happy-export/backend/lib/dataTransfer/index.js

/**
 * 数据迁移的门面：把「读包、预览、真导、备份」串起来，路由只调这一层。
 *
 * 预览为什么要开一个库副本：
 *   预览必须回答「照这样导，线上会变成什么样」。如果只是把行数点一遍，就答不了
 *   「哪一行会因为外键或非空约束失败」。所以预览在做一次 VACUUM INTO 出来的副本上
 *   **真跑一遍导入**，跑完把副本删掉 —— 报告里的数字与实际导入完全同源。
 *   库只有几 MB，拷一份比维护两套判断逻辑便宜得多。
 */
const fs = require('fs')
const path = require('path')
const crypto = require('crypto')
const Database = require('better-sqlite3-multiple-ciphers')

const { CATEGORY_LIST, TABLES, resolveCategories } = require('./catalog')
const { writeBundle, readBundle, tableExists } = require('./bundle')
const { importInto } = require('./import')

const escapePath = file => String(file).replace(/'/g, "''")

/** 本机各类别各表当前行数（界面用来显示「这次会导出多少」） */
function listCatalog(db) {
  return CATEGORY_LIST.map(category => ({
    key: category.key,
    label: category.label,
    description: category.description,
    tables: TABLES.filter(spec => spec.category === category.key).map(spec => {
      let rows = 0
      if (tableExists(db, spec.table)) {
        try { rows = db.prepare(`SELECT COUNT(*) AS n FROM "${spec.table}"`).get().n } catch { rows = 0 }
      }
      return { table: spec.table, label: spec.label, rows }
    })
  }))
}

/** 导出到可写流（HTTP 响应直接接） */
async function exportToStream(db, output, options) {
  await writeBundle(db, output, options)
}

/** 导出的文件名：带时间戳，便于在服务器上分辨先后 */
function exportFileName(date = new Date()) {
  const pad = value => String(value).padStart(2, '0')
  const stamp = `${date.getFullYear()}${pad(date.getMonth() + 1)}${pad(date.getDate())}-${pad(date.getHours())}${pad(date.getMinutes())}${pad(date.getSeconds())}`
  return `happy-export-${stamp}.hdata.gz`
}

function readFileMeta(bundle, filePath, fileName) {
  return {
    name: fileName || path.basename(filePath),
    size: bundle.size,
    exportedAt: bundle.header.exportedAt,
    exporter: bundle.header.exporter || null,
    app: bundle.header.app,
    version: bundle.header.version,
    categories: bundle.categories
  }
}

/**
 * 预览：在库副本上真跑一遍导入，返回与实际导入同源的报告。
 * 副本用完即删，线上库不受任何影响。
 */
async function previewImport(sourceDb, filePath, { categories, user, workDir, fileName }) {
  fs.mkdirSync(workDir, { recursive: true })
  const bundle = await readBundle(filePath)
  const selected = resolveCategories(categories && categories.length ? categories : bundle.categories)
  const available = selected.filter(key => bundle.categories.includes(key))
  if (!available.length) {
    const error = new Error('所选类别在数据包里没有内容，无法导入')
    error.statusCode = 400
    throw error
  }

  const copyPath = path.join(workDir, `preview-${crypto.randomBytes(6).toString('hex')}.db`)
  if (fs.existsSync(copyPath)) fs.unlinkSync(copyPath)
  sourceDb.exec(`VACUUM INTO '${escapePath(copyPath)}'`)

  const copy = new Database(copyPath)
  try {
    copy.pragma('foreign_keys = ON')
    copy.pragma('busy_timeout = 5000')
    const plan = copy.transaction(() => importInto(copy, bundle, { categories: available, user }))()
    return { file: readFileMeta(bundle, filePath, fileName), plan, availableCategories: available }
  } finally {
    try { copy.close() } catch { /* 副本关闭失败不影响结果 */ }
    try { fs.unlinkSync(copyPath) } catch { /* 副本已被占用就留着，下次预览会覆盖 */ }
  }
}

/**
 * 真正导入：先给线上库做一份完整备份（留档可回滚），再在一个事务里写入。
 * 任何未预期的异常都会整体回滚，备份文件保留在 backupDir。
 */
async function applyImport(db, filePath, { categories, user, backupDir, fileName }) {
  fs.mkdirSync(backupDir, { recursive: true })
  const bundle = await readBundle(filePath)
  const selected = resolveCategories(categories && categories.length ? categories : bundle.categories)
  const available = selected.filter(key => bundle.categories.includes(key))
  if (!available.length) {
    const error = new Error('所选类别在数据包里没有内容，无法导入')
    error.statusCode = 400
    throw error
  }

  /* 时间戳精确到秒，再拼一段随机后缀：同一秒内连续导两次也不会撞出
     「output file already exists」，而文件名照样按时间可排序。 */
  const stamp = new Date().toISOString().replace(/[-:TZ.]/g, '').slice(0, 14)
  const backupPath = path.join(backupDir, `pre-import-${stamp}-${crypto.randomBytes(3).toString('hex')}.db`)
  db.exec(`VACUUM INTO '${escapePath(backupPath)}'`)

  const plan = db.transaction(() => importInto(db, bundle, { categories: available, user }))()
  return { file: readFileMeta(bundle, filePath, fileName), plan, backup: backupPath, availableCategories: available }
}

/** 清理过期的上传包与预览副本（默认 24 小时） */
function cleanupUploads(dir, { maxAgeMs = 24 * 60 * 60 * 1000 } = {}) {
  let removed = 0
  try {
    for (const name of fs.readdirSync(dir)) {
      const file = path.join(dir, name)
      try {
        const stat = fs.statSync(file)
        /* 只删普通文件：目录里还套着 multer 的 uploads/，别把目录本身当过期文件删 */
        if (!stat.isFile()) continue
        if (Date.now() - stat.mtimeMs > maxAgeMs) { fs.unlinkSync(file); removed++ }
      } catch { /* 单个文件失败不影响其它清理 */ }
    }
  } catch { /* 目录还不存在 */ }
  return removed
}

module.exports = {
  listCatalog,
  exportToStream,
  exportFileName,
  previewImport,
  applyImport,
  cleanupUploads
}