@zakkster/lite-signal-decorators 0.2.0 → 0.3.0

package/CHANGELOG.md

@@ -4,6 +4,57 @@ All notable changes to `@zakkster/lite-signal-decorators` are documented here.
 The format follows Keep a Changelog; this project adheres to Semantic
 Versioning.
 
+## [0.3.0] - 2026-08-26
+
+The class-reactivity benchmark release. Zero runtime changes: the 11-export
+surface ships logic-identical to 0.2.0 (version constants aside); what this
+release adds is the proof. Stage gate, all measured at closeout: 171/171
+tests on both lanes; torture 13 scenarios / 13 sabotage controls all
+green-and-breakable; the full benchmark matrix 42/42 lanes `sink=ok`, exit 0.
+
+### Added
+
+- `bench/` -- a private, never-shipped benchmark sub-package: six admitted
+  engines (`lsd` decorators, `lsd-define` buildless, the `lite-raw-boxes`
+  honesty baseline, MobX 7, signal-utils + signal-polyfill, a hand-rolled
+  alien-signals class) x seven class-shaped scenarios (vm-write, fleet-read
+  and fleet-tick over 10k instances, cascade 64/16+aggregate, deep-vm 64-deep,
+  churn as the headline, retention as gates). Ported lite-signal rig: anti-DCE
+  sink with analytic expected-sum oracles (skipped work rejects the lane),
+  median-of-5 + min, GC-fenced runs with real delta/retained heap columns,
+  machine provenance stamps with resolved adapter versions, a sabotaged-adapter
+  self-test proving the harness can fail, and machine-gated effect liveness on
+  lifecycle lanes. Candidates `classy-solid` and `@reactively/decorate` probed
+  and excluded with recorded one-line blockers (solid's stock-Node SSR
+  resolution; legacy-only decorators). Reproducible from a clean clone:
+  `cd bench && npm install && npm run bench`.
+- `fleet-soak` -- the 13th torture scenario (soak group): 2000 VMs (16000 pool
+  nodes exactly) under sustained ticks with partial churn rotations for a
+  wall-clock budget; per-sample F-0, flat retained heap, gcGate maxMajor 0;
+  its sabotage control leaks one VM per rotation and is caught at the first
+  sample.
+- `decisions/0006-kill-criteria.md` -- the formal verdicts. Criterion 1 (the
+  2.0x line vs the hand-written instance-field baseline) CLEARED WITH MARGIN:
+  vm-write 0.94x, fleet-read 1.10x (parity, per 0003's tied-layout finding).
+  Criterion 2 (churn cleanliness) CLEARED: 0 major and 0 minor collections
+  over 4096 lifecycle cycles on every lite lane, pools at floor (asserted
+  in-lane), retained settled at-or-below baseline -- and the decorated churn
+  loop emits ~12.6x less transient garbage than the hand-rolled equivalent.
+  Every competitor number published under the same stamp (AD-6 honesty
+  tiering); the wide-shape cascade/deep-vm gap vs raw boxes is recorded as an
+  S4 investigation item, not hidden.
+
+### Changed
+
+- README: the measured-numbers section now cites the cross-framework matrix
+  and the 0006 verdicts alongside the package's own committed probes; version
+  references bumped.
+- CHANGELOG release links now point at the package `repository`
+  (github.com/PeshoVurtoleta/lite-signal-decorators), matching the new
+  `repository`/`homepage`/`funding` fields in package.json.
+- `llms.txt` scope note: 0.3.0 adds only dev-side proof; the export surface
+  is unchanged and frozen for 0.x.
+
 ## [0.2.0] - 2026-08-26
 
 The adversarial-torture release: the 0.2 surface is frozen for 0.x, proven by
@@ -54,9 +105,8 @@ gc major=0 minor=0`.
 
 ## [0.2.0-preview.1] - 2026-08-26
 
-The full runtime surface. Exports grow from 8 to 11. Never published: the
-preview publish was pending npm authentication when 0.2.0 superseded it; the
-changes below first shipped with 0.2.0.
+The full runtime surface. Exports grow from 8 to 11. Published under the
+`preview` dist-tag on the same day 0.2.0 went to `latest`.
 
 ### Added
 
@@ -152,6 +202,7 @@ Initial release -- the decorator core.
 - Torture skeleton (`@zakkster/lite-leak` + `@zakkster/lite-gc-profiler`):
   retention, conservation, lifecycle, and zero-GC lanes.
 
-[0.2.0]: https://github.com/zakkster/lite-signal-decorators/releases/tag/v0.2.0
-[0.2.0-preview.1]: https://github.com/zakkster/lite-signal-decorators/releases/tag/v0.2.0-preview.1
-[0.1.0]: https://github.com/zakkster/lite-signal-decorators/releases/tag/v0.1.0
+[0.3.0]: https://github.com/PeshoVurtoleta/lite-signal-decorators/releases/tag/v0.3.0
+[0.2.0]: https://github.com/PeshoVurtoleta/lite-signal-decorators/releases/tag/v0.2.0
+[0.2.0-preview.1]: https://github.com/PeshoVurtoleta/lite-signal-decorators/releases/tag/v0.2.0-preview.1
+[0.1.0]: https://github.com/PeshoVurtoleta/lite-signal-decorators/releases/tag/v0.1.0

package/README.md

@@ -245,7 +245,7 @@ Symbol keys work (`Reflect.ownKeys`). A spec key colliding with an own property
 | Export | Value |
 |---|---|
 | `ReactiveDisposedError` | `extends Error`; `name: "ReactiveDisposedError"`; fields `className`, `key`. Thrown on ANY touch of a disposed instance's surface. |
-| `VERSION` | `"0.2.0"` |
+| `VERSION` | `"0.3.0"` |
 
 ### The rejection matrix
 
@@ -348,7 +348,7 @@ The gates that hold it (run on every change, all green at 0.2.0):
 - Torture: **12/12 scenarios** (zero-GC read/write lanes at `maxMajor 0, maxPauseMs 4`; 4096-cycle leak gate at 0 live / 0 findings / 0 warnings; capacity atomicity at every overflow point; a 300-seed x 20k-op oracle with zero divergences) -- plus **12/12 sabotage controls** proving each gate can actually fail.
 - `churn-soak`: sustained construct/use/dispose for a wall-clock budget; pools at floor and retained heap flat at every sample.
 
-The cross-framework churn benchmark (vs MobX-style layers and signal-utils, with the same anti-DCE rig) is the 0.3.0 milestone; the numbers above are the package's own committed probes.
+The cross-framework matrix lives in `bench/` (private, never shipped): six engines -- both our tiers, the hand-written `lite-raw-boxes` baseline, MobX 7, signal-utils/signal-polyfill, and a hand-rolled alien-signals class -- across seven class-shaped scenarios, checksum-verified for identical work, stamped into `bench/results.txt`. The formal verdicts are in [`decisions/0006-kill-criteria.md`](decisions/0006-kill-criteria.md): the decorated path measured **0.94x** the hand-written baseline on vm-write and **1.10x** on a 10k-instance fleet read (the 2.0x kill line cleared with margin), and **0 major + 0 minor GC over 4096 construct/use/dispose cycles** with pools at floor -- while emitting ~12.6x less transient garbage per churn run than the hand-rolled class it replaces.
 
 </details>
 
@@ -473,7 +473,7 @@ Because an instance whose base-class boxes live in one pool and whose subclass b
 No, and the README says so with numbers: 22.12 ns/op vs 15.25 raw vs 11.67 unbatched on the reference rig -- a thunk + rest-array per call. Use it for actions; keep per-frame writes on plain accessors.
 
 **Where are `costOf`, labels, the audit hook, private members?**
-Not in 0.2.0 -- the runtime surface is complete and frozen for 0.x, and the introspection layer is roadmapped behind the 0.3.0 benchmark. The `llms.txt` scope note tracks exactly what is and isn't included.
+Not in 0.3.0 -- the runtime surface froze for 0.x at 0.2.0, and the introspection layer (capacity accounting, devtools identity) is the 0.4.0 milestone. The `llms.txt` scope note tracks exactly what is and isn't included.
 
 ---
 

package/SignalDecorators.d.ts

@@ -273,4 +273,4 @@ export class ReactiveDisposedError extends Error {
 // --- Version ------------------------------------------------------------------
 
 /** Package version. Kept in lockstep with package.json and llms.txt. */
-export const VERSION: "0.2.0";
+export const VERSION: "0.3.0";

package/SignalDecorators.js

@@ -1,5 +1,5 @@
 /**
- * @zakkster/lite-signal-decorators v0.2.0
+ * @zakkster/lite-signal-decorators v0.3.0
  * --------------------
  * Stage-3 decorator layer over @zakkster/lite-signal. Turns a plain class into
  * a reactive view-model with measured per-instance cost and deterministic
@@ -1298,4 +1298,4 @@ export function rootOf(vm) {
 // --- Version ------------------------------------------------------------------
 
 /** Package version. Kept in lockstep with package.json and llms.txt. */
-export const VERSION = "0.2.0";
+export const VERSION = "0.3.0";

package/llms.txt

@@ -1,6 +1,6 @@
 # @zakkster/lite-signal-decorators
 
-VERSION 0.2.0
+VERSION 0.3.0
 
 > Stage-3 decorator layer over @zakkster/lite-signal. Turns a plain class into a
 > reactive view-model where each instance has a measured per-property cost, a
@@ -55,7 +55,7 @@ slot for a poison handle, so any later read/write throws a named
   `forEachOwned` / lite-devtools. Throws `ReactiveDisposedError` after dispose.
 - `ReactiveDisposedError` -- `extends Error`, `name` `"ReactiveDisposedError"`,
   fields `className` and `key`.
-- `VERSION` -- `"0.2.0"`.
+- `VERSION` -- `"0.3.0"`.
 
 ## Registry law (one registry per host chain)
 
@@ -137,7 +137,8 @@ routed per-registry via `createRegistry()`).
 
 ## Scope note
 
-0.2.0 ships the full runtime surface: the decorator core plus
-`@reactiveEffect`, `@batched`, `@reactiveHost({ registry })`, and
-`defineReactive`. Not yet included: cost introspection (`costOf`/`capacityFor`),
-labels, an audit hook, and private-member support.
+0.3.0 = the 0.2.0 runtime surface (unchanged -- the same 11 exports, frozen for
+0.x) plus the dev-side class-reactivity benchmark (`bench/`, private, never
+shipped) and the fleet-soak torture scenario; the formal kill-criteria verdicts
+live in decisions/0006-kill-criteria.md. Not yet included: cost introspection
+(`costOf`/`capacityFor`), labels, an audit hook, and private-member support.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@zakkster/lite-signal-decorators",
-  "version": "0.2.0",
+  "version": "0.3.0",
   "description": "Stage-3 decorator layer over @zakkster/lite-signal. The reactive class layer where an instance has a measured cost, deterministic teardown, and a churn benchmark.",
   "author": "Zahary Shinikchiev <shinikchiev@yahoo.com>",
   "license": "MIT",