@vzn/vx-ci 0.0.594 → 0.0.595

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vzn/vx-ci",
-  "version": "0.0.594",
+  "version": "0.0.595",
   "license": "MIT",
   "description": "GitHub Actions integration for vx — a telemetry plugin that writes the run as a GitHub Actions job summary.",
   "keywords": [
@@ -33,7 +33,7 @@
     "access": "public"
   },
   "peerDependencies": {
-    "@vzn/vx": "^0.0.594"
+    "@vzn/vx": "^0.0.595"
   },
   "repository": {
     "type": "git",

package/src/checks.ts

@@ -3,6 +3,7 @@
 // `GITHUB_TOKEN` with `checks: write`; the plugin declines the check (not
 // the whole sink) without one, so the summary still works token-less.
 import type { RunSummaryRecord } from '@vzn/vx'
+import { truncateUtf8 } from './summary.js'
 
 export type FetchFn = (
   url: string,
@@ -67,14 +68,11 @@ function isCertificateRefusal(err: unknown): boolean {
  * summary's does.
  */
 export function clampSummary(markdown: string): string {
-  const MAX = 65_535
-  if (markdown.length * 3 <= MAX) return markdown
-  const bytes = new TextEncoder().encode(markdown)
-  if (bytes.byteLength <= MAX) return markdown
-  const suffix = '\n\n…truncated by @vzn/vx-ci (65535-char Checks API limit)'
-  let end = MAX - new TextEncoder().encode(suffix).byteLength
-  while (end > 0 && (bytes[end]! & 0xc0) === 0x80) end--
-  return new TextDecoder().decode(bytes.subarray(0, end)) + suffix
+  return truncateUtf8(
+    markdown,
+    65_535,
+    '\n\n…truncated by @vzn/vx-ci (65535-char Checks API limit)',
+  )
 }
 
 export function buildCheckRunPayload(args: {

package/src/summary.ts

@@ -60,16 +60,24 @@ function hitSplit(s: RunSummaryRecord): string {
 export const MAX_JOB_SUMMARY_BYTES = 1024 * 1024
 
 export function clampJobSummary(markdown: string, room = MAX_JOB_SUMMARY_BYTES): string {
+  return truncateUtf8(
+    markdown,
+    room,
+    '\n\n…truncated by @vzn/vx-ci (GitHub caps a job summary at 1 MiB)\n',
+  )
+}
+
+/** Cuts `text` to `maxBytes` of UTF-8, `suffix` included, never splitting a character. */
+export function truncateUtf8(text: string, maxBytes: number, suffix: string): string {
   // The cap is BYTES, and the page is not ASCII: every status label is an
   // emoji (4 bytes, 2 UTF-16 units) and every separator a `·` or a `—`
   // (2 and 3 bytes, 1 unit). Measured in `.length`, a clamped page of those
   // rows came out past 1 MiB and GitHub refused it whole (item 806). A unit
   // is at most 3 bytes, so a short page skips the encode.
-  if (markdown.length * 3 <= room) return markdown
-  const bytes = new TextEncoder().encode(markdown)
-  if (bytes.byteLength <= room) return markdown
-  const suffix = '\n\n…truncated by @vzn/vx-ci (GitHub caps a job summary at 1 MiB)\n'
-  let end = room - new TextEncoder().encode(suffix).byteLength
+  if (text.length * 3 <= maxBytes) return text
+  const bytes = new TextEncoder().encode(text)
+  if (bytes.byteLength <= maxBytes) return text
+  let end = maxBytes - new TextEncoder().encode(suffix).byteLength
   if (end <= 0) return ''
   // Back up off a continuation byte, so the cut never splits a character.
   while (end > 0 && (bytes[end]! & 0xc0) === 0x80) end--