From bf8d4d6db35dbc8dab518eea70fa57212aff38d8 Mon Sep 17 00:00:00 2001
From: "David J. Hamilton" <dhamilton@linkedin.com>
Date: Thu, 26 Oct 2017 15:31:46 -0700
Subject: [PATCH] Ember Data notes 01 november 2017

---
 ember-data/2017-11/november-01.md | 101 ++++++++++++++++++++++++++++++
 1 file changed, 101 insertions(+)
 create mode 100644 ember-data/2017-11/november-01.md

diff --git a/ember-data/2017-11/november-01.md b/ember-data/2017-11/november-01.md
new file mode 100644
index 00000000..be7b6a8b
--- /dev/null
+++ b/ember-data/2017-11/november-01.md
@@ -0,0 +1,101 @@
+# ember-data Meeting Notes - 2017-11-01
+
+## Attendees
+
+- bm @bmac Brendan McLoughlin
+- cm @pangratz Clemens Müller
+- ct @runspired Chris Thoburn
+- dh @hjdivad David J. Hamilton
+- dg @dgeb Dan Gebhardt
+- it @igort Igor Terzic
+- rj @rwjblue Robert Jackson
+
+## Agenda
+
+- addon transitive-unload
+- addon soft-delete
+- emberjs/data#5247 createrecord
+- data#5230 multi-table inheritance
+- unique indexes against simple attributes
+
+## Discussion Notes
+
+### unloadRecord
+
+ct: I want to follow up on the unload addons we discussed previously
+
+dh: No progress on either of them.
+
+dh: We'll want to restore `destroyRecord`'s implicit unload
+
+it: We had this before because of the `firstObject` `lastObject` bugs
+
+dh: Those have been fixed in ember
+
+it: When did those fixes land?
+
+rj: They landed in ember 2.15.  Keep in mind the 2.16 lts releases soon.
+
+### createRecord async
+
+rj: I wanted to discuss something I ran into while working on the codemod for the new testing APIs.  The issue is that there's a runloop assertion because [createRecord](https://github.com/emberjs/data/issues/5247) is async.
+
+it: I'm not sure why it's async.
+
+bm: It's using a runloop for record array optimization
+
+rj: What's the best thing for me to do for the testing codemod?
+
+ct: We could `run.join`
+
+rj: The downside there is it changes semantics.
+
+rj: one way or another the codemod will probably have to wrap `this.subject` for model unit tests in a runloop to be consistent with old semantics.  Unsure about long term semantics.
+
+it: it seems odd for `createRecord` to be async
+
+ct: When exactly do we guarantee live arrays contain the new record?
+
+it: The optimizations are useful for `push` but they don't make much sense for local changes.  We could do async for push and sync for local.
+
+dh: seems good, but we'll have to be careful about the brittleness of sync and async code paths for live array updates.
+
+rj: seems good to me although we may need to have sync state transitions
+
+it: are state transition really the thing that causes async?
+
+it: `updateRecordArrays` is idempotent so it should be fine to call it in `createRecord` to get sync updates in this case
+
+### new qunit api
+
+rj: we may want to spike out test helpers for unit tests now that we don't need to do `ember-qunit` specific stuff.  This makes it easier for addons to expose helpers.  See [my blog post](http://rwjblue.com/2017/10/23/ember-qunit-simplication/#why)
+
+it: it may be a bit tricky, eg consider setting up a record that was loaded from the server but has a dirty attribute.  How would a helper for that look?
+
+rj: I don't know, that's why I'm bringing it up.
+
+rj: for sure there's a difficulty both in setting things up and enabling the user to indicate what they want
+
+ct: the difficulty is in unit testing
+
+it: this is easier to do via `record.serialize()` in an integration test
+
+dh: it's almost always better to test serializers in integration tests precisely because of these setup concerns
+
+*Some discussion around unit testing vs integration testing serializers*
+
+dh: whether we add test helpers or not, it seems like we should have a blueprint as Rob suggests, that shows a recommended way for testing serializers via `push` and `record.serialize`
+
+### multi-table inheritance
+
+ct: i'll pair with @hjdivad on this we talked out of band and it looks like we have a path forward
+
+### custom indexes
+
+dh: this can be enabled via hooks for "saw record" and "record went away" for the case @runspired wants, ie static non-compound keys.
+
+dh: a common use case for this is unifying the id-space of polymorphic types.
+
+ct: yeah that plus the create case.
+
+