sori.studio/lib/media-catalog.js

import { basename, extname, posix } from 'node:path'
import { z } from 'zod'

const mediaUrlSchema = z.string().trim().min(1).max(2048)
const imageExtensions = new Set(['.jpg', '.jpeg', '.png', '.webp', '.gif'])
const videoExtensions = new Set(['.mp4', '.webm', '.mov'])
const audioExtensions = new Set(['.mp3', '.wav', '.ogg', '.m4a'])

/**
 * PostgreSQL LIKE 검색어의 wildcard 문자를 리터럴로 이스케이프한다.
 * SQL에서는 `ESCAPE '!'`와 함께 사용한다.
 * @param {string} value - 검색어
 * @returns {string} LIKE 리터럴
 */
export const escapePostgresLikeLiteral = (value) => String(value || '')
  .replace(/!/g, '!!')
  .replace(/%/g, '!%')
  .replace(/_/g, '!_')

/**
 * 업로드 URL을 카탈로그의 공개 URL과 저장 키로 정규화한다.
 * @param {unknown} value - 입력 URL
 * @returns {{ url: string, storageKey: string }} 정규화 결과
 */
export const normalizeMediaCatalogUrl = (value) => {
  const input = mediaUrlSchema.parse(value)
  const pathOnly = input.split(/[?#]/, 1)[0]
  let decodedPath = ''

  try {
    decodedPath = decodeURIComponent(pathOnly.replace(/%(?![0-9a-f]{2})/gi, '%25'))
  } catch {
    throw new Error('MEDIA_URL_INVALID_ENCODING')
  }

  if (decodedPath.includes('\0') || decodedPath.includes('\\')) {
    throw new Error('MEDIA_URL_INVALID_PATH')
  }

  const normalizedPath = posix.normalize(decodedPath)
  const rawSegments = decodedPath.split('/')
  const normalizedSegments = normalizedPath.split('/')

  if (rawSegments.some((segment) => segment === '.' || segment === '..')
    || normalizedSegments.some((segment) => segment === '..')
    || normalizedPath !== decodedPath.replace(/\/+/g, '/')) {
    throw new Error('MEDIA_URL_TRAVERSAL')
  }

  if (!normalizedPath.startsWith('/uploads/') || normalizedPath === '/uploads/') {
    throw new Error('MEDIA_URL_OUTSIDE_UPLOADS')
  }

  const storageKey = normalizedPath.slice('/uploads/'.length)

  return {
    url: `/uploads/${storageKey}`,
    storageKey
  }
}

/**
 * 확장자에서 미디어 종류를 판별한다.
 * @param {string} fileName - 파일명
 * @returns {'image'|'video'|'audio'|'file'} 미디어 종류
 */
export const detectMediaKind = (fileName) => {
  const extension = extname(String(fileName || '')).toLowerCase()

  if (imageExtensions.has(extension)) {
    return 'image'
  }
  if (videoExtensions.has(extension)) {
    return 'video'
  }
  if (audioExtensions.has(extension)) {
    return 'audio'
  }
  return 'file'
}

/**
 * 저장 키에서 미디어 역할과 원본 후보 키를 판별한다.
 * @param {string} storageKey - 업로드 루트 기준 저장 키
 * @returns {{ role: string, parentStorageKey: string|null }} 역할 판별 결과
 */
export const detectMediaRole = (storageKey) => {
  const key = String(storageKey || '').replace(/^\/+/, '')
  const fileName = basename(key)
  const directory = posix.dirname(key)
  const thumbsDirectory = directory.endsWith('/thumbs')
  const derivedMatch = fileName.match(/^(.*)-(card|page-cover|body(?:-(?:720|1600|2560))?|lightbox)\.webp$/i)

  if (key.startsWith('exports/')) {
    return { role: 'export', parentStorageKey: null }
  }
  if (key.includes('/members/avatars/') || key.startsWith('members/avatars/')) {
    return { role: 'avatar', parentStorageKey: null }
  }
  if (/^(system|site)\//.test(key) || key.startsWith('tags/')) {
    return { role: 'system', parentStorageKey: null }
  }
  if (thumbsDirectory && derivedMatch) {
    const originalDirectory = posix.dirname(directory)
    const parentCandidates = ['.jpg', '.jpeg', '.png', '.webp', '.gif']
    const parentStorageKey = parentCandidates
      .map((extension) => posix.join(originalDirectory, `${derivedMatch[1]}${extension}`))[0]
    const variant = derivedMatch[2].toLowerCase()
    const role = variant === 'card'
      ? 'card'
      : variant === 'page-cover'
        ? 'page_cover'
        : 'body'

    return { role, parentStorageKey }
  }

  return { role: 'original', parentStorageKey: null }
}

/**
 * 문자열 필드에서 로컬 미디어 URL 참조를 중복 없이 추출한다.
 * @param {Record<string, unknown>} fields - 필드명별 값
 * @returns {Array<{ mediaUrl: string, referenceField: string }>} 참조 목록
 */
export const extractMediaReferences = (fields) => {
  const references = new Map()
  const mediaPattern = /\/uploads\/[^\s"'<>()[\]{}]+/g

  for (const [referenceField, rawValue] of Object.entries(fields || {})) {
    if (typeof rawValue !== 'string' || !rawValue.includes('/uploads/')) {
      continue
    }

    for (const match of rawValue.matchAll(mediaPattern)) {
      const candidate = match[0].replace(/[.,;:!?]+$/g, '')

      try {
        const { url } = normalizeMediaCatalogUrl(candidate)
        references.set(`${referenceField}\0${url}`, {
          mediaUrl: url,
          referenceField
        })
      } catch {
        continue
      }
    }
  }

  return [...references.values()]
}