初始化
缺陷：主题配色需要优化整体的同风格
This commit is contained in:
692
app/Service/MaterialService.php
Normal file
692
app/Service/MaterialService.php
Normal file
@@ -0,0 +1,692 @@
|
||||
<?php
|
||||
|
||||
namespace App\Service;
|
||||
|
||||
use App\BaseApp\BaseService;
|
||||
use App\Models\FileFolderModel;
|
||||
use App\Models\FileModel;
|
||||
use App\Service\common\MediaUrlService;
|
||||
use App\Service\common\oss\OssRuntimeConfigService;
|
||||
use App\Service\common\oss\OssStorageFactory;
|
||||
use App\Service\common\oss\OssStorageInterface;
|
||||
use Exception;
|
||||
use Illuminate\Support\Facades\DB;
|
||||
use Illuminate\Support\Facades\Schema;
|
||||
use Psr\Container\ContainerExceptionInterface;
|
||||
use Psr\Container\NotFoundExceptionInterface;
|
||||
use Throwable;
|
||||
|
||||
/**
|
||||
* 素材库
|
||||
*
|
||||
* 三个动作串成一条链，顺序不能颠倒：
|
||||
* syncFromOss 把 bucket 里的对象补进 nl_file（历史文件根本没有上传记录）
|
||||
* scanReferences 按 config/media_refs 把「谁在用」数出来，回填 ref_count
|
||||
* reclaim 只删 ref_count = 0 且扫过的，先删远端再软删记录
|
||||
* 少了中间那步，刚同步进来的素材 ref_count 默认就是 0，一键回收等于清空 bucket，
|
||||
* 所以 reclaim 里对 last_scan_at 的校验不是冗余检查。
|
||||
*/
|
||||
class MaterialService extends BaseService
|
||||
{
|
||||
/** 单次同步默认条数：一屏够看，又不至于把请求跑到超时 */
|
||||
private const SYNC_DEFAULT_LIMIT = 200;
|
||||
|
||||
/** 单次同步上限，防止前端把 limit 传成十万 */
|
||||
private const SYNC_MAX_LIMIT = 1000;
|
||||
|
||||
/** 分批取值 / 分批回写的批大小 */
|
||||
private const CHUNK_SIZE = 1000;
|
||||
|
||||
private MediaUrlService $media;
|
||||
|
||||
public function __construct()
|
||||
{
|
||||
// Job 版本在 CLI / 队列里跑，那里没有 token；HTTP 入口仍然要求登录
|
||||
if (app()->runningInConsole()) {
|
||||
$this->isAuth = false;
|
||||
}
|
||||
parent::__construct();
|
||||
$this->model = FileModel::class;
|
||||
$this->selectField = [
|
||||
'id', 'user_id', 'oss_config_id', 'folder_id', 'name', 'url', 'path', 'ext',
|
||||
'type', 'size', 'width', 'height', 'hash', 'ref_count', 'last_scan_at',
|
||||
'source', 'created_at', 'updated_at',
|
||||
];
|
||||
$this->queryField = [
|
||||
'folder_id' => '=',
|
||||
'type' => '=',
|
||||
'ext' => '=',
|
||||
'oss_config_id' => '=',
|
||||
'source' => '=',
|
||||
];
|
||||
$this->media = MediaUrlService::getInstance();
|
||||
}
|
||||
|
||||
/**
|
||||
* 素材列表
|
||||
*
|
||||
* keyword 要同时命中 name / path / url，而基类的 queryField 只会按列 AND，
|
||||
* 拼不出这个 OR 组，所以这里自己组查询，返回结构与 getPageList 保持一致。
|
||||
*
|
||||
* @throws ContainerExceptionInterface
|
||||
* @throws NotFoundExceptionInterface
|
||||
*/
|
||||
public function list(): array
|
||||
{
|
||||
$this->getWhere();
|
||||
$keyword = trim((string) request()->get('keyword', ''));
|
||||
$unused = (int) request()->get('unused', 0);
|
||||
$searchTime = request()->get('search_time');
|
||||
|
||||
$query = FileModel::where($this->where)
|
||||
->when($keyword !== '', function ($q) use ($keyword) {
|
||||
$q->where(function ($sub) use ($keyword) {
|
||||
$sub->where('name', 'like', '%' . $keyword . '%')
|
||||
->orWhere('path', 'like', '%' . $keyword . '%')
|
||||
->orWhere('url', 'like', '%' . $keyword . '%');
|
||||
});
|
||||
})
|
||||
->when($unused === 1, fn ($q) => $q->where('ref_count', 0))
|
||||
->when(!empty($searchTime), function ($q) use ($searchTime) {
|
||||
$this->getWhereBetween($searchTime);
|
||||
$q->whereBetween($this->whereBetween[0], $this->whereBetween[1]);
|
||||
});
|
||||
|
||||
return $this->toPage($query->select($this->selectField)->orderByDesc('id'));
|
||||
}
|
||||
|
||||
/**
|
||||
* 详情，带上文件夹名，免得前端为了显示一个名字再请求一次目录树
|
||||
*
|
||||
* @throws Exception
|
||||
*/
|
||||
public function detail($id): mixed
|
||||
{
|
||||
$info = $this->getDetail($id);
|
||||
$info->url = $this->media->toPublic($info->url);
|
||||
$info->folder_name = (string) (FileFolderModel::where('id', (int) $info->folder_id)->value('name') ?? '');
|
||||
return $info;
|
||||
}
|
||||
|
||||
/**
|
||||
* 只允许改素材名与所属文件夹
|
||||
*
|
||||
* path / hash / size 是同步结果，手工改会让引用扫描直接失准：
|
||||
* path 一旦被改歪，原本在用的素材就匹配不上任何引用，转头出现在可回收列表里。
|
||||
*
|
||||
* @throws Exception
|
||||
*/
|
||||
public function update($id, $params): mixed
|
||||
{
|
||||
$data = [];
|
||||
if (array_key_exists('name', $params)) {
|
||||
$data['name'] = mb_substr(trim((string) $params['name']), 0, 255);
|
||||
}
|
||||
if (array_key_exists('folder_id', $params)) {
|
||||
$data['folder_id'] = $this->assertFolder((int) $params['folder_id']);
|
||||
}
|
||||
if (empty($data)) {
|
||||
$this->utils->errorThrow('没有可更新的内容');
|
||||
}
|
||||
return $this->save($id, $data);
|
||||
}
|
||||
|
||||
/**
|
||||
* 只软删素材记录，不动远端对象
|
||||
*
|
||||
* 远端删除必须走 reclaim 的二次校验；这里如果顺手删了远端，
|
||||
* 误删的文件就再也找不回来了，而软删记录随时可以恢复。
|
||||
*
|
||||
* @throws Exception
|
||||
*/
|
||||
public function delete($ids): mixed
|
||||
{
|
||||
return $this->del(is_array($ids) ? $ids : [$ids]);
|
||||
}
|
||||
|
||||
/**
|
||||
* 素材概览：总量、占用空间与可回收部分
|
||||
*/
|
||||
public function stat(): array
|
||||
{
|
||||
$base = static fn () => FileModel::where('deleted_at', 0);
|
||||
$types = $base()->selectRaw('type, COUNT(*) as count')
|
||||
->groupBy('type')
|
||||
->orderBy('type')
|
||||
->get()
|
||||
->map(static fn ($row) => ['type' => (int) $row->type, 'count' => (int) $row->count])
|
||||
->all();
|
||||
|
||||
return [
|
||||
'total' => $base()->count(),
|
||||
'total_size' => (int) $base()->sum('size'),
|
||||
'unused' => $base()->where('ref_count', 0)->count(),
|
||||
'unused_size' => (int) $base()->where('ref_count', 0)->sum('size'),
|
||||
'types' => $types,
|
||||
];
|
||||
}
|
||||
|
||||
/**
|
||||
* 从 OSS 增量拉对象进素材表
|
||||
*
|
||||
* 一次只处理一页：返回的 next_marker 非空就带着它再调一次，前端点几下即可拉完。
|
||||
* 幂等靠三级匹配 —— 先按对象键，再按完整地址（换过域名的老记录），
|
||||
* 最后才按哈希兜住「地址早就变了、内容没变」的历史上传记录。
|
||||
* 反复调用只会更新 size / hash，不会重复插入。
|
||||
*
|
||||
* @param array $params oss_config_id / prefix / marker / limit
|
||||
* @throws Exception
|
||||
*/
|
||||
public function syncFromOss(array $params): array
|
||||
{
|
||||
$configId = (int) ($params['oss_config_id'] ?? 0);
|
||||
if ($configId <= 0) {
|
||||
$this->utils->errorThrow('请选择要同步的存储配置');
|
||||
}
|
||||
$limit = (int) ($params['limit'] ?? self::SYNC_DEFAULT_LIMIT);
|
||||
$limit = $limit > 0 ? min($limit, self::SYNC_MAX_LIMIT) : self::SYNC_DEFAULT_LIMIT;
|
||||
|
||||
$page = $this->driverOf($configId)->listObjects(
|
||||
(string) ($params['prefix'] ?? ''),
|
||||
(string) ($params['marker'] ?? ''),
|
||||
$limit
|
||||
);
|
||||
$items = array_values(array_filter(
|
||||
(array) ($page['items'] ?? []),
|
||||
static fn ($item) => !empty($item['key']) && !str_ends_with((string) $item['key'], '/')
|
||||
));
|
||||
$result = [
|
||||
'inserted' => 0,
|
||||
'updated' => 0,
|
||||
'next_marker' => (string) ($page['next_marker'] ?? ''),
|
||||
'finished' => (bool) ($page['finished'] ?? true),
|
||||
];
|
||||
if (empty($items)) {
|
||||
return $result;
|
||||
}
|
||||
|
||||
[$byPath, $byUrl, $byHash] = $this->existingIndexOf($items);
|
||||
$now = time();
|
||||
$pending = [];
|
||||
$seen = [];
|
||||
foreach ($items as $item) {
|
||||
$key = (string) $item['key'];
|
||||
if (isset($seen[$key])) {
|
||||
continue;
|
||||
}
|
||||
$seen[$key] = true;
|
||||
$url = (string) ($item['url'] ?? '');
|
||||
$hash = (string) ($item['hash'] ?? '');
|
||||
$ext = strtolower((string) pathinfo($key, PATHINFO_EXTENSION));
|
||||
$row = $byPath[$key] ?? $byUrl[$url] ?? ($hash !== '' ? ($byHash[$hash] ?? null) : null);
|
||||
|
||||
if ($row !== null) {
|
||||
FileModel::where('id', $row->id)->update(
|
||||
$this->backfillOf($row, $key, $url, $hash, $ext, (int) ($item['size'] ?? 0), $configId, $now)
|
||||
);
|
||||
$result['updated']++;
|
||||
continue;
|
||||
}
|
||||
$pending[] = [
|
||||
'user_id' => $this->userId,
|
||||
'oss_config_id' => $configId,
|
||||
'folder_id' => 0,
|
||||
'name' => basename($key),
|
||||
'url' => $url,
|
||||
'path' => $key,
|
||||
'ext' => $ext,
|
||||
'type' => FileModel::typeOfExt($ext),
|
||||
'size' => (int) ($item['size'] ?? 0),
|
||||
'width' => 0,
|
||||
'height' => 0,
|
||||
'hash' => $hash,
|
||||
'ref_count' => 0,
|
||||
// 新补录的素材必须是 0：非 0 会被 reclaim 当成「扫过且没人用」直接删掉
|
||||
'last_scan_at' => 0,
|
||||
'source' => FileModel::SOURCE_OSS,
|
||||
// 用对象的实际修改时间当上传时间，否则批量补录出来的素材
|
||||
// 创建时间全挤在同一秒，「N 天前的无引用文件」这个筛选就没意义了
|
||||
'created_at' => (int) ($item['last_modified'] ?? 0) ?: $now,
|
||||
'updated_at' => 0,
|
||||
'deleted_at' => 0,
|
||||
];
|
||||
}
|
||||
if (!empty($pending)) {
|
||||
FileModel::insert($pending);
|
||||
$result['inserted'] = count($pending);
|
||||
}
|
||||
return $result;
|
||||
}
|
||||
|
||||
/**
|
||||
* 全库引用扫描，回填 ref_count 与 last_scan_at
|
||||
*
|
||||
* 匹配策略：先按完整对象键，再退回文件名兜底。上传时的文件名是随机串，
|
||||
* 现实中不会撞车；真撞车（同名多条）就放弃文件名兜底，宁可少算一次引用，
|
||||
* 也不能把引用记到错的素材头上 —— 记错的那一头会被判成可回收。
|
||||
*
|
||||
* 软删的业务行也照样算引用：那些记录随时可能被恢复，
|
||||
* 把它们的封面提前删掉等于让恢复出来的数据全是裂图。
|
||||
*/
|
||||
public function scanReferences(array $params = []): array
|
||||
{
|
||||
[$byPath, $byName] = $this->buildMaterialIndex();
|
||||
$counts = [];
|
||||
$scanned = 0;
|
||||
|
||||
foreach ((array) config('media_refs', []) as $ref) {
|
||||
$conn = (string) ($ref['connection'] ?? 'mysql');
|
||||
$table = (string) ($ref['table'] ?? '');
|
||||
$field = (string) ($ref['field'] ?? '');
|
||||
$kind = (string) ($ref['kind'] ?? 'single');
|
||||
if ($table === '' || $field === '') {
|
||||
continue;
|
||||
}
|
||||
try {
|
||||
$schema = Schema::connection($conn);
|
||||
if (!$schema->hasTable($table) || !$schema->hasColumn($table, $field)) {
|
||||
// 登记了但库里还没这列（比如 catalogue.video 尚未上线）：跳过而不是报错，
|
||||
// 否则一个规划中的字段就能让整次扫描前功尽弃
|
||||
continue;
|
||||
}
|
||||
} catch (Throwable $e) {
|
||||
continue;
|
||||
}
|
||||
|
||||
DB::connection($conn)->table($table)
|
||||
->select(['id', $field])
|
||||
->orderBy('id')
|
||||
->chunk(self::CHUNK_SIZE, function ($rows) use ($field, $kind, $byPath, $byName, &$counts, &$scanned) {
|
||||
foreach ($rows as $row) {
|
||||
$raw = $row->{$field} ?? null;
|
||||
if ($raw === null || trim((string) $raw) === '') {
|
||||
continue;
|
||||
}
|
||||
foreach ($this->extractUrls((string) $raw, $kind) as $url) {
|
||||
$key = $this->normalizeRefKey($url);
|
||||
if ($key === '') {
|
||||
continue;
|
||||
}
|
||||
$scanned++;
|
||||
$id = $byPath[$key] ?? ($byName[basename($key)] ?? null);
|
||||
if ($id !== null) {
|
||||
$counts[$id] = ($counts[$id] ?? 0) + 1;
|
||||
}
|
||||
}
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
$now = time();
|
||||
// 先整表归零并盖上扫描时间戳，再把有引用的批量改回去：
|
||||
// 逐条 update 在几万条素材上是几万次往返
|
||||
FileModel::where('deleted_at', 0)->update(['ref_count' => 0, 'last_scan_at' => $now]);
|
||||
$grouped = [];
|
||||
foreach ($counts as $id => $count) {
|
||||
$grouped[(int) $count][] = (int) $id;
|
||||
}
|
||||
foreach ($grouped as $count => $ids) {
|
||||
foreach (array_chunk($ids, self::CHUNK_SIZE) as $slice) {
|
||||
FileModel::whereIn('id', $slice)->update(['ref_count' => $count]);
|
||||
}
|
||||
}
|
||||
|
||||
return [
|
||||
'scanned' => $scanned,
|
||||
'referenced' => count($counts),
|
||||
'unused' => FileModel::where('deleted_at', 0)->where('ref_count', 0)->count(),
|
||||
];
|
||||
}
|
||||
|
||||
/**
|
||||
* 无引用素材清单
|
||||
*
|
||||
* 除了 ref_count = 0 还要卡「上传超过 N 天」：用户上传完图片、表单还没提交时，
|
||||
* 这张图确实没人引用，但它马上就要被用上，不能出现在回收列表里。
|
||||
*/
|
||||
public function unusedList(array $params = []): array
|
||||
{
|
||||
$days = (int) ($params['days'] ?? 30);
|
||||
$days = $days > 0 ? $days : 30;
|
||||
$query = FileModel::where('deleted_at', 0)
|
||||
->where('ref_count', 0)
|
||||
->where('created_at', '<', time() - $days * 86400)
|
||||
->select($this->selectField)
|
||||
// 先给大文件，回收一页就能腾出可观的空间
|
||||
->orderByDesc('size')
|
||||
->orderByDesc('id');
|
||||
return $this->toPage($query, (int) ($params['pageSize'] ?? 20));
|
||||
}
|
||||
|
||||
/**
|
||||
* 回收：删远端对象，成功后软删记录
|
||||
*
|
||||
* 两道闸门都不能省 —— ref_count = 0 说明扫描时没人用，last_scan_at > 0 说明真的扫过。
|
||||
* 单条失败不中断整批：一批几百个对象，前面已经删掉的必须留下软删记录，
|
||||
* 否则库里还挂着记录、远端对象已经没了，素材库里全是点开 404 的幽灵。
|
||||
*
|
||||
* @param array $ids 素材 ID
|
||||
* @throws Exception
|
||||
*/
|
||||
public function reclaim(array $ids): array
|
||||
{
|
||||
$ids = array_values(array_unique(array_filter(array_map('intval', $ids))));
|
||||
if (empty($ids)) {
|
||||
$this->utils->errorThrow('请选择要回收的素材');
|
||||
}
|
||||
|
||||
$rows = FileModel::whereIn('id', $ids)->where('deleted_at', 0)->get();
|
||||
$deleted = 0;
|
||||
$failed = [];
|
||||
$drivers = [];
|
||||
foreach ($rows as $row) {
|
||||
$id = (int) $row->id;
|
||||
if ((int) $row->last_scan_at <= 0) {
|
||||
$failed[] = ['id' => $id, 'reason' => '尚未扫描过引用，请先执行引用扫描'];
|
||||
continue;
|
||||
}
|
||||
if ((int) $row->ref_count !== 0) {
|
||||
$failed[] = ['id' => $id, 'reason' => '仍被引用 ' . (int) $row->ref_count . ' 处'];
|
||||
continue;
|
||||
}
|
||||
$key = trim((string) $row->path) !== ''
|
||||
? (string) $row->path
|
||||
: $this->normalizeRefKey((string) $row->url);
|
||||
if ($key === '') {
|
||||
$failed[] = ['id' => $id, 'reason' => '没有可定位的对象键，请先同步一次'];
|
||||
continue;
|
||||
}
|
||||
try {
|
||||
$configId = (int) $row->oss_config_id;
|
||||
if (!isset($drivers[$configId])) {
|
||||
$drivers[$configId] = $this->driverOf($configId);
|
||||
}
|
||||
if (!$drivers[$configId]->deleteObject($key)) {
|
||||
$failed[] = ['id' => $id, 'reason' => '远端删除失败'];
|
||||
continue;
|
||||
}
|
||||
} catch (Throwable $e) {
|
||||
$failed[] = ['id' => $id, 'reason' => $e->getMessage()];
|
||||
continue;
|
||||
}
|
||||
$now = time();
|
||||
FileModel::where('id', $id)->update(['deleted_at' => $now, 'updated_at' => $now]);
|
||||
$deleted++;
|
||||
}
|
||||
|
||||
$found = $rows->pluck('id')->map(static fn ($value) => (int) $value)->all();
|
||||
foreach (array_diff($ids, $found) as $missing) {
|
||||
$failed[] = ['id' => (int) $missing, 'reason' => '素材不存在或已删除'];
|
||||
}
|
||||
|
||||
return ['deleted' => $deleted, 'failed' => array_values($failed)];
|
||||
}
|
||||
|
||||
/**
|
||||
* 批量移动到文件夹（folder_id = 0 表示移回根目录）
|
||||
*
|
||||
* @throws Exception
|
||||
*/
|
||||
public function moveToFolder(array $ids, int $folderId): array
|
||||
{
|
||||
$ids = array_values(array_unique(array_filter(array_map('intval', $ids))));
|
||||
if (empty($ids)) {
|
||||
$this->utils->errorThrow('请选择要移动的素材');
|
||||
}
|
||||
$folderId = $this->assertFolder($folderId);
|
||||
$moved = FileModel::whereIn('id', $ids)->where('deleted_at', 0)->update([
|
||||
'folder_id' => $folderId,
|
||||
'updated_at' => time(),
|
||||
]);
|
||||
return ['moved' => (int) $moved, 'folder_id' => $folderId];
|
||||
}
|
||||
|
||||
/**
|
||||
* 分页结果统一成前端约定的结构
|
||||
*/
|
||||
private function toPage($query, int $pageSize = 0): array
|
||||
{
|
||||
$pageSize = $pageSize > 0 ? $pageSize : (int) request()->get('pageSize', 20);
|
||||
$result = $query->paginate($pageSize > 0 ? $pageSize : 20)->toArray();
|
||||
foreach ($result['data'] as &$item) {
|
||||
$item['url'] = $this->media->toPublic($item['url'] ?? '');
|
||||
}
|
||||
unset($item);
|
||||
return [
|
||||
'page' => $result['current_page'],
|
||||
'size' => $result['per_page'],
|
||||
'page_count' => $result['last_page'],
|
||||
'total' => $result['total'],
|
||||
'items' => $result['data'],
|
||||
];
|
||||
}
|
||||
|
||||
/**
|
||||
* 一次把本批可能命中的已有记录捞出来，避免逐条 select 打 N 次库
|
||||
*
|
||||
* byHash 只收 path still 为空的行：按内容哈希去认亲很容易把「两个键、同一份内容」
|
||||
* 的对象合成一条，只用来给还没记过对象键的历史上传记录补档。
|
||||
*
|
||||
* @param array<int, array> $items
|
||||
* @return array{0: array<string, mixed>, 1: array<string, mixed>, 2: array<string, mixed>}
|
||||
*/
|
||||
private function existingIndexOf(array $items): array
|
||||
{
|
||||
$paths = array_values(array_unique(array_column($items, 'key')));
|
||||
$urls = array_values(array_unique(array_filter(array_column($items, 'url'))));
|
||||
$hashes = array_values(array_unique(array_filter(array_column($items, 'hash'))));
|
||||
|
||||
$rows = FileModel::where('deleted_at', 0)
|
||||
->where(function ($q) use ($paths, $urls, $hashes) {
|
||||
$q->whereIn('path', $paths);
|
||||
if (!empty($urls)) {
|
||||
$q->orWhereIn('url', $urls);
|
||||
}
|
||||
if (!empty($hashes)) {
|
||||
$q->orWhere(function ($sub) use ($hashes) {
|
||||
$sub->where('path', '')->whereIn('hash', $hashes);
|
||||
});
|
||||
}
|
||||
})
|
||||
->get(['id', 'name', 'url', 'path', 'ext', 'type', 'hash', 'oss_config_id']);
|
||||
|
||||
$byPath = [];
|
||||
$byUrl = [];
|
||||
$byHash = [];
|
||||
foreach ($rows as $row) {
|
||||
$path = (string) $row->path;
|
||||
$url = (string) $row->url;
|
||||
$hash = (string) $row->hash;
|
||||
if ($path !== '') {
|
||||
$byPath[$path] = $row;
|
||||
}
|
||||
if ($url !== '' && !isset($byUrl[$url])) {
|
||||
$byUrl[$url] = $row;
|
||||
}
|
||||
if ($path === '' && $hash !== '' && !isset($byHash[$hash])) {
|
||||
$byHash[$hash] = $row;
|
||||
}
|
||||
}
|
||||
return [$byPath, $byUrl, $byHash];
|
||||
}
|
||||
|
||||
/**
|
||||
* 已有记录的更新字段
|
||||
*
|
||||
* size / hash 每次都覆盖（对象可能被同名替换过），其余列只在原值为空时补，
|
||||
* 免得把用户在素材库里改过的名字、归过的文件夹又冲回默认值。
|
||||
*/
|
||||
private function backfillOf(
|
||||
mixed $row,
|
||||
string $key,
|
||||
string $url,
|
||||
string $hash,
|
||||
string $ext,
|
||||
int $size,
|
||||
int $configId,
|
||||
int $now
|
||||
): array {
|
||||
$update = ['size' => $size, 'updated_at' => $now];
|
||||
if ($hash !== '') {
|
||||
$update['hash'] = $hash;
|
||||
}
|
||||
if (trim((string) $row->path) === '') {
|
||||
$update['path'] = $key;
|
||||
}
|
||||
if (trim((string) $row->url) === '' && $url !== '') {
|
||||
$update['url'] = $url;
|
||||
}
|
||||
if (trim((string) $row->ext) === '' && $ext !== '') {
|
||||
$update['ext'] = $ext;
|
||||
$update['type'] = FileModel::typeOfExt($ext);
|
||||
}
|
||||
if (trim((string) $row->name) === '') {
|
||||
$update['name'] = basename($key);
|
||||
}
|
||||
if ((int) $row->oss_config_id === 0) {
|
||||
$update['oss_config_id'] = $configId;
|
||||
}
|
||||
return $update;
|
||||
}
|
||||
|
||||
/**
|
||||
* 素材索引：对象键 → id，文件名 → id
|
||||
*
|
||||
* 文件名索引里同名多条时置 null（放弃兜底），见 scanReferences 的说明。
|
||||
*
|
||||
* @return array{0: array<string, int>, 1: array<string, int|null>}
|
||||
*/
|
||||
private function buildMaterialIndex(): array
|
||||
{
|
||||
$byPath = [];
|
||||
$byName = [];
|
||||
FileModel::where('deleted_at', 0)
|
||||
->select(['id', 'path', 'url'])
|
||||
->orderBy('id')
|
||||
->chunk(self::CHUNK_SIZE, function ($rows) use (&$byPath, &$byName) {
|
||||
foreach ($rows as $row) {
|
||||
$id = (int) $row->id;
|
||||
// path 与 url 都进索引：老记录只有 url，新记录两者都有
|
||||
foreach ([(string) $row->path, (string) $row->url] as $candidate) {
|
||||
$key = $this->normalizeRefKey($candidate);
|
||||
if ($key === '') {
|
||||
continue;
|
||||
}
|
||||
$byPath[$key] = $id;
|
||||
$name = basename($key);
|
||||
if ($name === '') {
|
||||
continue;
|
||||
}
|
||||
$byName[$name] = array_key_exists($name, $byName) && $byName[$name] !== $id
|
||||
? null
|
||||
: $id;
|
||||
}
|
||||
}
|
||||
});
|
||||
return [$byPath, $byName];
|
||||
}
|
||||
|
||||
/**
|
||||
* 把引用地址收敛成能与素材 path 比对的对象键
|
||||
*
|
||||
* 同一个文件在库里可能是绝对地址、/storage 相对地址、带 ?imageView2 处理参数
|
||||
* 三种写法，不先归一化就只能匹配上碰巧写法一致的那批。
|
||||
*/
|
||||
private function normalizeRefKey(string $value): string
|
||||
{
|
||||
$value = $this->media->stripProcessParams($value);
|
||||
if ($value === '') {
|
||||
return '';
|
||||
}
|
||||
$path = (string) (parse_url($value, PHP_URL_PATH) ?: $value);
|
||||
$path = ltrim((string) preg_replace('#/{2,}#', '/', rawurldecode($path)), '/');
|
||||
// 本地存储出库地址带 /storage 前缀，对象键里没有
|
||||
if (str_starts_with($path, 'storage/')) {
|
||||
$path = substr($path, 8);
|
||||
}
|
||||
return $path;
|
||||
}
|
||||
|
||||
/**
|
||||
* 按登记的 kind 从字段值里取出地址
|
||||
*
|
||||
* @return array<int, string>
|
||||
*/
|
||||
private function extractUrls(string $value, string $kind): array
|
||||
{
|
||||
$value = trim($value);
|
||||
if ($value === '') {
|
||||
return [];
|
||||
}
|
||||
return match ($kind) {
|
||||
'multi' => array_values(array_filter(
|
||||
array_map('trim', explode(',', $value)),
|
||||
static fn ($item) => $item !== ''
|
||||
)),
|
||||
'rich' => $this->extractFromRichText($value),
|
||||
default => [$value],
|
||||
};
|
||||
}
|
||||
|
||||
/**
|
||||
* 富文本 / Markdown 里的地址
|
||||
*
|
||||
* 三种写法都要抽：HTML 属性（src/href/poster）、内联样式的 url(…)、
|
||||
* Markdown 的 。只抽 src 会漏掉正文里手写的 Markdown 图片。
|
||||
*
|
||||
* @return array<int, string>
|
||||
*/
|
||||
private function extractFromRichText(string $content): array
|
||||
{
|
||||
$found = [];
|
||||
$patterns = [
|
||||
'#(?:src|href|poster|data-src)\s*=\s*[\'"]([^\'"]+)[\'"]#i',
|
||||
'#url\(\s*[\'"]?([^\'")]+)[\'"]?\s*\)#i',
|
||||
'#!\[[^\]]*\]\(\s*([^\s)]+)#',
|
||||
];
|
||||
foreach ($patterns as $pattern) {
|
||||
if (preg_match_all($pattern, $content, $matches)) {
|
||||
foreach ($matches[1] as $hit) {
|
||||
$hit = trim((string) $hit);
|
||||
if ($hit !== '') {
|
||||
$found[] = $hit;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
return array_values(array_unique($found));
|
||||
}
|
||||
|
||||
/**
|
||||
* 取指定配置的存储驱动
|
||||
*
|
||||
* oss_config_id 为 0 的是扩表之前的老记录，只能按当前启用配置去删；
|
||||
* 换过存储的站点这类记录得先同步一次把 config_id 补上，否则会删错 bucket。
|
||||
*
|
||||
* @throws Exception
|
||||
*/
|
||||
private function driverOf(int $configId): OssStorageInterface
|
||||
{
|
||||
$runtime = OssRuntimeConfigService::getInstance();
|
||||
$config = $configId > 0 ? $runtime->getConfigById($configId) : $runtime->getActiveConfig();
|
||||
return OssStorageFactory::getInstance()->make($config);
|
||||
}
|
||||
|
||||
/**
|
||||
* 0 表示根目录，其余必须是存在的文件夹
|
||||
*
|
||||
* 不校验就会把素材移进一个查不到的 folder_id，那批文件在素材库里再也点不出来
|
||||
*
|
||||
* @throws Exception
|
||||
*/
|
||||
private function assertFolder(int $folderId): int
|
||||
{
|
||||
if ($folderId <= 0) {
|
||||
return 0;
|
||||
}
|
||||
if (!FileFolderModel::where('id', $folderId)->where('deleted_at', 0)->exists()) {
|
||||
$this->utils->notFound('文件夹不存在');
|
||||
}
|
||||
return $folderId;
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user