enhance: Union types construct class instances in ctrl.fetch() resolution value

[ntucker]

Motivation

ctrl.fetch() resolution should always be denormalized if possible. Previously, we would skip class instantiation (and other denorm things) with Polymorphic types, as they use a special normalized format to select the schema for denormalization.

Solution

Polymorphic (Union) types should still denormalize when handling passthrough (non-normalized) data

When denormalizing non-normalized (like return of ctrl.fetch), it is still expected to handle
all steps like constructing class instances if possible. However, to do this for Polymorphic
types we need to fallback to using part of the normalize process to find out which schema
to use for the remainder of denormalization.

This also removes extra console.warn that would appear when using useSuspense() or other hooks, even though the normalize+denormalize worked fine - since ctrl.fetch() always runs denormalize in 'passthrough' mode (without entities or normalized input).

export class Link extends FeedItem {
  readonly type = 'link' as const;
  readonly url: string = '';
  readonly title: string = '';
}
export class Post extends FeedItem {
  readonly type = 'post' as const;
  readonly content: string = '';
}
export const getFeed = new RestEndpoint({
  path: '/feed',
  schema: new schema.Array(
    {
      link: Link,
      post: Post,
    },
    'type',
  ),
});

Before

const feed = await ctrl.fetch(getFeed);
assert feed[0] is pojo

After

const feed = await ctrl.fetch(getFeed);
assert feed[0] instanceOf Post || feed[0] instanceOf Link

[changeset-bot]

🦋 Changeset detected

Latest commit: e218d8f

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 8 packages

Name	Type
@data-client/endpoint	Patch
@data-client/graphql	Patch
@data-client/rest	Patch
example-benchmark	Patch
normalizr-github-example	Patch
normalizr-redux-example	Patch
normalizr-relationships	Patch
rdc-website	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.49%. Comparing base (a66d51e) to head (e218d8f).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #3063   +/-   ##
=======================================
  Coverage   98.49%   98.49%           
=======================================
  Files         122      122           
  Lines        2196     2199    +3     
  Branches      449      451    +2     
=======================================
+ Hits         2163     2166    +3     
  Misses         21       21           
  Partials       12       12           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[github-actions]

Benchmark

Benchmark suite	Current: e218d8f	Previous: 4bc9145	Ratio
normalizeLong	432 ops/sec (±2.08%)	435 ops/sec (±1.81%)	1.01
infer All	9767 ops/sec (±0.66%)	9500 ops/sec (±1.31%)	0.97
denormalizeLong	345 ops/sec (±0.92%)	331 ops/sec (±2.76%)	0.96
denormalizeLong donotcache	892 ops/sec (±0.26%)	890 ops/sec (±0.57%)	1.00
denormalizeShort donotcache 500x	1345 ops/sec (±0.39%)	1356 ops/sec (±0.89%)	1.01
denormalizeShort 500x	973 ops/sec (±0.24%)	979 ops/sec (±0.44%)	1.01
denormalizeShort 500x withCache	4876 ops/sec (±0.35%)	5094 ops/sec (±0.37%)	1.04
denormalizeLong with mixin Entity	292 ops/sec (±0.51%)	293 ops/sec (±0.31%)	1.00
denormalizeLong withCache	6185 ops/sec (±0.25%)	6217 ops/sec (±0.17%)	1.01
denormalizeLong All withCache	6646 ops/sec (±0.17%)	6757 ops/sec (±0.15%)	1.02
denormalizeLong Query-sorted withCache	6503 ops/sec (±0.48%)	6620 ops/sec (±0.51%)	1.02
denormalizeLongAndShort withEntityCacheOnly	1521 ops/sec (±0.52%)	1483 ops/sec (±1.82%)	0.98
getResponse	5830 ops/sec (±0.68%)	6315 ops/sec (±1.08%)	1.08
getResponse (null)	6297913 ops/sec (±1.10%)	5367929 ops/sec (±0.53%)	0.85
getResponse (clear cache)	288 ops/sec (±0.47%)	278 ops/sec (±0.61%)	0.97
getSmallResponse	2660 ops/sec (±0.28%)	2669 ops/sec (±0.48%)	1.00
getSmallInferredResponse	2040 ops/sec (±0.25%)	2018 ops/sec (±0.52%)	0.99
getResponse Query-sorted	5406 ops/sec (±0.80%)	5570 ops/sec (±0.80%)	1.03
getResponse Collection	5786 ops/sec (±0.68%)	6532 ops/sec (±0.92%)	1.13
get Collection	4940 ops/sec (±1.20%)	5837 ops/sec (±1.12%)	1.18
setLong	438 ops/sec (±1.88%)	432 ops/sec (±2.32%)	0.99
setLongWithMerge	184 ops/sec (±1.81%)	187 ops/sec (±1.20%)	1.02
setLongWithSimpleMerge	194 ops/sec (±1.25%)	193 ops/sec (±1.82%)	0.99
setSmallResponse 500x	860 ops/sec (±1.34%)	857 ops/sec (±1.29%)	1.00

This comment was automatically generated by workflow using github-action-benchmark.