// Checks existing commits against the CONTRIBUTING.md spec. The spec allows one
// line only, so a subject must both parse (emoji + type + Chinese description)
// and be the whole message: any body or footer is reported too.
//
// This is a reporting tool, not a hook: it never blocks a commit.
//
// Usage:
//   node scripts/check-commit-log.mjs          # newest 10 commits
//   node scripts/check-commit-log.mjs 50       # newest 50 commits
//   node scripts/check-commit-log.mjs all      # every commit
//   DSD_SPEC=path/to/CONTRIBUTING.md node scripts/check-commit-log.mjs
import { execFileSync } from 'node:child_process'
import { readFileSync } from 'node:fs'
import { dirname, resolve } from 'node:path'
import { fileURLToPath } from 'node:url'

const repoRoot = resolve(dirname(fileURLToPath(import.meta.url)), '..')
const specPath = process.env.DN_SPEC ?? resolve(repoRoot, 'CONTRIBUTING.md')
const doc = readFileSync(specPath, 'utf8')
const count = process.argv[2] ?? '10'

/** The regex the spec documents for a subject line. */
const SUBJECT = new RegExp(
  '^(?:\\p{Extended_Pictographic}\\uFE0F?\\s+)?' +
  '(?<type>[a-z]+)' +
  '(?:\\((?<scope>[^)]*)\\))?' +
  '(?<breaking>!)?' +
  ':\\s(?<desc>.+)$',
  'u',
)

// The spec's table is the single source of truth for the type → emoji mapping,
// so a table edit cannot drift from what this script checks.
const emojiOfType = new Map()
for (const row of doc.split('\n').filter((line) => /^\| `[a-z]+` \| /.test(line))) {
  const cells = row.split('|').map((cell) => cell.trim())
  emojiOfType.set(cells[1].replaceAll('`', ''), cells[2])
}
if (emojiOfType.size === 0) throw new Error(`${specPath} has no type table to read`)

/** Strip the variation selector, so '♻' and '♻️' compare equal. */
const base = (text) => text.replace(/\uFE0F/g, '')
/** Display width, counting anything above U+2000 as double. */
const widthOf = (text) => [...text].reduce((n, c) => n + (c.codePointAt(0) > 0x2000 ? 2 : 1), 0)

// %B is the raw message body, so a record can hold several lines: a record ends
// at the next hash that starts a line, which is what this splits on.
const range = count === 'all' ? [] : [`-${Number(count)}`]
const raw = execFileSync('git', ['log', ...range, '--format=%h%x09%B%x00'], {
  encoding: 'utf8',
  cwd: repoRoot,
})

const commits = raw.split('\0').map((chunk) => chunk.replace(/^\n+/, '')).filter((chunk) => chunk.trim() !== '')

let problems = 0
for (const chunk of commits) {
  const tab = chunk.indexOf('\t')
  const hash = chunk.slice(0, tab)
  const message = chunk.slice(tab + 1).replace(/\n+$/, '')
  const lines = message.split('\n')
  const subject = lines[0]
  const extra = lines.slice(1).filter((line) => line.trim() !== '')
  const notes = []

  const match = SUBJECT.exec(subject)
  if (match === null) {
    notes.push('首行缺少类型前缀，或整体格式不是 <emoji> <type>[(scope)][!]: <描述>')
  } else {
    const expected = emojiOfType.get(match.groups.type)
    const lead = [...subject][0]
    if (expected === undefined) notes.push(`类型不在规范表里：${match.groups.type}`)
    else if (base(lead) !== base(expected)) notes.push(`类型 ${match.groups.type} 的 emoji 应为 ${expected}，实际是 ${lead}`)
    const width = widthOf(subject)
    if (width > 72) notes.push(`首行宽度 ${width}，超过 72`)
    if (/[。.]$/.test(subject)) notes.push('首行结尾不应有句号')
  }
  if (extra.length > 0) notes.push(`带了 ${extra.length} 行正文或脚注，规范要求只有一行`)

  if (notes.length > 0) problems += 1
  const mark = notes.length === 0 ? 'ok  ' : 'FAIL'
  console.log(`  ${mark} ${hash} ${notes.length === 0 ? '' : notes.join('；') + ' — '}${subject}`)
}

console.log(problems === 0
  ? `\n检查了 ${commits.length} 条提交，全部符合规范`
  : `\n检查了 ${commits.length} 条提交，其中 ${problems} 条不符合规范（规范文件：${specPath}）`)
process.exit(problems === 0 ? 0 : 1)