docs(hook-bodies): 写集一节改为准确的覆盖面描述 —— #4305/#4344 之后"没有静态检查"已不成立 (#4351)

.changeset/hook-bodies-write-set-docs.md

@@ -0,0 +1,4 @@
+---
+---
+
+docs(hook-bodies): the write-set section said the write side had no static checking, which #4305/#4344 made false (#4351) — rewritten as an accurate coverage description (three literal patterns, advisory `hook-body-write-unknown-field` / `action-body-write-unknown-field`, and the shapes the parser must skip, so a missing warning is not read as proof of correctness), the "accepted gap with no planned closure" conclusion retracted (#3700 stays accurate history; the gap was closed from the other end, by parsing the write set), the runtime paragraph corrected against measured behaviour, and the three now-stale cross-links in `hooks.mdx` updated to the renamed anchor. Releases nothing.

content/docs/automation/hook-bodies.mdx

@@ -116,20 +116,49 @@ The CLI builder **rejects** any source that uses:
 
 Need outbound HTTP? Define a **Connector recipe** as metadata and call it via `ctx.connector(...)`. (Connector spec is tracked separately and ships after L1+L2 stabilises.)
 
-### Not statically checked: the write set
+### Write-set checking
 
 Static validation around a hook is asymmetric, and it is worth knowing exactly where the line is:
 
 - **Checked — read side.** `hook.condition` is validated at build time against the target object's fields by the expression validator (`@objectstack/lint`), including array-valued `hook.object` targets. A condition referencing a nonexistent field fails the lint.
 - **Checked — capability side.** `body.capabilities` gates which `ctx` APIs the body may call at all; the sandbox throws on an undeclared call.
-- **Not checked — write side.** Nothing validates *which fields* your body writes. `body.source` is an opaque JS string: the fields it assigns (`ctx.input.foo = …`) or writes cross-object (`ctx.api.object('x').update(id, { foo })`) are not represented anywhere in metadata, so no lint can verify they exist. This is an **accepted gap** with no planned closure — a structured `writes` declaration was considered and dropped ([#3700](https://github.com/objectstack-ai/objectstack/issues/3700), closed as not planned).
+- **Checked — write side, advisory and literal-only.** Since [#4271](https://github.com/objectstack-ai/objectstack/issues/4271), `body.source` is **parsed** (never executed, never type-checked) and the field names it writes are resolved against the target object's declarations. An unknown field raises `hook-body-write-unknown-field` — a **warning** carrying a did-you-mean suggestion, which never blocks a build. Action bodies get the same check on their `ctx.api` writes (`action-body-write-unknown-field`). Both run under `os validate`, `os lint` and `os compile`.
 
-Writing a field the target object doesn't have is not caught at runtime either — the write-path validator skips unknown keys. On a SQL driver the stray column reaches the database and the whole operation fails with a driver-level error, far from the authoring mistake; a schemaless driver (memory, MongoDB) silently persists the stray key.
+Three literal write shapes are recognized, and only these:
 
-Because the write set carries no static checking:
+| Write shape | Hook body | Action body |
+|---|---|---|
+| `ctx.input.<field> = …` / `ctx.input['<field>'] ⟨op⟩= …` (including `+=`, `??=`, …) | checked | not checked — an action's `ctx.input` is its **params bag**, not a record |
+| `Object.assign(ctx.input, { <field>: … })` | checked | not checked — same surface |
+| `ctx.api.object('<literal>').insert\|create\|update({ <field>: … })`, `.updateById(id, { <field>: … })` | checked | checked |
+
+**A missing warning is not a clean bill of health.** The rule bails *silently* on everything it cannot resolve statically, deliberately preferring a missed finding to a false one — a false positive kills an advisory lint, while a miss just leaves the gap open a little longer:
+
+- computed keys (`ctx.input[k] = …`), spreads, and non-literal payloads;
+- dynamic object names (`ctx.api.object(name)`);
+- `ctx.input` writes in a wildcard (`object: '*'`) hook — there is no single target to resolve against;
+- multi-target hooks where the field exists on *some* target: a body may legitimately branch per object, so only a field missing on **every** named target is flagged;
+- objects declared by another package;
+- aliased input (`const doc = ctx.input; doc.x = 1`) — v1 does no data-flow analysis.
+
+System/audit columns and the flat-input envelope keys (`id`, `options`, `ast`, `data`) are never flagged.
+
+A structured `writes` declaration was considered and dropped ([#3700](https://github.com/objectstack-ai/objectstack/issues/3700), closed as not planned) — but the gap it left was closed from the other end, by parsing the write set out of the source. The practical consequence of that route: coverage is bounded by what a parser can see, not by what an author remembered to declare.
+
+#### What still happens at runtime
+
+An unknown field is **not** caught at runtime, and it does not fail quietly either. The write-path validator walks the object's *declared* fields, so an undeclared key is neither rejected nor stripped, and the sandbox's mutations are copied back onto the payload verbatim. What happens next is the driver's call:
+
+- **SQL drivers** put the stray column into the statement, so the **whole write fails** with a driver-level error (`table deal has no column named stagee`) — nothing is stored, and the error surfaces far from the authoring mistake.
+- **Schemaless drivers** (memory, MongoDB) silently persist the stray key alongside the real ones.
+
+Neither outcome is the one you wanted, and the advisory warning is the earliest signal you get.
+
+Because the checking is advisory and literal-only:
 
-- **Check your write targets by hand.** Every field the body assigns must exist on the target object(s) — for an array or `"*"` hook, on every target.
-- **Prefer a flow `update_record` node when the write set is fixed.** Flow nodes declare their writes structurally in `fields`, so they get the static checking hook bodies can't.
+- **Treat `hook-body-write-unknown-field` as a build failure by convention.** It does not gate, but the rule is tuned for near-zero false positives — in practice a warning is a real typo.
+- **Check by hand what the parser cannot see.** Computed keys, spreads, aliased input and dynamic object names are invisible to the rule; for an array or `"*"` hook, every field must exist on every target.
+- **Prefer a flow `update_record` node when the write set is fixed — but not for *this* check.** A flow node's writes are structured config: they diff field-by-field, render in the Console designer, and a write to a `readonly:true` field is a **gating error** (`flow-update-readonly-field`) that hooks have no counterpart for. What an `update_record` node does *not* get today is a field-existence check — writing a field the object never declares is currently reported by nothing at all. On that one axis an L2 body is now the better-checked surface.
 - **Exercise the hook against a real object before shipping** — on SQL drivers the mistake surfaces on the first write; schemaless drivers won't tell you.
 
 ### Signature conventions

content/docs/automation/hooks.mdx

@@ -33,12 +33,15 @@ cannot express.
 
 Two structural reasons to prefer the flow when either could work:
 
-- **A flow's writes are validatable metadata; a hook body is opaque code.** An
+- **A flow's writes are structured metadata; a hook body is code.** An
   `update_record` node's `fields` is structural config that `os validate`
   checks — readonly targets, template dialects, declared expression slots. A
-  hook body's write set is **not** statically checked (see
-  [Hook & Action Bodies](/docs/automation/hook-bodies#not-statically-checked-the-write-set)):
-  a misspelled field name runs green and writes nothing.
+  hook body's write set is checked too, but only for the literal patterns a
+  parser can recognise and only as an advisory warning (see
+  [Hook & Action Bodies](/docs/automation/hook-bodies#write-set-checking)).
+  Note the one axis where this reverses: writing a field the target object never
+  declares warns in a hook body, and is reported nowhere in an `update_record`
+  node.
 - **A flow reviews as data.** The node graph diffs field-by-field and renders
   in the Console designer with per-node run history; a hook body reviews as
   code only.
@@ -84,9 +87,10 @@ a hook can be, so review it as such:
   belongs behind a `condition` (evaluated before the handler) rather than an
   early `return` inside the body.
 - **Reads and writes both widen.** A wildcard hook holding an L2 body with
-  `api.write` can write anywhere. Whether the fields it writes exist is
-  [not statically checked](/docs/automation/hook-bodies#not-statically-checked-the-write-set),
-  so a wildcard write set is the least verifiable shape in the system.
+  `api.write` can write anywhere. Its `ctx.input` writes are also the one shape
+  the [write-set lint](/docs/automation/hook-bodies#write-set-checking) has to
+  skip — with no single target object there is nothing to resolve field names
+  against — so a wildcard write set is the least verifiable shape in the system.
 
 Where a wildcard is the honest answer, say so in `description` — it is the one
 place a reviewer can find out why the broad target was chosen.
@@ -179,8 +183,9 @@ ctx = {
 - Use `ctx.api.object('x')` for cross-object access so writes stay in scope
 - Handle errors gracefully
 - Verify every field your hook writes exists on the target object(s) — the
-  write set of a hook body is **not statically validated**, unlike `condition`
-  (see [Hook & Action Bodies](/docs/automation/hook-bodies#not-statically-checked-the-write-set))
+  write-set lint catches the literal cases, but it is advisory, and computed
+  keys, spreads and aliased input are invisible to it (see
+  [Hook & Action Bodies](/docs/automation/hook-bodies#write-set-checking))
 
 ❌ **DON'T:**
 - Query in loops