mirror of
https://github.com/hicccc77/WeFlow.git
synced 2026-10-25 17:36:32 +08:00
153 lines
5.6 KiB
TypeScript
153 lines
5.6 KiB
TypeScript
/**
|
||
* 账号目录解析器(Worker 线程 / 主进程通用)
|
||
*
|
||
* 职责:在 dbPath 根目录下,根据传入的账号 ID,找出实际写入数据
|
||
* 的那个账号子目录,例如:
|
||
* dbPath = <数据根目录>
|
||
* accountId = customid_abcd 或 customid
|
||
* 期望返回 <数据根目录>/customid_abcd(带后缀、有 session.db 的那个)
|
||
*
|
||
* 与 ConfigService.getAccountDir 行为保持一致;二者实现独立是因为本文件
|
||
* 也会在 Worker 线程中被加载,无法依赖 electron-store。
|
||
*/
|
||
import { existsSync, readdirSync, statSync } from 'fs'
|
||
import { join } from 'path'
|
||
|
||
// 解析结果缓存(进程内,避免重复 IO)。key = `${dbPath}|${cleanedAccountId}`
|
||
const accountDirCache = new Map<string, string>()
|
||
|
||
/**
|
||
* 把账号 ID 字符串"标准化"为目录前缀。
|
||
* - 账号ID_后缀(4 位) → 账号ID(例如 customid_abcd → customid)
|
||
* - 其他形式 → 原样返回
|
||
*
|
||
* 注意:清洗只是为了得到"前缀"用于扫描匹配,并不代表清洗结果就是真实目录名。
|
||
* 真实目录名仍需在 dbPath 下按"前缀 + 任意后缀"扫描得出。
|
||
*/
|
||
const cleanAccountDirName = (dirName: string): string => {
|
||
const trimmed = dirName.trim()
|
||
if (!trimmed) return trimmed
|
||
|
||
const suffixMatch = trimmed.match(/^(.+)_([a-zA-Z0-9]{4})$/)
|
||
if (suffixMatch) return suffixMatch[1]
|
||
|
||
return trimmed
|
||
}
|
||
|
||
const isDirectory = (path: string): boolean => {
|
||
try {
|
||
return statSync(path).isDirectory()
|
||
} catch {
|
||
return false
|
||
}
|
||
}
|
||
|
||
/**
|
||
* 解析账号目录的真实绝对路径。
|
||
*
|
||
* 匹配流程:
|
||
* 1. 扫描 dbPath 下所有子目录;
|
||
* 2. 同时接受**精确匹配**(`entry == cleanedAccountId`) 与
|
||
* **后缀匹配**(`entry.startsWith(cleanedAccountId + '_')`) 两种命中方式;
|
||
* 3. 用 {@link accountDirLooksValid} 过滤掉"看起来根本不像账号目录"的项
|
||
* (没有 db_storage 也没有 FileStorage/Image[2]);
|
||
* 4. 在剩余候选中按以下优先级排序,取最优:
|
||
* - **有 session.db** > 没有:区分"真正写入数据"与"残留空目录";
|
||
* - **后缀匹配** > 精确匹配;
|
||
* - **修改时间更新** > 更旧:兜底。
|
||
*/
|
||
export const resolveAccountDir = (dbPath?: string, accountId?: string): string | null => {
|
||
if (!dbPath || !accountId) return null
|
||
|
||
const cleanedAccountId = cleanAccountDirName(accountId)
|
||
const normalized = dbPath.replace(/[\\/]+$/, '')
|
||
const cacheKey = `${normalized}|${cleanedAccountId.toLowerCase()}`
|
||
|
||
// 命中缓存且目标仍存在则直接返回;目标已被删除的过期缓存项会被剔除
|
||
const cached = accountDirCache.get(cacheKey)
|
||
if (cached && existsSync(cached)) return cached
|
||
if (cached && !existsSync(cached)) {
|
||
accountDirCache.delete(cacheKey)
|
||
}
|
||
|
||
const lowerAccountId = cleanedAccountId.toLowerCase()
|
||
|
||
try {
|
||
const entries = readdirSync(normalized)
|
||
type Candidate = { entryPath: string; isExact: boolean; hasSession: boolean; mtime: number }
|
||
const candidates: Candidate[] = []
|
||
|
||
for (const entry of entries) {
|
||
const entryPath = join(normalized, entry)
|
||
if (!isDirectory(entryPath)) continue
|
||
|
||
const lowerEntry = entry.toLowerCase()
|
||
const isExactMatch = lowerEntry === lowerAccountId
|
||
const isSuffixMatch = lowerEntry.startsWith(`${lowerAccountId}_`)
|
||
// 既不是精确命中、也不是前缀命中 → 与本账号 ID 无关,跳过
|
||
if (!isExactMatch && !isSuffixMatch) continue
|
||
|
||
// 看起来不像账号目录(连 db_storage 与 FileStorage/Image 都没有)→ 跳过
|
||
if (!accountDirLooksValid(entryPath)) continue
|
||
|
||
let mtime = 0
|
||
try { mtime = statSync(entryPath).mtimeMs } catch { /* 忽略 stat 异常 */ }
|
||
candidates.push({
|
||
entryPath,
|
||
isExact: isExactMatch,
|
||
hasSession: accountDirHasSessionDb(entryPath),
|
||
mtime,
|
||
})
|
||
}
|
||
|
||
if (candidates.length > 0) {
|
||
candidates.sort((a, b) => {
|
||
// 1) 优先选有 session.db 的(真实写入数据的目录)
|
||
if (a.hasSession !== b.hasSession) return a.hasSession ? -1 : 1
|
||
// 2) 其次优先选"带后缀"的
|
||
if (a.isExact !== b.isExact) return a.isExact ? 1 : -1
|
||
// 3) 最后按修改时间倒序(最新的优先)
|
||
return b.mtime - a.mtime
|
||
})
|
||
const best = candidates[0].entryPath
|
||
accountDirCache.set(cacheKey, best)
|
||
return best
|
||
}
|
||
} catch { /* 扫描目录失败时直接 fallthrough 返回 null */ }
|
||
|
||
return null
|
||
}
|
||
|
||
/**
|
||
* 浅层判定一个目录"看起来像不像账号目录":
|
||
* 存在 db_storage 子目录,或存在 FileStorage/Image[2] 子目录之一即认为是。
|
||
*
|
||
* 用于在候选阶段剔除"同名但实际无数据"的残留空目录。
|
||
*/
|
||
const accountDirLooksValid = (entryPath: string): boolean => {
|
||
return (
|
||
existsSync(join(entryPath, 'db_storage')) ||
|
||
existsSync(join(entryPath, 'FileStorage', 'Image')) ||
|
||
existsSync(join(entryPath, 'FileStorage', 'Image2'))
|
||
)
|
||
}
|
||
|
||
/**
|
||
* 检测账号目录下是否存在 session.db。
|
||
*
|
||
* 是排序优先级里"区分真实写入数据 vs 仅有空 db_storage 骨架"的关键判据,
|
||
* 同时兼容两种已知布局:
|
||
* - db_storage/session/session.db (嵌套布局)
|
||
* - db_storage/session.db (扁平布局)
|
||
*/
|
||
const accountDirHasSessionDb = (entryPath: string): boolean => {
|
||
const candidates = [
|
||
join(entryPath, 'db_storage', 'session', 'session.db'),
|
||
join(entryPath, 'db_storage', 'session.db'),
|
||
]
|
||
for (const candidate of candidates) {
|
||
if (existsSync(candidate)) return true
|
||
}
|
||
return false
|
||
}
|