fix(aggregation): the lifecycle cluster — flush stamps, behind-stamp catches up incrementally, the native rebuild finally gets invoked, deletes are never silently skipped
Some checks failed
CI / Node 22 (push) Has been cancelled
CI / Node 24 (push) Has been cancelled
CI / Bun (latest) (push) Has been cancelled

SELF-ENGINE-LIFECYCLE-SPRINT + BRAINY-PROD-LATENCY-TRIAD, the four asks:

(a) brain.flush() persists aggregation state stamped at the committed
    generation. The stamp used to advance only at close(), so a long-lived
    writer that flushes but never closes — the primary production shape —
    left every write window behind the stamp, and ANY unclean exit forced a
    whole-store backfill walk (per-entity work, measured >60s and
    door-starving on a 9k-row production brain) on the first stats call.

(b) BEHIND-stamp adoption becomes adopt + INCREMENTAL CATCH-UP: the exact
    missing window (stamp, committed] resolves its affected-id set from the
    fact log and reconciles each entity with time-travel before/after reads
    (asOf at both window bounds) through the same delta algebra the live
    hooks use — cost bounded by writes since the last flush, never store
    size, and exact under interleaving because reconciliation targets the
    FIXED window end while later writes chain through hooks. Oversized
    windows (>5000 affected) and unreadable windows demote to the announced
    rescan — never a silent partial serve.

(c) The native provider's parallel rebuildAggregate — on the contract since
    8.x but never invoked anywhere — is now the backfill walk's preferred
    door: one call per aggregate with source-matched entities, replacing
    the per-entity FFI stream.

(d) A delete whose before-image is unavailable can no longer SKIP the
    aggregation hook silently (counts drifted upward forever): both delete
    paths (remove() and transact) flag an exact rescan, loudly.

Pins: integration (flush stamp; unclean-exit reopen → exact counts through
an add + group-move + delete window with the walk spy proving ZERO
whole-store walks) + unit (provider rebuild invoked once with filtered
entities; flagAllForRescan; reconcile delta algebra). Gates: unit 1913/1913
· integration 760 · conformance 27/27.
This commit is contained in:
David Snelling 2026-08-05 15:49:12 -07:00
parent 607b6b56f2
commit 1dc861d299
5 changed files with 796 additions and 38 deletions

View file

@ -371,6 +371,15 @@ export class AggregationIndex {
*/
private pendingAdopt = new Set<string>()
/**
* Aggregates adopted with a BEHIND stamp: name the exact generation
* window `(from, to]` whose writes the adopted state has not seen. The
* owner (Brainy) drains this via {@link getPendingCatchUps} +
* {@link reconcileEntity} + {@link finishCatchUp} BEFORE serving queries
* cost bounded by the window's affected entities, never store size.
*/
private pendingCatchUp = new Map<string, { from: number; to: number }>()
/**
* In-flight rescan targets. While a name has a staging map, ALL
* contributions (the walk's and concurrent write hooks') land there instead
@ -437,25 +446,47 @@ export class AggregationIndex {
}
/**
* May this persisted state be ADOPTED? When the store exposes its committed
* watermark, the state's `sourceGeneration` must EQUAL it: behind means
* later writes are missing from the state (unclean shutdown); ahead means
* it counts writes that no longer exist (e.g. a fact-log truncation on a
* copied store pulled the watermark back). Either way: one exact rescan,
* said out loud never a silent adopt. Stores without the capability (and
* pre-stamp state on them) fall back to hash-only adoption.
* The adoption verdict for persisted state, against the store's committed
* watermark (SELF-ENGINE-LIFECYCLE-SPRINT ask (b) behind-stamp is no
* longer a whole-store rescan):
*
* - `'adopt'` stamp equals the watermark (clean), or the store has no
* watermark capability (hash-only adoption, the pre-stamp behavior).
* - `'catchup'` stamp is BEHIND the watermark (an unclean exit after
* later writes, or a long-lived writer whose last flush predates recent
* writes). The state is exact AS OF its stamp, so it is adopted and the
* missing window `(stamp, committed]` is reconciled INCREMENTALLY per
* affected entity via time-travel reads bounded by writes since the
* last flush, never by store size. The owner drains
* {@link getPendingCatchUps} before serving queries.
* - `'rescan'` no stamp (pre-stamp state on a stamped store) or stamp
* AHEAD of the watermark (e.g. a fact-log truncation on a copied store
* pulled the watermark back): the state over-counts unverifiably; one
* exact rescan, said out loud.
*/
private stateGenerationAdoptable(name: string, stateData: unknown): boolean {
private stateAdoptionVerdict(
name: string,
stateData: unknown
): 'adopt' | 'catchup' | 'rescan' {
const committed = this.storage.committedGeneration?.() ?? null
if (committed === null) return true
if (committed === null) return 'adopt'
const raw = (stateData as Record<string, unknown>).sourceGeneration
const stamped = typeof raw === 'number' ? raw : null
if (stamped === committed) return true
if (stamped === committed) return 'adopt'
if (stamped !== null && stamped < committed) {
this.pendingCatchUp.set(name, { from: stamped, to: committed })
prodLog.info(
`[Aggregation] '${name}': persisted state is at generation ${stamped}, store is at ` +
`${committed} — adopting and reconciling the ${committed - stamped}-generation window ` +
`incrementally (no store rescan)`
)
return 'catchup'
}
prodLog.warn(
`[Aggregation] '${name}': persisted state is at generation ${stamped ?? 'unstamped'} ` +
`but the store's committed generation is ${committed} — rescanning instead of adopting`
)
return false
return 'rescan'
}
private async loadPersisted(): Promise<void> {
@ -476,20 +507,21 @@ export class AggregationIndex {
const appHash = this.definitionHashes.get(def.name) || ''
if (appHash === savedHash && this.pendingAdopt.has(def.name)) {
const stateData = await this.storage.getMetadata(`${STATE_KEY_PREFIX}${def.name}__`)
if (
stateData &&
stateData.groups &&
this.stateGenerationAdoptable(def.name, stateData)
) {
const verdict =
stateData && stateData.groups
? this.stateAdoptionVerdict(def.name, stateData)
: 'rescan'
if (verdict !== 'rescan') {
const groupMap = new Map<string, AggregateGroupState>()
for (const group of stateData.groups as AggregateGroupState[]) {
for (const group of stateData!.groups as AggregateGroupState[]) {
groupMap.set(serializeGroupKey(group.groupKey), group)
}
this.states.set(def.name, groupMap)
this.pendingAdopt.delete(def.name)
this.needsBackfill.delete(def.name)
prodLog.info(
`[Aggregation] '${def.name}': adopted persisted state (${groupMap.size} groups) — no rescan`
`[Aggregation] '${def.name}': adopted persisted state (${groupMap.size} groups) — ` +
(verdict === 'catchup' ? 'incremental catch-up pending' : 'no rescan')
)
}
// No/invalid persisted state: stays in pendingAdopt and resolves
@ -504,22 +536,23 @@ export class AggregationIndex {
const currentHash = hashDefinition(def)
const stateData = await this.storage.getMetadata(`${STATE_KEY_PREFIX}${def.name}__`)
if (
stateData &&
stateData.groups &&
savedHash === currentHash &&
this.stateGenerationAdoptable(def.name, stateData)
) {
// Definition unchanged — load state
const restoreVerdict =
stateData && stateData.groups && savedHash === currentHash
? this.stateAdoptionVerdict(def.name, stateData)
: 'rescan'
if (restoreVerdict !== 'rescan') {
// Definition unchanged — load state (exact as of its stamp; a
// 'catchup' verdict reconciles the missing window incrementally).
const groupMap = new Map<string, AggregateGroupState>()
for (const group of stateData.groups as AggregateGroupState[]) {
for (const group of stateData!.groups as AggregateGroupState[]) {
const serialized = serializeGroupKey(group.groupKey)
groupMap.set(serialized, group)
}
this.states.set(def.name, groupMap)
this.needsBackfill.delete(def.name)
prodLog.info(
`[Aggregation] '${def.name}': restored definition + adopted persisted state (${groupMap.size} groups)`
`[Aggregation] '${def.name}': restored definition + adopted persisted state (${groupMap.size} groups)` +
(restoreVerdict === 'catchup' ? ' — incremental catch-up pending' : '')
)
} else {
// Definition changed or no saved state — start fresh and backfill from
@ -747,6 +780,119 @@ export class AggregationIndex {
this.dirty.add(name)
}
// ============= Incremental Catch-Up (behind-stamp adoption) =============
/** The aggregates adopted behind the watermark, with their exact missing windows. */
getPendingCatchUps(): Array<{ name: string; from: number; to: number }> {
return Array.from(this.pendingCatchUp, ([name, w]) => ({ name, ...w }))
}
/**
* Reconcile ONE entity's contribution across a catch-up window using the
* same exact delta algebra the write-time hooks use: remove the
* contribution the adopted state counted (the entity AS OF the stamp),
* add the contribution it should count (AS OF the window's end). `null`
* on either side means the entity did not exist then. Composes exactly
* with live hooks because every application is a precise old/new pair
* order between catch-up and post-window writes cannot drift the totals.
*/
reconcileEntity(
name: string,
id: string,
before: Record<string, unknown> | null,
after: Record<string, unknown> | null
): void {
const def = this.definitions.get(name)
if (!def) return
if (before && after) {
if (isAggregateEntity(after)) return
const oldMatches = matchesSource(before, def.source)
const newMatches = matchesSource(after, def.source)
if (this.nativeProvider && (oldMatches || newMatches)) {
this.applyNativeResults(
name,
this.nativeProvider.incrementalUpdate(name, def, after, 'update', before)
)
return
}
if (oldMatches) this.removeContribution(name, def, before)
if (newMatches) this.addContribution(name, def, after)
return
}
if (after) {
if (isAggregateEntity(after) || !matchesSource(after, def.source)) return
if (this.nativeProvider) {
this.applyNativeResults(name, this.nativeProvider.incrementalUpdate(name, def, after, 'add'))
} else {
this.addContribution(name, def, after)
}
return
}
if (before) {
if (isAggregateEntity(before) || !matchesSource(before, def.source)) return
if (this.nativeProvider) {
this.applyNativeResults(name, this.nativeProvider.incrementalUpdate(name, def, before, 'delete'))
} else {
this.removeContribution(name, def, before)
}
}
}
/** Whether the native provider offers the parallel whole-rebuild path. */
hasProviderRebuild(): boolean {
return typeof this.nativeProvider?.rebuildAggregate === 'function'
}
/** The catch-up window for `name` is fully reconciled; state is current. */
finishCatchUp(name: string): void {
this.pendingCatchUp.delete(name)
this.dirty.add(name)
}
/**
* A catch-up could not complete (window unreadable, affected set over the
* bound, ): demote to an exact rescan, loudly never serve un-reconciled.
*/
demoteCatchUpToBackfill(name: string, reason: string): void {
this.pendingCatchUp.delete(name)
this.needsBackfill.add(name)
prodLog.warn(`[Aggregation] '${name}': catch-up demoted to full rescan — ${reason}`)
}
/**
* Rebuild an aggregate through the native provider's parallel path
* (SELF-ENGINE-LIFECYCLE-SPRINT ask (c) `rebuildAggregate` existed on
* the provider contract but was never invoked; the JS walk fed
* per-entity FFI calls instead). Returns false when no provider rebuild
* exists the caller streams the JS walk as before.
*/
rebuildWithProvider(name: string, entities: Array<Record<string, unknown>>): boolean {
const def = this.definitions.get(name)
if (!def || !this.nativeProvider?.rebuildAggregate) return false
const rebuilt = this.nativeProvider.rebuildAggregate(
def,
entities.filter(e => !isAggregateEntity(e) && matchesSource(e, def.source))
)
this.states.set(name, rebuilt)
this.backfillStaging.delete(name)
this.needsBackfill.delete(name)
this.dirty.add(name)
return true
}
/**
* A write-path hook could not see the entity it needed (e.g. a delete
* whose before-image was unavailable): flag EVERY defined aggregate for
* an exact rescan, loudly the counts must never silently drift
* (SELF-ENGINE-LIFECYCLE-SPRINT ask (d): the gated hook used to SKIP).
*/
flagAllForRescan(reason: string): void {
for (const name of this.definitions.keys()) this.needsBackfill.add(name)
prodLog.warn(
`[Aggregation] all ${this.definitions.size} aggregate(s) flagged for rescan — ${reason}`
)
}
// ============= Write-Time Hooks =============
/**