fix(8.0): restore() reloads a native entity-id mapper before graphIndex.rebuild()

A logical snapshot restore could silently lose graph edges on a brain backed by
a native provider: the graph adjacency keys on the shared id-mapper's interned
ints (sourceInt/targetInt), which are derived + never persisted, and restore()
never reloaded the mapper — so a native graph rebuild resolved verb endpoints
against a stale/empty mapper and dropped edges.

restore() now calls entityIdMapper.rebuild() — a new OPTIONAL method on the
EntityIdMapperProvider contract — BEFORE rebuilding the indexes, so a native
mapper reloads its int<->uuid from the restored binary KV first and the graph
resolves endpoints correctly. The JS mapper deliberately has NO rebuild() and
needs none: MetadataIndex.rebuild() re-derives it from the restored entities via
append-only getOrAssign, consistently with the bitmaps it builds — forcing a
reload there would blank a still-referenced mapping and break find() after a
same-instance restore.

Contract: graphIndex.rebuild() resolves sourceId/targetId -> ints through the
shared mapper itself (brainy does not re-feed resolved endpoints); brainy's only
job is to ensure the mapper is reloaded first.

Test: relationships survive a snapshot round-trip (db-mvcc.test.ts). 84
generation/temporal/visibility tests green; tsc clean.
This commit is contained in:
David Snelling 2026-06-23 12:02:17 -07:00
parent 3783e61b30
commit 4d0b64f455
4 changed files with 72 additions and 2 deletions

View file

@ -6408,8 +6408,23 @@ export class Brainy<T = any> implements BrainyInterface<T> {
await this.storage.restoreFromDirectory(path) await this.storage.restoreFromDirectory(path)
await this.generationStore.reopenAfterRestore(floorGeneration) await this.generationStore.reopenAfterRestore(floorGeneration)
// Every in-memory index is now stale — rebuild all three from the // If the entity-id mapper is a NATIVE provider with a `rebuild()`, reload it
// restored canonical records (each rebuild clears its own state first). // from the restored snapshot FIRST. The graph index resolves every verb
// endpoint (sourceId/targetId → stable int) through this mapper, so a native
// adjacency (keyed on those ints) needs the mapper to reflect the snapshot's
// assignments BEFORE graphIndex.rebuild() — otherwise edges resolve to
// stale/missing ints and are silently dropped (CTX-BR-RESTORE-REBUILD). The
// JS mapper has no `rebuild()` and needs no reload: MetadataIndex.rebuild()
// re-derives it from the restored entities via append-only getOrAssign,
// consistently with the bitmaps it builds.
const idMapper = this.metadataIndex.getIdMapper() as { rebuild?: () => Promise<void> }
if (typeof idMapper.rebuild === 'function') {
await idMapper.rebuild()
}
// Every in-memory index is now stale — rebuild all three from the restored
// canonical records (each rebuild clears its own state first). The graph
// rebuild resolves endpoints via the (now-reloaded, for native) mapper.
await Promise.all([ await Promise.all([
this.metadataIndex.rebuild(), this.metadataIndex.rebuild(),
this.index.rebuild(), this.index.rebuild(),

View file

@ -771,6 +771,18 @@ export interface VectorIndexProvider {
*/ */
export interface EntityIdMapperProvider { export interface EntityIdMapperProvider {
init(): Promise<void> init(): Promise<void>
/**
* @description OPTIONAL: reload the uuidint mapping from the CURRENT storage,
* discarding in-memory state called by `brain.restore()` AFTER the storage
* has been replaced from a snapshot and BEFORE the graph index rebuilds, so
* the graph resolves each verb endpoint through a mapper that reflects the
* snapshot's assignments (without it, native adjacency edges keyed on these
* ints resolve to stale/missing ints and are silently dropped). A native
* mapper reloads from its restored binary KV; the JS fallback re-reads its
* persisted metadata. Optional so a mapper that already reloads via `init()`
* stays compatible `restore()` falls back to `init()` when this is absent.
*/
rebuild?(): Promise<void>
getOrAssign(uuid: string): number getOrAssign(uuid: string): number
getUuid(intId: number): string | undefined getUuid(intId: number): string | undefined
getInt(uuid: string): number | undefined getInt(uuid: string): number | undefined

View file

@ -344,6 +344,17 @@ export class EntityIdMapper implements EntityIdMapperProvider {
await this.flush() await this.flush()
} }
// No `rebuild()` on the JS mapper by design (CTX-BR-RESTORE-REBUILD): after a
// `brain.restore()`, `MetadataIndex.rebuild()` re-derives this mapper from the
// restored entities via append-only `getOrAssign` (the ints it picks are
// internally consistent with the bitmaps it builds), so the JS path needs no
// explicit post-restore reload — and forcing one (blanking + reloading) would
// drop a still-referenced mapping when the snapshot omits the mapper file. The
// `rebuild()` reload is the NATIVE mapper's concern: cor's binary KV holds the
// authoritative int↔uuid the native adjacency is keyed on, so it must reload
// from the restored KV before the native graph rebuild. `restore()` therefore
// calls `rebuild()` only when the provider implements it (see brainy.ts).
/** /**
* Get statistics about the mapper * Get statistics about the mapper
*/ */

View file

@ -1096,6 +1096,38 @@ describe('8.0 Db API — generational MVCC', () => {
await after.release() await after.release()
}) })
it('restore() rebuilds graph adjacency — relationships survive a snapshot round-trip (id-mapper reloaded before graph)', async () => {
const { brain } = await openFsBrain()
const a = uid('rr-a')
const b = uid('rr-b')
const c = uid('rr-c')
await brain.transact([
{ op: 'add', id: a, type: NounType.Person, data: 'a', vector: vec(1), metadata: {} },
{ op: 'add', id: b, type: NounType.Person, data: 'b', vector: vec(2), metadata: {} },
{ op: 'add', id: c, type: NounType.Document, data: 'c', vector: vec(3), metadata: {} },
{ op: 'relate', from: a, to: b, type: VerbType.RelatedTo },
{ op: 'relate', from: a, to: c, type: VerbType.References }
])
expect((await brain.related(a)).map((r) => r.to).sort()).toEqual([b, c].sort())
const snapDir = path.join(makeTempDir(), 'rel-snapshot')
const db = brain.now()
await db.persist(snapDir)
await db.release()
// Drop a + its edges live so the restore must genuinely RECONSTRUCT them.
await brain.remove(a)
expect(await brain.related(a)).toEqual([])
await brain.restore(snapDir, { confirm: true })
// a's relationships are back: the graph adjacency rebuilt by resolving each
// verb's endpoints (sourceId/targetId → ints) through the id-mapper that
// restore() reloaded from the snapshot BEFORE graphIndex.rebuild(). Without
// that reload, a native adjacency keyed on those ints loses the edges.
expect((await brain.related(a)).map((r) => r.to).sort()).toEqual([b, c].sort())
})
it('Db values are first-class: Brainy.open() + chained with() overlays compose', async () => { it('Db values are first-class: Brainy.open() + chained with() overlays compose', async () => {
const brain = await Brainy.open({ requireSubtype: false, storage: { type: 'memory' } }) const brain = await Brainy.open({ requireSubtype: false, storage: { type: 'memory' } })
brains.push(brain) brains.push(brain)