Upgrade Changesets to v3 and the benchmark harness to tinybench 6

.changeset/changesets-3-and-tinybench-6.md

@@ -0,0 +1,3 @@
+---
+
+---

.changeset/config.json

@@ -1,5 +1,5 @@
 {
-  "$schema": "https://unpkg.com/@changesets/config@3.1.2/schema.json",
+  "$schema": "https://unpkg.com/@changesets/config@4.0.1/schema.json",
   "changelog": "@changesets/cli/changelog",
   "commit": false,
   "fixed": [

.github/workflows/release.yml

@@ -36,12 +36,11 @@ jobs:
       - run: pnpm install --frozen-lockfile
       - run: pnpm check
       - id: changesets
-        uses: changesets/action@v1
+        uses: changesets/action@v2.1.2
         with:
-          version: pnpm version-packages
-          publish: pnpm release
-        env:
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          version-script: pnpm version-packages
+          publish-script: pnpm release
+          github-token: ${{ secrets.GITHUB_TOKEN }}
 
   deploy-website:
     needs: release

package.json

@@ -36,7 +36,7 @@
     "release": "pnpm build && changeset publish"
   },
   "devDependencies": {
-    "@changesets/cli": "2.31.0",
+    "@changesets/cli": "3.0.3",
     "@effect/vitest": "4.1.1",
     "@typeonce/oxlint-plugin-effect-machine": "workspace:*",
     "@types/node": "26.6.6",
@@ -45,7 +45,7 @@
     "oxlint": "1.90.0",
     "oxlint-tsgolint": "7.0.2005",
     "pagefind": "1.5.2",
-    "tinybench": "2.9.0",
+    "tinybench": "6.2.2",
     "tstyche": "7.2.5",
     "typedoc": "0.28.20",
     "typescript": "6.0.3",

packages/effect-machine/package.json

@@ -76,7 +76,7 @@
     "@effect/vitest": "4.1.1",
     "@types/node": "26.6.6",
     "effect": "4.1.1",
-    "tinybench": "2.9.0",
+    "tinybench": "6.2.2",
     "tstyche": "7.2.5",
     "typescript": "6.0.3",
     "vitest": "5.0.5"

scripts/runtime-performance.mjs

@@ -83,6 +83,7 @@ const bench = new Bench({
   warmupTime: configuration.warmupTimeMilliseconds,
   iterations: configuration.minimumIterations,
   warmupIterations: configuration.minimumWarmupIterations,
+  retainSamples: true,
   throws: true
 })
 const benchmarkDefinitions = new Map()
@@ -334,7 +335,7 @@ for (const implementation of runtimeReferenceImplementations) {
 }
 
 try {
-  await bench.warmup()
+  // Bench.run warms every task up before measuring it.
   await bench.run()
 } finally {
   for (const implementation of implementations) {
@@ -373,21 +374,26 @@ const benchmarks = bench.tasks.map((task) => {
   if (definition === undefined || result === undefined) {
     throw new Error(`Missing benchmark result for ${task.name}`)
   }
-  if (result.error !== undefined) {
+  if (result.state === "errored") {
     throw result.error
   }
+  if (result.state !== "completed" || result.latency.samples === undefined) {
+    throw new Error(`Benchmark ${task.name} did not complete: ${result.state}`)
+  }
   const operations = definition.operationsPerIteration
-  const medianBatchMilliseconds = percentile(result.samples, 0.5)
+  const { latency } = result
+  const samples = latency.samples
+  const medianBatchMilliseconds = percentile(samples, 0.5)
   return {
     ...definition,
     medianThroughput: operations * 1_000 / medianBatchMilliseconds,
-    meanThroughput: result.hz * operations,
-    meanNanosecondsPerOperation: result.mean * 1_000_000 / operations,
+    meanThroughput: operations * 1_000 / latency.mean,
+    meanNanosecondsPerOperation: latency.mean * 1_000_000 / operations,
     medianBatchMilliseconds,
-    p95BatchMilliseconds: percentile(result.samples, 0.95),
-    p99BatchMilliseconds: result.p99,
-    relativeMarginOfError: result.rme,
-    sampleCount: result.samples.length
+    p95BatchMilliseconds: percentile(samples, 0.95),
+    p99BatchMilliseconds: percentile(samples, 0.99),
+    relativeMarginOfError: latency.rme,
+    sampleCount: samples.length
   }
 })