@pls-compile-owo/fisheye 0.2.1 → 0.4.2

package/lib/util.mjs

@@ -0,0 +1,78 @@
+// 通用工具：时间格式化、参数解析、文件头探测
+import fs from 'node:fs'
+import { t } from './i18n.mjs'
+
+// 时长格式化：超过 1 小时显示 01:05:30，否则 MM:SS
+
+export const fmt = (s) => {
+  const h = Math.floor(s / 3600)
+  const m = Math.floor((s % 3600) / 60)
+  const sec = Math.floor(s % 60)
+  const mm = String(m).padStart(2, '0')
+  const ss = String(sec).padStart(2, '0')
+  return h ? `${h}:${mm}:${ss}` : `${mm}:${ss}`
+}
+// Windows 文件名不能带冒号，落盘用秒数
+
+export const fsec = (s) => String(Math.round(s)).padStart(4, '0')
+
+// 从 JPEG 头读尺寸（沙箱禁管道，不能用 ffprobe 捕获输出）
+
+export function jpegSize(file) {
+  try {
+    const b = fs.readFileSync(file)
+    let i = 2
+    while (i < b.length - 8) {
+      if (b[i] !== 0xFF) { i++; continue }
+      const marker = b[i + 1]
+      const len = (b[i + 2] << 8) | b[i + 3]
+      if (marker >= 0xC0 && marker <= 0xCF && marker !== 0xC4 && marker !== 0xC8 && marker !== 0xCC) {
+        return { h: (b[i + 5] << 8) | b[i + 6], w: (b[i + 7] << 8) | b[i + 8] }
+      }
+      i += 2 + len
+    }
+  } catch {}
+  return null
+}
+
+// 参数校验（在 CLI 分发处用；Number('')===0、Infinity、NaN 都要挡掉）
+
+export const fail = (msg) => {
+  console.error(`${t('err.argPrefix')}${msg}`)
+  process.exit(2)
+}
+
+export const num = (v, name, { min, max } = {}) => {
+  const n = Number(v)
+  if (v === '' || !Number.isFinite(n)) fail(t('err.num.notNumber', { name, value: JSON.stringify(v) }))
+  if (min !== undefined && n < min) fail(t('err.num.min', { name, min, n }))
+  if (max !== undefined && n > max) fail(t('err.num.max', { name, max, n }))
+  return n
+}
+
+export const gridStr = (v) => {
+  const s = String(v ?? '3x3').toLowerCase()
+  if (!/^[1-9]x[1-9]$/.test(s)) fail(t('err.grid.format', { value: JSON.stringify(v) }))
+  return s
+}
+// 时间参数：支持 330（秒）/ 5:30 / 1:05:30
+
+export const parseTime = (v, name) => {
+  if (v === null || v === undefined || v === '') return null
+  const s = String(v).trim()
+  if (/^\d+(\.\d+)?$/.test(s)) return Number(s)
+  const parts = s.split(':')
+  if (parts.length < 2 || parts.length > 3) fail(t('err.time.format', { name, value: JSON.stringify(v) }))
+  const nums = parts.map(Number)
+  if (nums.some((n) => !Number.isFinite(n) || n < 0)) fail(t('err.time.invalid', { name, value: JSON.stringify(v) }))
+  return parts.length === 2 ? nums[0] * 60 + nums[1] : nums[0] * 3600 + nums[1] * 60 + nums[2]
+}
+
+export const pageNum = (v) => {
+  if (v === null || v === undefined || v === '') return 1
+  const n = Number(v)
+  if (!Number.isInteger(n) || n < 1 || n > 200) fail(t('err.page.range', { value: JSON.stringify(v) }))
+  return n
+}
+
+// 统一取视频元信息（带超时 + code 校验；之前三处各自 fetch，有的没查 code）

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@pls-compile-owo/fisheye",
-  "version": "0.2.1",
+  "version": "0.4.2",
   "description": "Token-budget-aware video understanding for AI agents: official subtitles -> burned-in subtitle band -> danmaku highlights -> adaptive keyframe grids -> targeted refine.",
   "type": "module",
   "bin": {
@@ -8,11 +8,15 @@
   },
   "files": [
     "fisheye.mjs",
+    "lib/",
     "danmaku-filter.mjs",
     "analyze_subs.py",
     "test-danmaku-filter.js",
+    "test/",
     "credentials.example.json",
     "README.md",
+    "README.en.md",
+    "CHANGELOG.md",
     "LICENSE",
     "docs/"
   ],
@@ -20,7 +24,7 @@
     "node": ">=18"
   },
   "scripts": {
-    "test": "node test-danmaku-filter.js",
+    "test": "node --test test/ && node test-danmaku-filter.js",
     "selftest": "node fisheye.mjs selftest --deep"
   },
   "keywords": [

package/test/danmaku.test.mjs

@@ -0,0 +1,70 @@
+// danmaku 模块单测：广告段检测 / 热区摘要 / 过滤规则
+import test from 'node:test'
+import assert from 'node:assert/strict'
+import { AD_SEGMENT_SIGNAL, detectAdSegments, summarizeDanmaku } from '../lib/danmaku.mjs'
+import { spamReason, filterDanmaku } from '../danmaku-filter.mjs'
+
+test('AD_SEGMENT_SIGNAL：命中广告关键词', () => {
+  assert.ok(AD_SEGMENT_SIGNAL.test('感谢金主爸爸'))
+  assert.ok(AD_SEGMENT_SIGNAL.test('本期恰饭'))
+  assert.ok(AD_SEGMENT_SIGNAL.test('优惠码在评论区'))
+  assert.ok(!AD_SEGMENT_SIGNAL.test('这个视频真好看'))
+})
+
+test('detectAdSegments：空 / 短视频返回空数组', () => {
+  assert.deepEqual(detectAdSegments([], 300), [])
+  assert.deepEqual(detectAdSegments([{ t: 1, text: 'x' }], 60), [])
+})
+
+test('detectAdSegments：密度低谷 + 广告词 → 识别出段', () => {
+  // 300 秒视频：中段 100~160s 弹幕稀少，其余密集
+  const danmaku = []
+  for (let t = 0; t < 300; t += 2) {
+    if (t >= 100 && t < 160) continue
+    danmaku.push({ t, text: '666' })
+  }
+  danmaku.push({ t: 120, text: '恰饭时间' })
+  const segs = detectAdSegments(danmaku, 300)
+  assert.ok(Array.isArray(segs))
+  assert.ok(segs.length >= 1)
+  for (const s of segs) {
+    assert.ok(s.to > s.from)
+    assert.ok(s.ts.includes('-'))
+    assert.ok(['high', 'low'].includes(s.confidence))
+  }
+})
+
+test('summarizeDanmaku：空输入返回 null', () => {
+  assert.equal(summarizeDanmaku([]), null)
+})
+
+test('summarizeDanmaku：按时间桶聚合，返回时间与样本', () => {
+  const list = [
+    { t: 1, text: '666' },
+    { t: 2, text: '666' },
+    { t: 3, text: '哈哈' },
+    { t: 11, text: '好' }
+  ]
+  const out = summarizeDanmaku(list, 5, 6)
+  assert.ok(Array.isArray(out))
+  assert.equal(out.length, 2)
+  assert.equal(out[0].start, 0)
+  assert.equal(out[0].count, 3)
+  assert.ok(out[0].samples.includes('666×2'))
+  assert.equal(out[1].start, 10)
+})
+
+test('danmaku-filter：只过滤明确引流，不碰梗', () => {
+  assert.ok(spamReason('加微信领取福利'))
+  assert.ok(spamReason('私信我'))
+  assert.ok(!spamReason('666'))
+  assert.ok(!spamReason('awsl'))
+  assert.ok(!spamReason('哈哈哈哈'))
+  const { kept, dropped } = filterDanmaku([
+    { t: 1, text: '666' },
+    { t: 2, text: '加微信' },
+    { t: 3, text: '前面高能' }
+  ])
+  assert.equal(kept.length, 2)
+  assert.ok(dropped.ad >= 1)
+})

package/test/util.test.mjs

@@ -0,0 +1,71 @@
+// util 模块单测：时间格式化 / 参数解析（非法输入走 fail() 会 exit，用子进程断言退出码）
+import test from 'node:test'
+import assert from 'node:assert/strict'
+import { spawnSync } from 'node:child_process'
+import { fileURLToPath, pathToFileURL } from 'node:url'
+import path from 'node:path'
+import { fmt, fsec, parseTime, num, gridStr, pageNum } from '../lib/util.mjs'
+
+const utilPath = fileURLToPath(new URL('../lib/util.mjs', import.meta.url))
+
+// 在子进程里执行一段 import util 的表达式，返回退出码（fail() 会 exit(2)）
+function runInChild(fn, argExpr) {
+  const code = `import(${JSON.stringify(pathToFileURL(utilPath).href)}).then(({ ${fn} }) => { ${fn}(${argExpr}) })`
+  const r = spawnSync(process.execPath, ['--input-type=module', '-e', code], { encoding: 'utf8' })
+  return r.status
+}
+
+test('fmt：超过 1 小时显示 H:MM:SS，否则 MM:SS', () => {
+  assert.equal(fmt(0), '00:00')
+  assert.equal(fmt(65), '01:05')
+  assert.equal(fmt(600), '10:00')
+  assert.equal(fmt(3930), '1:05:30')
+})
+
+test('fsec：落盘文件名用 4 位秒数', () => {
+  assert.equal(fsec(0), '0000')
+  assert.equal(fsec(59.6), '0060')
+  assert.equal(fsec(3600), '3600')
+})
+
+test('parseTime：数字 / M:SS / H:MM:SS', () => {
+  assert.equal(parseTime(90, 't'), 90)
+  assert.equal(parseTime('90', 't'), 90)
+  assert.equal(parseTime('1:30', 't'), 90)
+  assert.equal(parseTime('1:05:30', 't'), 3930)
+  assert.equal(parseTime(null, 't'), null)
+  assert.equal(parseTime('', 't'), null)
+})
+
+test('parseTime：非法格式退出码为 2', () => {
+  assert.equal(runInChild('parseTime', "'abc', 't'"), 2)
+  assert.equal(runInChild('parseTime', "'1:2:3:4', 't'"), 2)
+})
+
+test('num：数字与边界', () => {
+  assert.equal(num('5', 'x'), 5)
+  assert.equal(num(5, 'x'), 5)
+  assert.equal(num(0, 'x'), 0)
+})
+
+test('num：非法 / 越界退出码为 2', () => {
+  assert.equal(runInChild('num', "'abc', 'x'"), 2)
+  assert.equal(runInChild('num', "'', 'x'"), 2)
+  assert.equal(runInChild('num', "999, 'x', { min: 0, max: 10 }"), 2)
+})
+
+test('gridStr：归一化与校验', () => {
+  assert.equal(gridStr('3x3'), '3x3')
+  assert.equal(gridStr('4X4'), '4x4')
+  assert.equal(gridStr(undefined), '3x3')
+  assert.equal(runInChild('gridStr', "'10x10'"), 2)
+  assert.equal(runInChild('gridStr', "'abc'"), 2)
+})
+
+test('pageNum：分P序号', () => {
+  assert.equal(pageNum(null), 1)
+  assert.equal(pageNum('2'), 2)
+  assert.equal(runInChild('pageNum', '0'), 2)
+  assert.equal(runInChild('pageNum', '201'), 2)
+  assert.equal(runInChild('pageNum', "'x'"), 2)
+})