From 7f3cfb060d3fcbf0ec72874d813dce89ddb5d468 Mon Sep 17 00:00:00 2001 From: Mario Zechner Date: Wed, 28 Oct 2026 14:55:45 +0000 Subject: [PATCH] docs(chord): slice 6b spec revised after the first spec review --- packages/chord/docs/next/slice-6b.md | 443 ++++++++++++++++----------- 1 file changed, 257 insertions(+), 186 deletions(-) diff --git a/packages/chord/docs/next/slice-6b.md b/packages/chord/docs/next/slice-6b.md index 02f685ac0..52d9cd21e 100644 --- a/packages/chord/docs/next/slice-6b.md +++ b/packages/chord/docs/next/slice-6b.md @@ -1,245 +1,316 @@ # Slice 6b: Durable and jobs on the store (spec) -Status: **decided, ready for the spec reviews** (2026-10-27; D1-D9 decided by the owner), written by the redesign session after reading -Durable's `Storage` interface (`packages/durable/src/types.ts`), its SQLite storage, schema and migrations, the storage -conformance suite and benchmark, `session/session.ts`, `session/transaction.ts`, `session/forks.ts`, and Chord v2's -store engine, applier, schema, commit step and jobs, in full. Durable's memory, JSONL and IndexedDB storages were not -read (they stay as they are, S15). +Status: **revised after the first spec review** (2026-10-27): D1-D9 and R1-R4 decided by the owner (decision log, +entries "Slice 6b …" and "owner names and claims"); the four reviews (`/tmp/s6b-review-{astra,fable,opus,sol}.md`) +kept every decision and found the gaps this revision closes. Written by the redesign session after reading, in full, +Durable's `types.ts`, its SQLite storage, schema and migrations, the storage conformance suite and benchmark, +`session/session.ts`, `session/transaction.ts`, `session/forks.ts`, the scheduler's restarts, `writeAtClose` and `abort`, +and Chord v2's store, engine, applier, commit step and jobs. Durable's memory, JSONL and IndexedDB storages stay as they +are (S15). -The single source for building 6b, together with: `design.md` read in full; the decision log -(`packages/chord/docs/next/decisions.md`, later entries win), in particular "The store", S1, S2, S4, S10, S11, S13, S15, -W1, W12, W13, "Slice 6 first decisions", "Durable observers", "Durable's synchronous observers", "Text is plain UTF-8", -"Failures are watch calls", the "One vocabulary" entries, "The performance and memory bar"; `slice-6a.md` (what 6a -built); the feature map and the carry-over of `slices.md` (slice 6b); the root `AGENTS.md` "Chord v2 Code Style". +The single source for building 6b, together with `design.md` read in full; the decision log `decisions.md` (later +entries win), in particular "The store", S1, S2, S4, S10, S11, S13, S15, W1, W12, "Slice 6 first decisions", "Durable's +synchronous observers", "Text is plain UTF-8", "Failures are watch calls", the "One vocabulary" entries, "The performance +and memory bar", and every "Slice 6b" entry; `slice-6a.md` (what 6a built); the feature map and the carry-over of +`slices.md` (6b); the root `AGENTS.md` "Chord v2 Code Style". 6b puts Durable on the store (`StoreStorage`), follows Durable's views as Chord sources (`self.follow`), persists jobs, -and extends the applier to every write kind. After 6b a worker has one store: Chord's objects and Durable's Session -commit through one queue, and a commit of either is one seq in one outbox. +and extends the applier to every write kind. A worker then has one store: Chord's objects and Durable's Session commit +through one queue, and a commit of either is one seq in one outbox. ## What an app can do after this slice ```ts -// the start code of a session's worker (p3 base code) +// p3 worker base code const store = await openStore(sqliteBackend(db), ctx); -const worker = createHost

({ address: Session({ id }), principal, store, onError }); +const host = createHost

({ address: Session({ id }), principal, store, onError }); +await host.load([{ name: "@p3/core", facet: coreWorkerFacet({ init, store }) }, ...plugins], ctx); -// the core facet: the harness object opens Durable on the same store -host.serve(Harness, (_, self) => { +// the core facet: the harness object opens Durable on the same store, its data under the harness's own type +host.serve(Harness, (_, self) => { // local, empty params (P3) const harness = self.open(async (context) => Harness.open(await StoreStorage.open(store, Harness.name, context), { models, registry, … }, context), (h, _reason, context) => h.close(context)); - const view = self.follow((observer, ctx) => harness.value.observeConversation(id, observer, ctx)); // section 2 - return { … }; + const view = self.follow((observer, ctx) => harness.value.observeConversation(rootId, observer, ctx, { min: 50 })); + return { view }; }); -// a persisted job: survives a crash +// a persisted job: survives a crash, a clean close and its object's retry const ImportJob = defineJob("notes.import", { version: 1, args: obj({ path: str() }), result: obj({ count: int() }), - state: ImportState, persist: { resume: "rerun" }, + state: ImportState /* persist: true, checkpointWhen */, persist: { resume: "rerun" }, }); const imports = self.job(ImportJob, { run, finish }); -await imports.abort(id, ctx); // stored; Decision needed (D7): the result +await imports.abort(id, ctx); // "marked" | "terminal", stored +await ctx.state.change((d) => { d.output = text; }, ctx, { persistence: { kind: "deferred" } }); ``` -## 1. `StoreStorage`: Durable's `Storage` on the store +## 1. Changes to 6a's store -Durable's `Storage` (types.ts, "Atomic persistence boundary for Session records") implemented over a `ChordStore`, so -Durable's Session runs unchanged on it and passes Durable's storage conformance suite -(`createStorageConformance`), across reopen too. +Every item below changes 6a's code; nothing is deployed, so the schema and the archive format change in place. -**Where it lives (D1, owner, 2026-10-27: "they are only used if durable is used"):** `StoreStorage` belongs to -Durable (D9 makes `durableView` a generic Chord primitive, `self.follow`), in its own `src` (Chord v2 is in pi upstream since `93029a0f1`), importing -`@earendil-works/chord/next` (Chord's `./next` package export comes with them). Durable already depends on Chord and -Chord imports nothing of Durable (S13); Durable's conformance suite runs in Durable's tests and CI. +### 1.1 Names and claims (decisions "R2", "owner names and claims") -### 1.1 Records +- Tables are `":"` (`"notes.store:edits"`), document owners `""`; no host kind. The applier and + import check two name parts. +- Every owner name is claimed once per store (in memory, for the store's lifetime): an object claims its type when it + first declares persisted data; `StoreStorage.open` claims its name; a second claim is refused `invalid/args`; a + claim is released by its owner's end (`StoreStorage.close`, the object's end), so a new Session can reopen over the + same store. -Durable's four record tables (conversations, entries, tasks, submissions) and its ID namespace. +### 1.2 One id allocator (D3) -**Layout (D2, owner, 2026-10-27: "the whole purpose of this slice"):** declared tables of the store, written as -ordinary `put` writes: `defineTable`s that `StoreStorage` declares for itself under its owner name (`StoreStorage.open(store, owner, ctx)`, -the name claimed once per store; SQL names `":
"`, e.g. `"p3.harness:entries"`; decision "owner names -and claims"), each row the record's indexed columns (`id`, `conversation_id`, `kind`, `status`, …) plus -the whole record in a `json(...)` column. The outbox, export, import and the applier handle them as every table; -Durable's layout changes later through `evolveTable`. The partial head index becomes an index on `(conversation_id, -has_head, id)` (index columns are non-`opt`); indexed strings that must stay lossless (`kind`, `requestId`, document -kinds and keys) keep Durable's JSON encoding in their columns. +- The store owns one synchronous allocator: `mintId()` takes the next id; the commit step takes its documents' ids from + the same allocator when a creation applies (no copy in the `Step`, no overwrite after `COMMIT`); explicit ids in + staged writes raise it at staging (`max`); a refused step leaves its ids unused (gaps, S10). Each commit persists + the allocator's current value. +- It starts at 2 (id 1 is Durable's root conversation). It saturates at `Number.MAX_SAFE_INTEGER`: an explicit write of + that id is accepted, the persisted value stays `MAX_SAFE_INTEGER`, every later `mintId` rejects "ID space is + exhausted", and a Chord creation at the saturated allocator refuses its own commit (1.4). +- Outbox commits carry the allocator's value after them (`{ seq, nextId, writes }`); the archive manifest's + `nextDocumentId` is `nextId`; import and batch apply restore it, so a replica mints above every id it holds. -- **Writes:** every `StorageWrite` of a table record becomes a `put` of its row (Durable replaces complete records; - conversations and entries are immutable and written once). `record_ids` is a declared table too (`id`, - `record_type`), one `put` per new id (decision "Slice 6 first decisions" (3)). -- **Reads:** `conversation`, `entry` (with its `commitSeq`: a column), `findLatestHeadMarker`, `scanEntries` (both - orders, the fork chain walked as `SqliteStorage` does), `task`, `scanTasks`, `submission`, `scanSubmissions`, - `submissionByRequest`, `scanConversations`: queries over the declared tables, Durable's cursors (`{ after, order }`, - `scan.ts`) kept as they are (built from the last id, not from Chord's table cursors). -- **Validation in the step** (D4): what `SqliteStorage` checks inside its transaction (`checkGlobalIds`, - `prepareDocumentActions`, `checkDocumentActions`) runs inside the store's step, which rolls back a refused commit. +### 1.3 The commit step -### 1.2 IDs +- **A staged write returns the writes it recorded:** `StagedWrite.apply(step): Promise`. + `StoreStorage` stages one composite write per Durable commit (1.3 of section 2). +- **Order:** within the step, writes apply in the order staged, except that a staged write that frees a live address + (a retirement of another incarnation) is applied before the creation at that address (rule 9's reason, the partial + unique index); Chord's own staging keeps 6a's order. The outbox records the writes in applied order; the applier + replays them in that order (no re-sort). +- **Stamps:** a staged write can take values only the step knows (`step.seq`): Durable's `entries.commit_seq`. The + outbox records the resolved values. +- **Adopt in the owner's batch:** the host gives the store its batch when it attaches (`attach({ failed, batch })`); + an adapter's fence runs its `adopt` in `batch.within(…)` (joins a running batch, opens one otherwise); Chord's own + transactions keep their `batch.commit`. Without a host, `adopt` runs plain. -**One counter per store (D3, owner, 2026-10-27):** the store's counter (`next_document_id` renamed `next_id`, the -manifest's `nextDocumentId` renamed `nextId`) numbers every id: `StoreStorage.mintId()` takes from it in memory, -explicit ids in Durable's writes raise it, Chord's commit step takes its documents' ids from it, and every commit -persists it as a high-water mark. `record_ids` stays for Durable's "ID n already belongs to entry" check. Durable's ids -rise in its creation order (its scans and windows rely on it); Chord's ids are gaps between them. Spec reviewers check -that nothing relies on Chord's document ids rising with seq. +### 1.4 Errors: Durable's Storage semantics (D4) -### 1.3 Documents +- The step records the error its own code threw inside the transaction callback. A rejection of the backend + transaction that is that very error is a **refusal**: the facade rolled back (a failed `ROLLBACK` reaches us as + another error, an `AggregateError` on Node), the commit rejects, nothing is adopted, the allocator keeps its raised + value, the store goes on. Any other rejection (`BEGIN`, `COMMIT`, `ROLLBACK`, a throwing `adopt`) is an unknown + outcome: the store stops (`storage_failed`). The Cloudflare facade is checked to keep the same identity; a test + injects failures at `BEGIN`, in the body, at `COMMIT` and at `ROLLBACK` on both drivers' facades. +- **The refusal's error reaches the owner unchanged** (Durable's suite matches its messages); the store classifies it + for Chord's owners: an `int` `CHECK` failure (SQLite's extended code `SQLITE_CONSTRAINT_CHECK`) is `invalid/args` and + fails only that transaction; every other refusal of a Chord commit (the engine's counted `UPDATE`/`DELETE`, the + live-address index, a primary key, the saturated allocator, a backend error in a statement) fails the object live + (`object_failed`, `broken`: memory and disk disagree). +- **6a's owners:** a refused `self.tx` rejects with the error above and the object fails as above; a refused + migration commit (a chunk, (A) or (Z)) fails the activation `broken`, the table stays `migrating`, committed chunks + stay; a refused job start rejects the start; a refused outcome or finish commit fails the object (records and + memory disagree); a refused commit that carried held changes (D8) fails the object. +- **Reads:** a read's callback returns data, never throws for a request: an adapter's request checks (an unknown + conversation, a cursor, history a document does not keep) happen after the guarded read returned, from tagged + results, so only a backend error fails the store. The store offers a read transaction for reads of several queries + that must see one state (S4). -Durable's documents on the documents engine (S13): address `(namespace "durable", owner, kind, key?)`, owner the -scope as a string (`session`, `conversation:7`, `task:31`), kind and key JSON-encoded as Durable's indexed strings. -6a built `latest`; 6b adds what Durable needs: +### 1.5 The documents engine -- **`rewindable` records** keep every revision and are not deleted at retirement: `retired_at` is set, revisions - stay (decision "Writes carry their versions" already defined the archive fields). -- **Reads at a point:** `findDocument(address, at)`, `document(id, at)` (the base at or below `at` plus the deltas up - to it, in one read transaction, S4), `scanDocuments(scope, at, kind?)`, with Durable's half-open lifetime - (`createdAt <= at < retiredAt`); a historical read of a `latest` (or session or task) record is a - `StorageRequestError` ("does not retain historical content"). -- **`document.copy`** (S11): `StoreStorage` reads the source at its point before the step and enters the step with a - plain create carrying the base; the conformance rules (source not changed in the copy batch, a matching record) are - checked there. -- **Stored version and `deltasSinceBase`** as `StoredDocument` reports them; Durable's records carry no fingerprint: - the engine's fingerprint column takes `""` for Durable records. -- **Export** carries every revision of a `rewindable` record, retired ones included, with its seq (format 1's - fields); import and the applier take them. +- **Addresses:** `(owner, namespace, kind, key?)`: `owner` the claimed name; `namespace` the adapter's partition + (Chord: `chord` for documents, `chord.jobs` and `chord.job-states` for persisted runs; Durable: its scope, + `session`, `conversation:`, `task:`); `kind` and `key` as the adapter writes them (Durable JSON-encodes both, + as its indexed strings). One live record per address (the partial unique index). +- **`history` per record (R1):** `none` (Chord: older revisions dropped at a new base; row and revisions deleted at + retirement), `latest` (Durable's session, task and latest conversation documents: older revisions dropped; the row + kept with `retired_at` at retirement), `rewindable` (nothing dropped; the row kept). `retireDocument` reads the record + and applies its rule. +- **Explicit ids:** a creation may carry its id (Durable mints ahead); otherwise the allocator gives one. +- **Changes without a known revision:** an adapter whose writes carry no revision (Durable) stages a change that reads + the record in the step and uses its revision; the counted conditional `UPDATE` stays the guard. +- **`meta`:** Durable's `fork` policy and `history` details live in the record's `meta`. +- **Reads at a point:** `find(address, at)`, `load(id, at)` (the base at or below `at` plus the deltas up to it, its + version the base's), `list(owner, namespace, at, kind?, limit, after?)` by id, with the half-open lifetime + (`createdAt <= at < retiredAt`); records carry `retiredAt`. New indexes (store schema version 2): + `(owner, namespace, kind, key, created_at)`, `(owner, namespace, id)`, revisions `(document_id, kind, seq)`; their + cost on Chord's commit gate is measured. -### 1.4 Commits +### 1.6 The archive and batch apply -- `StoreStorage.commit(writes, adopt, ctx)`: validates (1.1), resolves copies (1.3), stages the writes (table `put`s, - document creates, changes and retirements) and calls `store.commit(staged, { admit: () => true, adopt: (seq) => - adopt(seq) }, ctx)`. The store's step runs `adopt` inside `host.batch.commit(...)` when the store has a host, so - Durable's observers mark Chord sources in the step (W1, "Slice 6 first decisions" (1)); `adopt` never throws - (Durable's rule). -- **Durable's seq is the store's seq** (gaps allowed: Chord's commits take seqs too). -- **Errors:** +- **Format 1 gains** (nothing deployed): `document.create` carries `history`, `retiredAt?` and `meta`; a + `document.revision { id, seq, kind, version, content }` part per kept revision, in seq order after its create; for + `none` and `latest` records only the current base and its deltas are exported (as 6a), for `rewindable` every + revision, retired records included. Import checks: revisions in seq order, a base first, `seq <= lastSeq`, lifetimes + at one address never overlapping (a second live record is still refused), versions per revision; the `end` count + includes revisions. +- **Batch apply** (`applyBatch(store, commits, ctx)`, internal): only on a store with no attached host (no owner's + memory to update); `commits[0].seq` follows `lastSeq` (S5); one backend transaction for the batch, every write applied + with its commit's seq, `lastSeq` and `nextId` set from the last commit; the applier handles every `StoreWrite` kind + (`alter`, `dropIndex`, `update`/`updateWhere`/`delete`/`deleteWhere` with their recorded counts and versions, + `document.change`, `document.retire`, `document.revision`), checked against the store's records (a trust boundary); + a count that differs from the recorded one stops applying and marks the store diverged (S17): every later + `applyBatch` rejects. A test replays a store's outbox into an empty store and compares the exports, manifest + included. -**Errors (D4, owner, 2026-10-27):** the store takes Durable's `Storage` semantics (design §10 "Store errors"): a -step that fails before `COMMIT` rolls back and rejects that commit only; the store stops only on an unknown outcome -(`COMMIT` or `ROLLBACK` failed), a throwing `adopt`, or a read's backend error. Owners react as Durable's Session: -Durable's Session fails (`GuardedStorage`, unchanged); a Chord object whose commit the step refused fails live -(`object_failed`, `broken`); an `int` `CHECK` failure fails only its transaction (`invalid/args`). `StoreStorage` runs -Durable's commit checks in the step, as `SqliteStorage` does, so the conformance suite's refused commits leave the -store usable. 6b changes 6a's step and object failure accordingly. +## 2. `StoreStorage`: Durable's `Storage` on the store (D1, D2) -- `mintId`, `close(context)` (closes nothing of the store: the app owns it; later calls reject "closed"), reads after - a store failure reject as Durable's `GuardedStorage` expects. +In Durable's `src` (an ordinary pi module), importing `@earendil-works/chord/next` (Chord's `./next` export comes +with it: the store and the adapter surface below, marked internal for adapters). It passes Durable's storage +conformance suite on SQLite `:memory:` and across reopen (a wrapper that closes and reopens the store file after +every commit, Durable's reopen harness), and Durable's Session and harness test files run on it. -### 1.5 Statements per commit +### 2.1 Opening and the adapter surface -A target to measure, as 6a's: an entry append is the `record_ids` put, the entry put, `lastSeq` (plus the outbox); a -document delta is 6a's three statements. `SqliteStorage` does the same work today (`record_ids`, the row, metadata). +`StoreStorage.open(store, owner, ctx)`: claims `owner` (1.1); declares its tables through the store's table +preparation (the path `self.table` takes: created at the newest version or migrated); raises the allocator to +`max(record_ids.id) + 1` (one index read; covers stores built by import or batch apply); attaches to the store's host +batch if there is one. Chord exports for it: `StoreOwnerTables` (declare and prepare tables under an owner), +staged table writes and engine writes, the step reader, the read transaction, `mintId`, the commit with a fence. Its +`close(context)` releases the claim and refuses every later call ("StoreStorage is closed"); the store stays open (the +app owns it). -## 2. `self.follow`: Durable's views as Chord sources (D9, owner, 2026-10-27) +### 2.2 Tables -Durable's synchronous observers (`observeConversation`, `observeTaskGraph`, `observeDoc`; decision "Durable's -synchronous observers") have Chord's watch shapes (`advance(call, ctx)` with a `WatchCall`, `end(end)` with a -`WatchEnd`, pi `5f312dac4`), so Chord follows them with one generic primitive and Durable needs no adapter: +Every row carries the record's indexed columns and the whole record as `json(unknown(), { version: 1 })`; indexed +strings that must stay lossless (`kind`, `request_id`) hold Durable's JSON encoding; index columns are non-`opt`, so +absent values take sentinels that no encoded value can be (`0` for an absent owner id, ids start at 2; `""` for an +absent request id, JSON-encoded strings start with `"`). + +| Table | Key | Columns | Indexes | +| --- | --- | --- | --- | +| `record_ids` | `id` | `record_type str` | | +| `conversations` | `id` | `owner_conversation_id int`, `owner_task_id int`, `record` | `(owner_conversation_id, id)`, `(owner_task_id, id)` | +| `entries` | `id` | `conversation_id int`, `has_head int` (0/1), `commit_seq int` (stamped), `record` | `(conversation_id, id)`, `(conversation_id, has_head, id)` | +| `tasks` | `id` | `conversation_id int`, `kind str`, `status str`, `abort_requested int`, `background int`, `record` | `(status, id)`, `(conversation_id, id)`, `(kind, id)`, `(abort_requested, id)`, `(background, id)` | +| `submissions` | `id` | `conversation_id int`, `request_id str`, `status str`, `record` | `(conversation_id, request_id)`, `(conversation_id, id)`, `(status, id)` | + +### 2.3 Commits + +`commit(writes, adopt, ctx)` stages one composite write and calls the store's commit with `{ admit: () => true, +adopt: (seq) => adopt(seq) }` (adopted in the host's batch, 1.3). Inside the step the composite write does what +`SqliteStorage` does inside its transaction, with the step's reader: `checkGlobalIds` (an id already claimed by +another record type, a conversation, entry or document written twice: Durable's messages), the per-document actions +of `prepareDocumentActions`/`checkDocumentActions` (one content command per document, a delta on its version, one live +incarnation per address: "already has a current incarnation"), then the table `put`s (`record_ids` only for ids not yet +claimed by the same type), then per document: its creation (explicit id) or content, then its retirement; a +create-plus-retire is an empty lifetime (`createdAt = retiredAt = seq`). Copies (`document.copy`) are resolved before +the step (S11): the source read at its point, checked against the copy's record ("changed in the copy batch", +kind, key, history, fork), entered as a creation carrying the base. A refused check rejects with Durable's message (a +refusal, 1.4): the store goes on, Durable's Session fails (`GuardedStorage`). + +### 2.4 Reads + +`conversation`, `entry` (both overloads, `commitSeq` from its column), `findLatestHeadMarker`, `scanEntries` (both +orders, the fork chain walked as `SqliteStorage` does), `task`, `scanTasks`, `submission`, `scanSubmissions`, +`submissionByRequest`, `scanConversations`: queries over the tables, Durable's cursors (`scan.ts`) as they are; +`findDocument`, `document` (one read transaction), `scanDocuments`: the engine's reads at a point (1.5). Request errors +(`StorageRequestError`) are thrown after the guarded read (1.4). + +### 2.5 Cost + +Statements per commit, as the gate measures them: an entry append, a `record_ids` check read, the `record_ids` put, +the entry put, the meta update (4; `SqliteStorage` 5); a document delta, the record read, the revision insert, the +counted update, the meta update (about 4-5; `SqliteStorage` about 7). Per `put` one compiled row check and the JSON +encoding of indexed strings. + +## 3. `self.follow`: Durable's views as Chord sources (D9, R4) + +`self.follow(attach)` in a factory returns a `Watchable` (a state member, a derive source, an `on` source); +`attach(observer, ctx)` returns `Promise<{ value, detach } | undefined>` (Durable's `Observation`; `undefined` for an +absent `observeDoc` address). In 6b it is used inside the object that opened Durable (a native's value); reaching +Durable from other objects is slice 8's (the harness in a registry entry, carry-over). + +- **Activation:** after the natives opened, before the members are checked and derives compute: Chord makes the + observer and calls `attach`; calls arriving before it resolves are queued and applied in order on `value`; a rejected + or `undefined` attach fails the activation like a failed dependency (`transient`); a failed activation detaches + what attached. The attempt's end detaches. +- **Calls:** a call inside a running batch (Durable's `adopt` on the store) applies the queued calls first, then marks + the source in that batch; a call outside any batch (Durable's deferred commits, `#hold`, and the task graph's + `blocked` refresh, even on the store; any Durable storage that is not the store) is queued and one batch on the next + microtask applies everything queued. A derive over two views of one commit sees both changes in one computation. +- **Ends (R4):** `failed` fails the object `transient` (its retry reopens Durable and attaches again); `ended` is final: + a pending retirement call reaches the source's watches first (Durable's `endAfterDelivery`), then they end `ended`; + the member keeps its last value; a new `watch` on an ended follow source opens at that value and ends `ended` at once. + +## 4. Persisted jobs (D5-D8, R3) + +`persist: { resume: "rerun" | "abandon" }` only in `local` types with empty params (P3, at `self.job`); a persisted job +needs a state definition with `persist: true` and `checkpointWhen` if it has state, and a JSON-only result (no `bytes()`, +refused at declaration). + +### 4.1 Records + +Each unsettled run is a keyed persisted document in namespace `chord.jobs` (kind the job name, key the run id, +`history: none`); its state is a keyed persisted document in namespace `chord.job-states` (same kind and key), so two +jobs sharing a state definition never meet. The record: ```ts -host.serve(ChatView, ({ conversation }, self) => { - const harness = self.get(Harness, {}); - const view = self.follow((observer, ctx) => - harness.value.observeConversation(conversation, observer, ctx, { min: 50 })); - return { view, title: self.derive([view], ([v]) => v.title) }; -}); +{ version, fingerprint, args, key?, trace?, + phase: "pending" | "running" | "stopped" | "finishing" | "blocked", + restarts, finishFailures, abort?: true, blocked?: "finish_failed" | "version", + outcome?: { kind: "done", result } | { kind: "failed", error: StoredJobError } } +// StoredJobError: { kind: "threw", error: WireError } | { kind: "aborted", reason } | { kind: "restart_limit", restarts } ``` -- `self.follow(attach)` returns a `Watchable`: a state member, a derive source, an `on` source. `attach(observer, - ctx)` returns `Promise<{ value, detach }>` (Durable's `Observation`). -- **Attaching** is that one call during activation: Chord makes the observer, calls `attach`, takes `value` as the - source's value (calls arriving before the promise resolves are held and applied after it), and the object goes live. - The attempt's end calls `detach()`. An attach that rejects, or resolves `undefined` (`observeDoc` of an absent - address), fails the activation like a failed dependency. -- **Calls:** `advance` records the call. Inside a batch (on the store, Durable's `adopt` runs in the host's batch) the - source is marked in the step and its frame goes out in that flush; outside a batch (a Durable storage that is not - the store) the source collects the calls and opens one batch on the next microtask. A derive over two views of one - Durable commit sees both changes in one computation. -- **End:** Durable's failures are final: `end({ kind: "failed" })` ends the watches `failed`, `ended` ends them - `ended`. +Errors are stored in the wire error format (a `PluginError`'s code, message and details survive); a stored outcome is +rebuilt as the in-memory one. -## 3. Persisted jobs +### 4.2 Transitions -Design §15 "Persistence": `persist: { resume: "rerun" | "abandon" }`, only in `local` objects with empty params (P3, -at declaration); a run's record and state document in the owner's store; a recorded outcome resumes only `finish`; an -abort request is persisted; a restart limit; the job fingerprint stored with every record; the start trace carrier. +| Event | Writes (one commit unless noted) | After | +| --- | --- | --- | +| start | record `pending`, state `initial()` | invoked on the live attempt | +| reservation (before `run`) | record `running` | `run` invoked (a crash after it counts) | +| `ctx.state.change` | state; the record's `restarts: 0` only when it was not 0 | | +| abort (D7) | record `abort: true` (on the line); the signal fires after the commit, outside the batch; the line is released, then `abort` joins `run` | `"marked"` | +| `run` settles | held state flushed (D8), record `finishing` with the outcome | `finish` on the live attempt | +| `finish` fails | record `finishFailures + 1`; at 5 `blocked: "finish_failed"` | retried after the backoff, a blocked run on reload | +| `finish` succeeds | record deleted, state retired (Q9), with what `finish` wrote | the run settles | +| clean stop (host close, a native restart, R3's seal) | the stopped runs written `stopped`, held state flushed, by the close writer | | +| activation (once per object, not per retry) | each `running` record: `restarts + 1` (or, at 5, `failed { restart_limit }` and `finishing`, `job_interrupted` reported); `stopped` and `pending`: back to `pending` | maps, keys, active count, `running` rebuilt before anything runs; `rerun` invokes, `abandon` records `failed { aborted, restart }`; an abort-marked run finishes `aborted` without `run`; `finishing` resumes `finish`; `blocked` stays | -- **Where records live:** +- **The close writer:** at the object's end, after the drain and before the documents drop, one commit past the seal + (Durable's `writeAtClose`): the runs stopped by the end written `stopped`, held state flushed; a run whose `run` + outlived the drain is fenced and written `stopped` too; blocked persisted runs keep their records (dropped only by + `abort`). In-memory runs keep 6a's rules. +- **Pinning:** records and state documents of unsettled runs stay in memory (never evicted): `ctx.state.value` and + `running` read them synchronously. +- **Fingerprint:** the job's fingerprint (args, result, state) is stored; a record of the job's version with another + fingerprint is `blocked: "version"` (resolved by slice 7). -**Where records live (D5, owner, 2026-10-27):** each unsettled run's record is a keyed persisted document of its -object in the engine's namespace `chord.jobs` (owner the object's owner string, kind the job name, key the run id: -`{ version, fingerprint, args, key?, phase, outcome?, abort?, restarts, trace? }`); Chord sets that namespace, a plugin -never does (its documents are always `chord`), so nothing can collide. Its state document stays an ordinary document -of the run. Both change in the transactions that change them today (start, state change, outcome, finish, block, -drop); `finish` deletes the record and retires the state in its commit (Q9), so only unsettled runs are stored. At the -object's first activation the records are loaded through the engine's key listing and a load per run (no new read). +### 4.3 Deferred state (D8) -- **Resume at activation:** a run recorded `finishing` resumes `finish` only; a run recorded `running` was cut off by - a crash: `rerun` invokes `run` again with the stored args and state, `abandon` records `failed { aborted, reason: - "restart" }` and finishes. -- **Restart limit** (aligned with Durable's `restarts`): - -**The same rule as Durable (D6, owner, 2026-10-27: "yes, same rule for chord"):** at activation a run recorded -`running` was cut off by a crash: its `restarts` rises by one (stored), then it resumes (`rerun` invokes `run` again, -`abandon` finishes with `failed { aborted, reason: "restart" }`); a clean close (`host.close`, a native restart on -reload) writes its stopped runs back so the next activation counts nothing (Durable's `writeAtClose`); a -`ctx.state.change` resets the count (progress); at five in a row the run ends with a new `JobError` `{ kind: -"restart_limit", restarts }`, its `finish` runs, and `job_interrupted` is reported (Durable's `restart_limit`, -`task_interrupted`). No `blocked { restart_limit }`: supersedes design §15's parking, as Durable dropped parking in -#11139. - -- **Abort** (owner, 2026-10-27: async and stored in 6b): - -**`job.abort(id, ctx): Promise<"marked" | "terminal">` (D7, owner, 2026-10-27), Durable's words and behaviour:** -a running run gets its abort stored (a rerun after a crash then finishes `aborted` instead of running again) and its -signal fired, and `abort` resolves `"marked"` once `run` settled (as `abortTask` joins the run); a finishing run is only -marked (`"marked"`, as Durable's `completing`); a blocked run is dropped at once (`"marked"`, as Durable's orphan); an id -with no unsettled run resolves `"terminal"` (Chord keeps no settled runs, Q9, so a settled and an unknown id are the -same; Durable throws for an unknown one). - -- **Deferred state persistence:** - -**Per change, Durable's shape (D8, owner, 2026-10-27):** `ctx.state.change(fn, ctx, { persistence: { kind: -"deferred", withinMs? } })` (default 1,000 ms, as Durable's `CommitPersistence`): adopted and published at once, -stored with the next immediate change of the document, before the outcome is recorded (so `outcome.state` has it), at -close, or after `withinMs`; a crash loses what is not stored. Only run state documents; other persisted documents get -it when a real case needs it. - -- **Fingerprint:** the job's fingerprint (args, result and state schemas) stored with every record; a record of the - same version with another fingerprint is `blocked { version }` (slice 7's family; the carry-over's "a job - fingerprint over args, result and state"). -- **Trace carrier:** the start's carrier (design §29) stored with the record; used in slice 9. - -## 4. The applier for batch apply - -6a's applier applies an archive's writes; 6b extends it to every `StoreWrite` kind (`alter`, `dropIndex`, `update`, -`updateWhere`, `delete`, `deleteWhere` with their recorded counts and versions, `document.change`, `document.retire`), -checked against the store's records (a trust boundary), one code path with import. A test replays a store's outbox -into an empty store and compares the two stores' exports. Applying a batch with a different count than recorded stops -applying (S17). +As Durable's holds: the slot adopts and publishes the change at once (`valueVersion` ahead of what is stored), keeps +the held ops in a `PendingOps` (a base flag when `checkpointWhen` selected one), and the run's documents stay pinned; +the next immediate change of the state document stores the held ops and its own as one delta on the stored revision +(or a base); one foreground timer per object for the earliest deadline (`withinMs`, default 1,000 ms, `Infinity` waits +for the others; a negative or non-number `withinMs` is `invalid/args`); the outcome's commit and the close writer +store what is held; a held change of a fenced invocation is still stored; a crash loses it. A deferred change that +would reset `restarts` is stored at once (the reset must be durable). On an in-memory job the option is accepted and +does nothing. ## 5. Benchmarks and tests -- **The gate** (decisions "The commit-step benchmark", "The performance and memory bar"): Durable's own storage - benchmark (`test/storage.bench.ts`: the write and read cases of `testing/storage-benchmark.ts`) on `StoreStorage` - against `SqliteStorage`, and Durable's `test/conversation-view-bench.ts` on both: each hot path within about 5%, - memory (allocation per op, retained after GC) no worse beyond noise; persisted job starts and state changes against - in-memory ones. -- **Tests:** Durable's storage conformance suite on `StoreStorage` (memory `:memory:` and across reopen); Durable's - Session and harness test files that take a storage, run on `StoreStorage`; `self.follow` (one frame per commit, a - derive over two views, outside the store, early calls held, failure and end, an absent `observeDoc`); persisted jobs (rerun, abandon, a recorded outcome - resumes `finish`, restart counting and reset, the limit, abort stored, deferred state, fingerprints); the applier - (every kind, the outbox replay, refused counts); export and import with Durable records and `rewindable` history; - GC (observers detached with their attempt). -- **Examples:** `26-durable-on-the-store.ts`, `27-durable-views.ts`, `28-persisted-jobs.ts`. +- **The gate** ("The performance and memory bar"): Durable's storage benchmark (`test/storage.bench.ts`, every read and + write case, 1k and 10k) and `test/conversation-view-bench.ts` on `StoreStorage` against `SqliteStorage`, outbox off + (its cost reported separately): each hot path within about 5%, allocation per op and retained memory no worse beyond + noise; Chord's 6a commit gate again with the new indexes and the allocator; persisted job start and state change + reported against in-memory ones (a disk commit is not held to the in-memory number). +- **Tests:** the conformance suite (both fixtures); the D4 matrix (`BEGIN`, body, `COMMIT`, `ROLLBACK` failures; a + refused Durable commit keeps the store; a refused Chord commit fails its object; a `CHECK` failure only its + transaction; a refused migration chunk; a request error inside a multi-query read keeps the store); the allocator (a + mint during a running Chord step, saturation across reopen, a refused step's gap, replay restoring `nextId`); + create-plus-retire and change-plus-retire through the outbox and the applier; reads at a point on retired `latest` + and `rewindable` records; export and import of rewindable history and mixed Chord and Durable data; `applyBatch` + (every kind, the replay equality, a diverged count); `self.follow` (one frame per commit, a derive over two views, + calls outside a batch on the store, early calls, an absent `observeDoc`, `failed` and `ended`, a retirement before + the end); persisted jobs (every transition of 4.2, a crash between reservation and `run`, a crash during the close + writer, a seal and retry, restart counting and its reset, the limit, the finish failure count across a crash, abort + joined and stored, abort racing an outcome, deferred state merged, flushed at the outcome and at close, pinned past + the retention); GC (closed `StoreStorage`, ended follow sources, cancelled abort waiters, deferred timers, settled and + dropped runs). +- **Examples:** `26-durable-on-the-store.ts` (a worker with Durable's Session and a Chord store object on one store, + exporting the archive), `27-durable-views.ts`, `28-persisted-jobs.ts`. ## Carry-over taken -From `slices.md` "Slice 6b": persisted jobs (all of it), `StoreStorage` (ids, fork copies, `adopt` in the batch), -the Chord source over Durable's views (`self.follow`, the batch on and off the store), the applier for batch apply; `job.abort` async and -stored (decision of 2026-10-27). +From `slices.md` "Slice 6b": persisted jobs (all of it, the finish failure count included), `StoreStorage` (ids, fork +copies, `adopt` in the batch), the Chord source over Durable's views (`self.follow`), the applier for batch apply; +`job.abort` async and stored. ## Carry-over left -Slice 7: a job fingerprint mismatch's `blocked { version }` resolution on reload; runs of jobs the code no longer -declares (with D5's records). Slice 9: the trace carrier's use. +Slice 7: `blocked: "version"` resolved on reload; runs of jobs the code no longer declares; blocked runs retried on +reload. Slice 8: Durable's views from other objects (the harness in a registry entry). Slice 9: the trace carrier's +use.