalModelFromManyArray(this._retainedManyArray, internalModel, idx);


we should maybe do this in the async case?

we discussed this OOB, but LMK if you still think there's an issue here

fivetanley · 2018-01-10T17:42:39Z

@hjdivad I tried this out in our app. We're still stuck in this scenario:

import Ember from 'ember';

export default Ember.Controller.extend({
  actions: {
    unload() {
      const clients = this.store.peekAll('client');

      // there is one record that matches this, so one record unloaded
      clients.filterBy('isCool').forEach((client) => {
        client.unloadRecord();
      });
      
      // here clients is [Model, Model, null]
      // null should not be in the array
      
      // error here because filterBy calls Ember.get on null to try to find hasDirtyAttributes.
      clients.filterBy('hasDirtyAttributes').forEach(client => {
        client.unloadRecord();
      });
    }
  }
});

It looks like some nulls are still getting stuck in the RecordArray created for this.store.peekAll('client').

hjdivad · 2018-01-10T18:15:42Z

Thanks for the followup @fivetanley. I'll see about reproducing this in a test if you haven't already.

A number of use cases rely on unloadRecord's 2.12 behaviour of treating unloadRecord as a client-side delete on the inverse side of a sync relationship. This pr restores that functionality, while retaining the invalidate+refetch functionality for async relationships. This behaviour is codified in a number of tests within tests/integration/records/unload-test.js Fix #5136, #5137

hjdivad · 2018-01-11T04:19:44Z

@igorT this should be good to merge; issues we discussed are handled.

hjdivad · 2018-01-11T04:20:24Z

@fivetanley this doesn't resolve the filter issue you mentioned in your previous comment; that will be addressed in a follow-on issue.

sly7-7 · 2018-01-12T16:27:57Z

@fivetanley I wonder if in your use case, calling toArray() after filter would change anything ?

workmanw · 2018-01-17T23:06:01Z

This isn't tagged with the release channel. Can we please get this in 2.18 LTS?

workmanw · 2018-01-28T16:51:35Z

Shameless bump @hjdivad @rwjblue -- Can we get this PR and #5249 into 2.18.1? I know DS doesn't have the same level LTS support as Ember, but it would be nice to have a place to jump to in the 2.x series.

sly7-7 · 2018-02-09T10:40:29Z

Or at least in 3.0 beta ?
BTW @workmanw did you try running your app against master ? On my side this seems to tackle all the regression I had after 2.12.2

elgordino · 2018-02-13T10:46:11Z

FYI This is included in 3.1.0-beta.1 https://github.com/emberjs/data/releases

bmac · 2018-02-13T13:56:29Z

I just released 2.18.1 and 3.0.1 with this pr. Thanks to @sly7-7 for pinging me about including this pr.

workmanw · 2018-02-14T12:23:26Z

@bmac Thank you!!

jlami · 2018-02-15T15:42:58Z

@hjdivad What is the status of the followup on the problem @fivetanley was having?
I think this might be the same problem as I'm having. see also #5353 and #5354

Could somebody explain the reasoning behind the distinction in handling of the canonical state of sync and async relationships for unload? I'm inclined the client side should not be effected by this, as this is more of a server behaviour in my mind.

hjdivad force-pushed the hjdivad/restore-unloadRecord-client-side-delete-semantics branch from 61cee15 to ed208d0 Compare December 14, 2017 00:50

rwjblue reviewed Dec 18, 2017

View reviewed changes

hjdivad force-pushed the hjdivad/restore-unloadRecord-client-side-delete-semantics branch from 87c8129 to aa59926 Compare January 4, 2018 01:13

igorT reviewed Jan 9, 2018

View reviewed changes

igorT reviewed Jan 10, 2018

View reviewed changes

hjdivad force-pushed the hjdivad/restore-unloadRecord-client-side-delete-semantics branch from aa59926 to ad66b89 Compare January 11, 2018 01:42

Wesley Workman and others added 2 commits January 10, 2018 20:13

Added failing test for #5136.

bbe8b41

hjdivad force-pushed the hjdivad/restore-unloadRecord-client-side-delete-semantics branch from ad66b89 to 6e8a236 Compare January 11, 2018 04:18

hjdivad requested a review from igorT January 11, 2018 04:19

igorT merged commit b329a2c into master Jan 11, 2018

igorT deleted the hjdivad/restore-unloadRecord-client-side-delete-semantics branch January 11, 2018 16:36

This was referenced Jan 11, 2018

Added failing test for #5136. #5137

Closed

Loading a previously unloaded record doesn't update relationships #5136

Closed

al3xnag mentioned this pull request Feb 17, 2018

peekAll contains null entries after unloadRecord #5111

Closed

kennethlarsen mentioned this pull request Mar 26, 2018

Ember 3.1 Release Blog Post emberjs/website#3230

Merged

13 tasks

Kilowhisky mentioned this pull request Mar 26, 2018

unloadRecord() jumps the gun on peekAll arrays using dependentArrayWillChange observers #5395

Closed

runspired mentioned this pull request Apr 20, 2018

unloadRecord destroys manyArray #4835

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

WIP client-side-delete semantics `unloadRecord` #5273

WIP client-side-delete semantics `unloadRecord` #5273

hjdivad commented Dec 6, 2017 •

edited

Loading

hjdivad commented Dec 6, 2017

hjdivad commented Dec 14, 2017

rwjblue Dec 18, 2017

hjdivad Jan 3, 2018

igorT Jan 9, 2018

igorT Jan 9, 2018

igorT Jan 10, 2018

igorT Jan 9, 2018

igorT Jan 9, 2018

igorT Jan 9, 2018

igorT Jan 10, 2018

hjdivad Jan 11, 2018

fivetanley commented Jan 10, 2018

hjdivad commented Jan 10, 2018

hjdivad commented Jan 11, 2018

hjdivad commented Jan 11, 2018

sly7-7 commented Jan 12, 2018

workmanw commented Jan 17, 2018

workmanw commented Jan 28, 2018

sly7-7 commented Feb 9, 2018 •

edited

Loading

elgordino commented Feb 13, 2018

bmac commented Feb 13, 2018

workmanw commented Feb 14, 2018

jlami commented Feb 15, 2018

WIP client-side-delete semantics unloadRecord #5273

WIP client-side-delete semantics unloadRecord #5273

Conversation

hjdivad commented Dec 6, 2017 • edited Loading

hjdivad commented Dec 6, 2017

hjdivad commented Dec 14, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

fivetanley commented Jan 10, 2018

hjdivad commented Jan 10, 2018

hjdivad commented Jan 11, 2018

hjdivad commented Jan 11, 2018

sly7-7 commented Jan 12, 2018

workmanw commented Jan 17, 2018

workmanw commented Jan 28, 2018

sly7-7 commented Feb 9, 2018 • edited Loading

elgordino commented Feb 13, 2018

bmac commented Feb 13, 2018

workmanw commented Feb 14, 2018

jlami commented Feb 15, 2018

WIP client-side-delete semantics `unloadRecord` #5273

WIP client-side-delete semantics `unloadRecord` #5273

hjdivad commented Dec 6, 2017 •

edited

Loading

sly7-7 commented Feb 9, 2018 •

edited

Loading