@stonyx/orm 0.3.2-beta.153 → 0.3.2-beta.155

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -381,6 +381,174 @@ export default class GlobalAccess {
381
381
  }
382
382
  ```
383
383
 
384
+
385
+ ### The access context (second argument)
386
+
387
+ `access()` is called with **two** arguments:
388
+
389
+ ```js
390
+ access(request, { model, operation })
391
+ ```
392
+
393
+ The second is the **access context** — the structural facts about the request,
394
+ which the framework already holds at authorization time. Read these instead of
395
+ parsing anything.
396
+
397
+ | Key | Value |
398
+ |---|---|
399
+ | `model` | The model this route was mounted for, as a **model name**: kebab-case, exactly as declared under `config.orm.paths.model` and keyed in the store — `'owner'`, `'animal'`, `'phone-number'`. **Not** the pluralized, dasherized, mount-prefixed *route* name. |
400
+ | `operation` | One of **`'read'`, `'create'`, `'update'`, `'delete'`** — and no second vocabulary *on this path*. Never an HTTP method name like `'GET'`, and **not** the hook vocabulary either (see [below](#operation-is-not-the-hook-operation)). `undefined` when the dispatched method has no entry in the framework's method map. |
401
+
402
+ So a predicate can be written without reference to any URL:
403
+
404
+ ```js
405
+ export default class OwnerAccess {
406
+ models = ['owner'];
407
+
408
+ access(request, { model, operation }) {
409
+ if (model === 'owner' && operation === 'read') {
410
+ return record => record.id !== 'angela';
411
+ }
412
+
413
+ return ['read'];
414
+ }
415
+ }
416
+ ```
417
+
418
+ There is no string to parse, no variant to miss, and no way to fail open through
419
+ a URL shape nobody anticipated. `model` is fixed at mount time and no request
420
+ can influence it — not a mount prefix, not a query string, not a case-varied
421
+ path, not an absolute-form request target.
422
+
423
+ #### What the context does not tell you: which surface
424
+
425
+ It names **which model and which verb**, not **which route**. Measured over the
426
+ live router, six surfaces produce one identical context:
427
+
428
+ ```
429
+ GET /owners { model: 'owner', operation: 'read' }
430
+ GET /owners/gina { model: 'owner', operation: 'read' }
431
+ GET /owners/gina/pets { model: 'owner', operation: 'read' }
432
+ GET /owners/gina/relationships/pets { model: 'owner', operation: 'read' }
433
+ GET /owners/archived { model: 'owner', operation: 'read' }
434
+ GET /owners/gina?include=pets { model: 'owner', operation: 'read' }
435
+ ```
436
+
437
+ So a rule that depends on the **sub-path** still needs `request.path` —
438
+ mount-relative and query-free, and the one read of argument one that
439
+ [Identifying the collection](#identifying-the-collection) sanctions. The sample
440
+ access class shipped with this repo has such a rule: its `/archived` deny
441
+ **cannot be expressed from the context alone**, and a predicate migrated to
442
+ context-only would silently drop it — a deny becoming an allow.
443
+
444
+ Note also that the related-resource and `?include=` surfaces serve *another
445
+ model's* records under `model: 'owner'`, and the context gives a predicate no
446
+ signal that it is authorizing a related-resource route. That is
447
+ [#196](https://github.com/abofs/stonyx-orm/issues/196).
448
+
449
+ #### `operation` is not the hook `operation`
450
+
451
+ This module exposes a **second** `operation` vocabulary, on an identically-named
452
+ key of an identically-shaped context object:
453
+ [hook contexts](#hook-context-object) carry `list` / `get` / `create` /
454
+ `update` / `delete`. The access vocabulary collapses `list` and `get` into
455
+ `'read'`, so for one `GET /animals/1` a hook sees `'get'` while `access()` sees
456
+ `'read'` — and a predicate cannot distinguish a collection read from a
457
+ record read.
458
+
459
+ "No second vocabulary" above is a statement about the **access path**, where
460
+ both the context and the permission array come from one method map. It is not a
461
+ statement about the module. Writing `operation === 'get'` in a predicate never
462
+ matches, and a predicate that stops matching falls through to the permission
463
+ array — so the misreading is fail-open shaped. In TypeScript the exported
464
+ `AccessOperation` union makes it a compile error.
465
+
466
+ The four `operation` values are the same four strings the permission-array
467
+ return shape is written in (`['read', 'create', 'update', 'delete']`), because
468
+ both come from one method map inside the framework. The two forms cannot
469
+ disagree about the same request.
470
+
471
+ **`operation` is `undefined`, never defaulted, for an unmapped method.** Express
472
+ delivers `HEAD` to the `GET` handler, so this is reachable. It is deliberately
473
+ not defaulted to `'read'`: a fabricated operation would turn an unclassified
474
+ request into an authorized one. Treat `undefined` as *not classified* and deny.
475
+
476
+ **The second argument is additive.** JavaScript ignores extra arguments, so an
477
+ existing `access(request)` predicate keeps working exactly as it did. Nothing
478
+ needs to be migrated to keep running — but note that argument **one** is still
479
+ the raw request, so the warning in
480
+ [Identifying the collection](#identifying-the-collection) still applies to any
481
+ predicate that reads it.
482
+
483
+ #### `record` is not in the context
484
+
485
+ Deliberately, and it is not an oversight. `auth()` runs after route matching but
486
+ **before any handler executes**, so nothing has been fetched yet. Supplying a
487
+ record would force a pre-fetch on every request — a second store hit, a new
488
+ failure mode, and an ordering change in the middle of an authorization path.
489
+
490
+ It is also unnecessary: the **function** return shape already *is* the
491
+ per-record hook. Return `(record) => boolean` and the handlers apply it to every
492
+ record the request touches. Auth-time and record-time are separate decision
493
+ points, and the contract keeps them separate.
494
+
495
+ #### Reaching another model's predicate
496
+
497
+ The model → predicate map is published on the ORM instance at boot, before any
498
+ route is mounted, so a predicate can be resolved by model name and asked about a
499
+ request routed to a *different* model:
500
+
501
+ ```js
502
+ import Orm from '@stonyx/orm';
503
+
504
+ const predicate = Orm.instance.getAccess('animal');
505
+ if (!predicate) return deny;
506
+
507
+ const verdict = predicate(request, { model: 'animal', operation: 'read' });
508
+ ```
509
+
510
+ **`undefined` means no predicate could be resolved — not that the model is
511
+ unrestricted. Treat it as deny.** It covers a model with no access class *and* a
512
+ model whose access class failed to **load**: a load failure is caught and warned
513
+ about, and the partial map is published anyway, so a missing key is not evidence
514
+ of an unrestricted model. This is the same rule as `operation === undefined`
515
+ above, and for the same reason.
516
+
517
+ The raw map is `Orm.instance.accessFunctions`, keyed by model name; prefer
518
+ `getAccess()` — it is guarded against inherited `Object.prototype` members and a
519
+ direct index is not. Note that it maps a model name to the predicate of the
520
+ access *class* that claims it, which may claim many models: against this repo's
521
+ sample, `getAccess('owner') === getAccess('animal')`.
522
+
523
+ #### Passing the context makes a model-correct answer *possible*
524
+
525
+ It does not make the answer model-correct on its own. **The resolved predicate
526
+ has to read the context.** Measured against the access class shipped with this
527
+ repo, on a request Express dispatched to `GET /owners/angela`, asked about
528
+ **animals**:
529
+
530
+ ```
531
+ getAccess('animal')(ownersRequest, { model: 'animal', operation: 'read' })
532
+ -> record => record.id !== 'angela' && record.id !== 'restricted'
533
+ ```
534
+
535
+ That is the **owners** filter, and it returns `true` for animal 21 — the record
536
+ hidden on every animal surface. Under a mount that predicate recognizes neither
537
+ way it is worse still: it falls through to
538
+ `['read', 'create', 'update', 'delete']`, a full CRUD grant.
539
+
540
+ Either way the context was supplied and the answer is not the animal answer, and
541
+ it is wrong in the direction that **grants**. That predicate is single-argument
542
+ and identifies its collection from the request, so it answered about the
543
+ collection the request is *addressed to* while being asked about another one.
544
+ Every predicate in this repo, and in every consumer tree, is single-argument on
545
+ the day this ships, and a caller has no supported way to tell which kind it
546
+ resolved. The boot-time arity warning that would surface it is
547
+ [#213](https://github.com/abofs/stonyx-orm/issues/213).
548
+
549
+ So: pass the context, and do not treat a resolved predicate's answer as
550
+ model-specific until that predicate has been migrated to read it.
551
+
384
552
  ### Return values
385
553
 
386
554
  | `access()` returns | Effect |
@@ -413,7 +581,7 @@ a write to a *different* collection can still re-parent one. See
413
581
  | `GET /:models/:id/relationships/{relationship}` | `404` — same |
414
582
  | `PATCH /:models/:id` | `404`, no attribute is applied |
415
583
  | `DELETE /:models/:id` | `404`, the record is not removed and no SQL `DELETE` is issued |
416
- | `POST /:models` | `403`, and a record **this request inserted** is rolled back — see [Known limitations](#known-limitations) for the case where it did not insert one |
584
+ | `POST /:models` | `403`, and a record **this request inserted** is rolled back — see [Known limitations](#known-limitations) for why the rollback is conditional |
417
585
 
418
586
  **Denied record-level requests return 404, not 403.** This is deliberate and it
419
587
  is the property most easily "improved" away. 403 would confirm that the record
@@ -447,9 +615,64 @@ chooses the id and learns whether the create succeeded — so under a filter the
447
615
  caller does not choose the id. The refusal happens before any store lookup, so
448
616
  neither the status nor the response time depends on whether the id exists.
449
617
 
450
- Let the server assign the id and read it back from the response. Callers with no
451
- function-style filter are unaffected: `409` on a duplicate id and `200` on a free
452
- one both behave exactly as before.
618
+ Let the server assign the id and read it back from the response and read it
619
+ back rather than predicting it, because the value it returns is documented but
620
+ not stable across model kinds: a numeric-id model gets `max + 1` (or, at the
621
+ numeric ceiling, the lowest free integer), and a string-id model gets
622
+ `<model>-<n>`. See breaking change 8.
623
+
624
+ **What a server-assigned id is not.** It is not a secret. On a string-id
625
+ collection it is dense and enumerable from `1`, where previously it inherited
626
+ whatever entropy the last-inserted id happened to carry — a UUID-seeded store
627
+ answered a UUID-derived key. If a collection has **no** `access` config its
628
+ record-level routes are ungated, so the id was the only thing standing between
629
+ an unauthenticated caller and `GET`/`PATCH`/`DELETE` on a record. That was never
630
+ a control and must not become one; configure `access`.
631
+
632
+ **And the id itself is an occupancy signal — on both model kinds.**
633
+ `assignRecordId` reads the whole store, not the caller's filtered view — it
634
+ never sees `state.filter` — so the id it returns is a function of records the
635
+ caller may not be permitted to read. **This applies to numeric-id collections
636
+ as well as string-id ones**, and the conditions differ, so read both:
637
+
638
+ - **String-id collections, always.** The assigned `n` is the smallest positive
639
+ integer whose landing key is free, which tells the caller that every key
640
+ below it is taken, hidden or not.
641
+ - **Numeric-id collections, once one record sits at the numeric ceiling.** The
642
+ normal answer is `max + 1`, which discloses only the maximum. But `max + 1`
643
+ is not representable at or above 2^53, so the walk restarts from `1` (see
644
+ breaking change 8) and the assigned id becomes the smallest free integer —
645
+ the same occupancy predicate, now over arbitrary low keys. Each subsequent
646
+ no-id `POST` names the next free one, so a caller can enumerate the holes in
647
+ a range it cannot read.
648
+
649
+ **A ceiling record reaches a filter-protected collection even though `POST`
650
+ refuses caller ids on one.** Breaking change 3 makes
651
+ `POST /animals {"id": 9007199254740992}` answer `403`, but the same id lands
652
+ through a *relationship write on another collection* —
653
+ `POST /owners` carrying `attributes: { pets: [{ "id": 9007199254740992 }] }`
654
+ creates the animal under that key
655
+ ([#207](https://github.com/abofs/stonyx-orm/issues/207), the same channel the
656
+ **Known limitations** re-parenting note describes). So the precondition is
657
+ reachable by an unauthenticated caller on exactly the collections `access`
658
+ exists to protect. Measured on the sample fixture, with every animal hidden by
659
+ the `/animals` predicate and keys 4 and 7 deleted:
660
+
661
+ ```
662
+ GET /animals -> 200 [] (nothing visible)
663
+ GET /animals/4 -> 404 (free — indistinguishable from hidden)
664
+ POST /animals {"id":4} -> 403 (breaking change 3)
665
+ POST /owners {... pets:[{"id":9007199254740992}]} -> 200 (#207 plants the ceiling record)
666
+ POST /animals (no id) -> 200 id=4 <- names a hole in the hidden range
667
+ POST /animals (no id) -> 200 id=7 <- and the other one
668
+ POST /animals (no id) -> 200 id=13
669
+ POST /animals (no id) -> 200 id=14
670
+ ```
671
+
672
+ Closing this requires the assignment to be filter-aware, which is a change to
673
+ the `access` contract rather than a fix; it is stated here rather than left to
674
+ be discovered. Callers with no function-style filter are unaffected — there are
675
+ no hidden records to disclose.
453
676
 
454
677
  ### Identifying the collection
455
678
 
@@ -512,14 +735,24 @@ sub-paths beneath the mount, as the `/archived` deny above does.
512
735
  one collection is writable and another is filtered on a field the first can
513
736
  set. Blocking it requires checking animal 21 against the **animal** model's
514
737
  predicate while servicing an **owners** route — cross-model access resolution,
515
- which the current contract cannot express: `access()` never receives the model
516
- structurally ([#202](https://github.com/abofs/stonyx-orm/issues/202)) and
517
- `setup-rest-server.ts` discards the model→predicate map at boot
518
- ([#196](https://github.com/abofs/stonyx-orm/issues/196)). Tracked as
519
- [#207](https://github.com/abofs/stonyx-orm/issues/207), blocked on that chain
520
- (#202 → #196 → #207). Until it lands, do not rely on a filter to keep a record
521
- unmodifiable; keep the *writable* collections' predicates as tight as the
522
- hidden ones.
738
+ which the contract could not express before
739
+ [#202](https://github.com/abofs/stonyx-orm/issues/202): `access()` never
740
+ received the model structurally and `setup-rest-server.ts` discarded the
741
+ model→predicate map at boot. **#202 has landed and both halves now exist** —
742
+ see [The access context](#the-access-context-second-argument):
743
+ `Orm.instance.getAccess(modelName)` makes another model's predicate
744
+ **reachable**, and `context.model` makes a **model-correct answer possible**
745
+ possible, not guaranteed: the resolved predicate has to read the context, and
746
+ every predicate in tree is still single-argument
747
+ ([#213](https://github.com/abofs/stonyx-orm/issues/213)), so today it answers
748
+ about the collection the request is addressed to. **The mechanism exists; the
749
+ ORM does not yet use it on this path.** The re-parenting write above is still
750
+ **not refused** — that enforcement is
751
+ [#196](https://github.com/abofs/stonyx-orm/issues/196) and
752
+ [#207](https://github.com/abofs/stonyx-orm/issues/207), which were blocked on
753
+ #202 and are now free to proceed. Until they land, do not rely on a filter to
754
+ keep a record unmodifiable; keep the *writable* collections' predicates as tight as
755
+ the hidden ones.
523
756
  - **Authorization by identifying the collection is a consumer-side
524
757
  reconstruction of information the framework already holds.** `access()`
525
758
  receives a transport artifact and is asked to work out which model, which
@@ -565,9 +798,10 @@ sub-paths beneath the mount, as the `/archived` deny above does.
565
798
  transform output differs from its lookup key is the same defect. Filtered
566
799
  collections are unaffected — breaking change 3 refuses any client-supplied id
567
800
  — so this reaches consumers with **no** function-style filter. Tracked as
568
- [#205](https://github.com/abofs/stonyx-orm/issues/205), alongside
569
- [#203](https://github.com/abofs/stonyx-orm/issues/203), which is the other way
570
- a create can land on an id nobody named.
801
+ [#205](https://github.com/abofs/stonyx-orm/issues/205). Its sibling
802
+ [#203](https://github.com/abofs/stonyx-orm/issues/203) a **server-assigned**
803
+ id landing on an occupied slot — is **fixed**; see breaking change 8. #205 is
804
+ the client-supplied half and is still open.
571
805
  - **`context.record` is `undefined` for an after-`create` hook when a string-id
572
806
  model is given a numeric-looking id.** The post-create lookup uses the same id
573
807
  coercion as every other surface, which resolves `'9107'` to the number `9107`,
@@ -577,15 +811,18 @@ sub-paths beneath the mount, as the `/archived` deny above does.
577
811
  [#209](https://github.com/abofs/stonyx-orm/issues/209).
578
812
  - **A denied `POST` rolls back only a record it *inserted*.** The rollback
579
813
  requires the store to have grown, because removing by id alone is a write
580
- primitive keyed by a caller-supplied value. When `assignRecordId` lands a
581
- **server-assigned** id on an occupied slot
582
- ([#203](https://github.com/abofs/stonyx-orm/issues/203) — it returns
583
- last-*inserted* + 1, not max + 1, so a store whose insertion order is not
584
- ascending collides), `createRecord` updates that record **in place**: the map
585
- does not grow, the rollback correctly declines to remove a record this request
586
- did not create, and the `403` leaves the caller's attributes on someone else's
587
- record. Narrow it needs a non-ascending insertion order — but it is the
588
- reachability condition, so it is stated rather than implied.
814
+ primitive keyed by a caller-supplied value. **The reachability condition this
815
+ bullet used to state is gone**: it was
816
+ [#203](https://github.com/abofs/stonyx-orm/issues/203) — `assignRecordId`
817
+ returned last-*inserted* + 1, so a server-assigned id could land on an
818
+ occupied slot and `createRecord` would update it in place and #203 is fixed
819
+ (breaking change 8). A server-assigned create can no longer overwrite, so on a
820
+ collection whose only id channel is `createHandler` this guard has no
821
+ observable effect today. It is kept because a caller-supplied id reaching
822
+ `createRecord` from another route a relationship write,
823
+ [#207](https://github.com/abofs/stonyx-orm/issues/207) — puts the condition
824
+ back, and without the guard a denied `403` would delete a record the request
825
+ did not create.
589
826
 
590
827
  ### Breaking changes
591
828
 
@@ -646,6 +883,64 @@ they are recorded here.
646
883
  population breaking changes 3 and 4 explicitly exempt. If you were relying on
647
884
  a hex-shaped or whitespace-padded id creating a second record, it never did.
648
885
 
886
+ 8. **Server-assigned ids change value on string-id models, numeric ids stop
887
+ being monotonic at the numeric ceiling, and the create route gains a
888
+ `409`.** Three consumer-visible changes from
889
+ [#203](https://github.com/abofs/stonyx-orm/issues/203).
890
+
891
+ **The value.** A `POST` with no `id` against a model declaring
892
+ `id = attr('string')` previously produced the *last-inserted* id with `1`
893
+ concatenated onto it — an owner store holding `['gina', 'bob']` answered
894
+ `'bob1'`. It now answers `'owner-1'`: the model name, a hyphen, and the
895
+ lowest positive integer whose landing key is free. **No test in this repo
896
+ pinned the old value**, so a consumer relying on it gets no failing test, no
897
+ deprecation and no other signal — which is why it is recorded here. Numeric
898
+ id models (`id = attr('number')`, the default) are unaffected in shape: they
899
+ still get an integer, but it is now the **maximum** existing id plus one
900
+ rather than the last-inserted id plus one, which is the defect #203 is about.
901
+ They are **not** unaffected in *sequence* — see the monotonicity half below.
902
+
903
+ The value is deliberately **not** numeric-looking, and that is not cosmetic.
904
+ Every id-bearing surface resolves a numeric-looking string id to a **number**
905
+ (`GET /owners/1` looks up `1`), while a string-id model files its records
906
+ under the **string** key `'1'`. A server-assigned `'1'` would therefore be a
907
+ record that was created successfully and could not be fetched, updated or
908
+ deleted by id, and whose after-`create` hook received
909
+ `context.record === undefined`
910
+ ([#209](https://github.com/abofs/stonyx-orm/issues/209)).
911
+
912
+ **Numeric ids are no longer monotonic, and deleted ids can be re-issued.**
913
+ The precondition is narrow but it is reachable, and there is no signal when
914
+ it is met: **one record filed at or above 2^53** (`9007199254740992`). `max
915
+ + 1` is not representable there, so assignment restarts from `1` and walks
916
+ up to the lowest free key — which means the id of a *deleted* record is
917
+ handed to the next `POST`. Both `dev` and every prior release were strictly
918
+ monotonic and never re-issued a numeric id, so a consumer that relied on
919
+ that — audit rows, cursors, cached authorization decisions, external
920
+ references keyed on the id — now has a stale reference that silently points
921
+ at a **different record, created by a different caller**, rather than at a
922
+ deleted one. Nothing fails; the reference simply resolves to the wrong
923
+ record.
924
+
925
+ The restart is deliberate and is not itself optional: without it, one record
926
+ at the ceiling made every subsequent server-assigned create on that
927
+ collection fail permanently. Re-use is the cost of keeping the collection
928
+ writable. **If you need monotonic ids, assign them yourself** rather than
929
+ letting the server assign, and note that a ceiling record can be planted by
930
+ an unauthenticated caller — see *And the id itself is an occupancy signal*
931
+ under [Filter functions](#filter-functions) for the reachability path.
932
+ String-id models are unaffected by this half: their keys are
933
+ `<model>-<n>` and were never monotonic over an integer sequence.
934
+
935
+ **The status.** `POST /{collection}` can now answer `409` for a reason other
936
+ than a duplicate id: the server could not derive a free id. That requires a
937
+ **non-injective** id transform — one that maps distinct candidates onto the
938
+ same store key, such as `boolean`, or anything you registered on
939
+ `Orm.instance.transforms` and named as an id type. It is a configuration
940
+ fault rather than a request fault; the message is logged through
941
+ `stonyx/log`. Previously this case threw out of the handler and express
942
+ answered `500` with a stack trace.
943
+
649
944
  ### Include Parameter (Sideloading Relationships)
650
945
 
651
946
  The ORM supports JSON API-compliant relationship sideloading via the `include` query parameter. This reduces the need for multiple API requests by embedding related records in a single response.
package/dist/index.d.ts CHANGED
@@ -9,6 +9,7 @@ import { count, avg, sum, min, max } from './aggregates.js';
9
9
  export { default } from './main.js';
10
10
  export { store, relationships } from './main.js';
11
11
  export type { PersistErrorDetail } from './main.js';
12
+ export type { AccessContext, AccessFunction, AccessMethod, AccessOperation } from './types/orm-types.js';
12
13
  export { Model, View, Serializer };
13
14
  export { attr, belongsTo, hasMany, createRecord, updateRecord };
14
15
  export { count, avg, sum, min, max };
package/dist/main.d.ts CHANGED
@@ -1,4 +1,5 @@
1
1
  import Store from './store.js';
2
+ import type { AccessFunction } from './types/orm-types.js';
2
3
  interface OrmOptions {
3
4
  dbType?: string;
4
5
  }
@@ -32,12 +33,127 @@ export default class Orm {
32
33
  views: Record<string, unknown>;
33
34
  transforms: Record<string, (value: unknown) => unknown>;
34
35
  warnings: Set<string>;
36
+ /**
37
+ * Model name -> the `access` predicate of the access class that CLAIMS that
38
+ * model (abofs/stonyx-orm#202).
39
+ *
40
+ * Not "that model's own predicate". One access class may claim many models
41
+ * -- `GlobalAccess` in this repo's fixtures declares five, and `models = '*'`
42
+ * claims every model in the store -- and it declares ONE `access` method, so
43
+ * the same function object is registered under every one of those keys.
44
+ * `getAccess('owner') === getAccess('animal')` is `true` there. The
45
+ * one-to-one guarantee below is key -> function, never function -> model,
46
+ * and a caller must not read a resolved predicate as being animal-specific.
47
+ * What makes the ANSWER model-specific is the context the caller passes and
48
+ * the predicate actually reading it -- see {@link Orm#getAccess}.
49
+ *
50
+ * NAMED FOR WHAT IT HOLDS. It was `accessFiles` through review, inherited
51
+ * from the function-local in `setup-rest-server.ts` where the values came
52
+ * straight out of `forEachFileImport` and "files" was defensible. The values
53
+ * are `AccessFunction`s, and the sibling public registries on this class
54
+ * (`models`, `serializers`, `views`, `transforms`) are all plural nouns of
55
+ * the thing held. Renamed here because #202 is the last moment it is free.
56
+ *
57
+ * Populated by `setup-rest-server.ts` at boot, from the access classes under
58
+ * `config.orm.paths.access`, BEFORE any route is mounted -- so it is complete
59
+ * and reachable before the first request can be served. The mapping is
60
+ * one-to-one by construction: setup-rest-server throws if two access classes
61
+ * claim the same model.
62
+ *
63
+ * Keys are model names as declared and stored (kebab-case, e.g.
64
+ * `'phone-number'`), NOT pluralised or mount-prefixed route names.
65
+ *
66
+ * WHY THIS EXISTS AS A FIELD. It used to be a function-local in
67
+ * setup-rest-server that was discarded when that function returned, so at
68
+ * request time there was no way to get from a model name to that model's
69
+ * predicate at all. Each `OrmRequest` held only its OWN model's predicate.
70
+ * That made cross-model authorization -- asking model X's predicate about a
71
+ * request routed to model Y -- inexpressible, which is the capability
72
+ * abofs/stonyx-orm#196 and abofs/stonyx-orm#207 are built on.
73
+ *
74
+ * Empty when the REST server is disabled, and PARTIAL when one access file
75
+ * failed to load (`setup-rest-server.ts` catches, warns and assigns whatever
76
+ * it had). So a missing key does NOT mean the model has no access class.
77
+ * Prefer {@link Orm#getAccess} over indexing this directly -- it is guarded
78
+ * against the prototype chain and this is not.
79
+ */
80
+ accessFunctions: Record<string, AccessFunction>;
35
81
  options: OrmOptions;
36
82
  sqlDb?: SqlDb;
37
83
  db?: OrmDB | SqlDb;
38
84
  private _persistErrorHandler;
39
85
  constructor(options?: OrmOptions);
40
86
  init(): Promise<void>;
87
+ /**
88
+ * Resolve the `access` predicate registered for a model name
89
+ * (abofs/stonyx-orm#202).
90
+ *
91
+ * This is the supported way to reach another model's predicate while
92
+ * servicing a request routed to a different model. Call it with the model
93
+ * name and invoke the result with the live request and an explicit context
94
+ * naming THAT model:
95
+ *
96
+ * ```js
97
+ * const predicate = Orm.instance.getAccess('animal');
98
+ * if (!predicate) return deny;
99
+ * const verdict = predicate(request, { model: 'animal', operation: 'read' });
100
+ * ```
101
+ *
102
+ * WHAT IT RESOLVES. The predicate of the access CLASS that claims the model,
103
+ * which is not necessarily specific to it: one class may claim many models
104
+ * and declares one `access` method, so
105
+ * `getAccess('owner') === getAccess('animal')` is `true` against this repo's
106
+ * fixture. See {@link Orm#accessFunctions}.
107
+ *
108
+ * `undefined` means NO PREDICATE COULD BE RESOLVED for that name. That
109
+ * includes a model whose access class failed to LOAD -- `setup-rest-server`
110
+ * catches, warns and publishes the partial map -- so it is not the same claim
111
+ * as "this model is unrestricted". Treat it as DENY, the same way
112
+ * `AccessContext.operation === undefined` is treated.
113
+ *
114
+ * PASSING THE CONTEXT MAKES A MODEL-CORRECT ANSWER POSSIBLE. It does not, on
115
+ * its own, make the answer model-correct: the resolved predicate has to READ
116
+ * the context. Measured against this repo's shipped access class on a request
117
+ * express dispatched to `GET /owners/angela`, asked about ANIMALS:
118
+ *
119
+ * ```
120
+ * getAccess('animal')(ownersRequest, { model: 'animal', operation: 'read' })
121
+ * -> record => record.id !== 'angela' && record.id !== 'restricted'
122
+ * ```
123
+ *
124
+ * The OWNERS filter, which returns `true` for animal 21 -- the record hidden
125
+ * on every animal surface. Under a mount that predicate recognises neither
126
+ * way it falls through to `['read', 'create', 'update', 'delete']`, a full
127
+ * CRUD grant. Either way: context supplied, answer not the animal answer,
128
+ * wrong in the GRANTING direction, because that predicate is arity-1 and
129
+ * identifies its collection from the request. AC9 asserts the first case on a
130
+ * live dispatch.
131
+ *
132
+ * Every predicate in this repo and in every consumer tree is arity-1 today,
133
+ * and there is no supported way for the caller to tell which kind it got; the
134
+ * boot-time arity warning that would surface it is abofs/stonyx-orm#213. Pass
135
+ * the context, and do not treat a resolved predicate's answer as
136
+ * model-specific until that predicate reads it.
137
+ *
138
+ * OWN PROPERTIES ONLY. A bare `this.accessFunctions[modelName]` walks the
139
+ * prototype chain, so `getAccess('constructor')` resolved `Object` and
140
+ * `getAccess('toString')` resolved `Object.prototype.toString` -- both
141
+ * callable, and the documented `predicate?.(request, ctx)` pattern then
142
+ * returned a TRUTHY value (`Object(request)` is the request), bypassing the
143
+ * `undefined`-means-deny contract entirely. Nothing in the ORM calls
144
+ * `getAccess` yet, so it was not exploitable as shipped -- but #207 takes the
145
+ * model name from the REQUEST BODY (`data.relationships.<key>.data.type`),
146
+ * which would have made a one-field body an authorization bypass. Guarded
147
+ * here at the read point rather than by constructing the map with a null
148
+ * prototype, because the field is public and reassignable and the guard has
149
+ * to hold whatever object it is holding.
150
+ *
151
+ * @param modelName - Model name as declared and stored (kebab-case).
152
+ * @returns The predicate, or `undefined` when no predicate could be resolved
153
+ * for that name. `undefined` is NOT "this model is unrestricted" -- see the
154
+ * note above. Treat it as deny.
155
+ */
156
+ getAccess(modelName: string): AccessFunction | undefined;
41
157
  startup(): Promise<void>;
42
158
  shutdown(): Promise<void>;
43
159
  static get db(): OrmDB | SqlDb;