@arcblock/crawler 1.1.6 → 1.3.0

package/lib/cjs/crawler.d.ts

@@ -1,11 +1,15 @@
-import { JobState } from './store/job';
-import { SnapshotModel } from './store/snapshot';
-export declare function createCrawlQueue(): void;
+import { Page } from '@blocklet/puppeteer';
+import { JobState, SnapshotModel } from './store';
+type PageHandler = {
+    handleScreenshot?: (page: Page, params?: JobState) => Promise<Buffer | null>;
+    handleHtml?: (page: Page, params?: JobState) => Promise<string | null>;
+};
+export declare function createCrawlQueue(queue: string, handler?: PageHandler): any;
 export declare function getDataDir(): Promise<{
     htmlDir: string;
     screenshotDir: string;
 }>;
-export declare const getPageContent: ({ url, includeScreenshot, includeHtml, width, height, quality, timeout, fullPage, headers, cookies, localStorage, }: JobState) => Promise<{
+export declare const getPageContent: ({ url, includeScreenshot, includeHtml, width, height, quality, format, timeout, waitTime, fullPage, headers, cookies, localStorage, }: JobState, handler?: PageHandler) => Promise<{
     html: string | null;
     screenshot: Uint8Array<ArrayBufferLike> | null;
     meta: {
@@ -18,4 +22,7 @@ export declare const getPageContent: ({ url, includeScreenshot, includeHtml, wid
  * @param params
  * @param callback callback when job finished
  */
+export declare function enqueue(queue: any, params: Omit<JobState, 'jobId'>, callback?: (snapshot: SnapshotModel | null) => void): Promise<string>;
 export declare function crawlUrl(params: Omit<JobState, 'jobId'>, callback?: (snapshot: SnapshotModel | null) => void): Promise<string>;
+export declare function crawlCode(params: Omit<JobState, 'jobId'>, callback?: (snapshot: SnapshotModel | null) => void): Promise<string>;
+export {};

package/lib/cjs/crawler.js

@@ -15,7 +15,9 @@ Object.defineProperty(exports, "__esModule", { value: true });
 exports.getPageContent = void 0;
 exports.createCrawlQueue = createCrawlQueue;
 exports.getDataDir = getDataDir;
+exports.enqueue = enqueue;
 exports.crawlUrl = crawlUrl;
+exports.crawlCode = crawlCode;
 const queue_1 = __importDefault(require("@abtnode/queue"));
 const sequelize_1 = __importDefault(require("@abtnode/queue/lib/store/sequelize"));
 const crypto_1 = require("crypto");
@@ -23,44 +25,44 @@ const fs_extra_1 = __importDefault(require("fs-extra"));
 const path_1 = __importDefault(require("path"));
 const config_1 = require("./config");
 const puppeteer_1 = require("./puppeteer");
+const carbon_1 = require("./services/carbon");
 const snapshot_1 = require("./services/snapshot");
-const job_1 = require("./store/job");
-const snapshot_2 = require("./store/snapshot");
+const store_1 = require("./store");
 const utils_1 = require("./utils");
 const { BaseState } = require('@abtnode/models');
-let crawlQueue;
-function createCrawlQueue() {
-    const db = new BaseState(job_1.Job);
-    crawlQueue = (0, queue_1.default)({
-        store: new sequelize_1.default(db, 'crawler'),
+// eslint-disable-next-line import/no-mutable-exports
+const crawlQueue = createCrawlQueue('urlCrawler');
+const syncQueue = createCrawlQueue('syncCrawler');
+const codeQueue = createCrawlQueue('codeCrawler', {
+    handleScreenshot: carbon_1.createCarbonImage,
+});
+function createCrawlQueue(queue, handler) {
+    const db = new BaseState(store_1.Job);
+    return (0, queue_1.default)({
+        store: new sequelize_1.default(db, queue),
         concurrency: config_1.config.concurrency,
         onJob: (job) => __awaiter(this, void 0, void 0, function* () {
             config_1.logger.info('Starting to execute crawl job', job);
-            const canCrawl = yield (0, utils_1.isAcceptCrawler)(job.url);
-            if (!canCrawl) {
-                config_1.logger.error(`failed to crawl ${job.url}, denied by robots.txt`, job);
-                const snapshot = (0, snapshot_1.convertJobToSnapshot)({
-                    job,
-                    snapshot: {
-                        status: 'failed',
-                        error: 'Denied by robots.txt',
-                    },
-                });
-                yield snapshot_2.Snapshot.upsert(snapshot);
-                return snapshot;
+            // check robots.txt
+            if (!job.ignoreRobots) {
+                const canCrawl = yield (0, utils_1.isAcceptCrawler)(job.url);
+                if (!canCrawl) {
+                    config_1.logger.error(`failed to crawl ${job.url}, denied by robots.txt`, job);
+                    const snapshot = (0, snapshot_1.convertJobToSnapshot)({
+                        job,
+                        snapshot: {
+                            status: 'failed',
+                            error: 'Denied by robots.txt',
+                        },
+                    });
+                    yield store_1.Snapshot.upsert(snapshot);
+                    return snapshot;
+                }
             }
-            // if index reach autoCloseBrowserCount, close browser
-            // try {
-            //   if (index >= autoCloseBrowserCount) {
-            //     await closeBrowser({ trimCache: false });
-            //   }
-            // } catch (error) {
-            //   logger.error('failed to close browser when queue index reached autoCloseBrowserCount:', error);
-            // }
             const formattedJob = Object.assign(Object.assign({}, job), { cookies: (config_1.config.cookies || []).concat(job.cookies || []), localStorage: (config_1.config.localStorage || []).concat(job.localStorage || []), url: (0, utils_1.formatUrl)(job.url) });
             try {
                 // get page content later
-                const result = yield (0, exports.getPageContent)(formattedJob);
+                const result = yield (0, exports.getPageContent)(formattedJob, handler);
                 if (!result || (!result.html && !result.screenshot)) {
                     config_1.logger.error(`failed to crawl ${formattedJob.url}, empty content`, formattedJob);
                     const snapshot = (0, snapshot_1.convertJobToSnapshot)({
@@ -70,25 +72,43 @@ function createCrawlQueue() {
                             error: 'Failed to crawl content',
                         },
                     });
-                    yield snapshot_2.Snapshot.upsert(snapshot);
+                    yield store_1.Snapshot.upsert(snapshot);
                     return snapshot;
                 }
-                // save html and screenshot to data dir
-                const { screenshotPath, htmlPath } = yield saveSnapshotToLocal({
-                    screenshot: result.screenshot,
-                    html: result.html,
-                });
-                // const lastModified = job.lastmodMap?.get(url) || new Date().toISOString();
-                const snapshot = (0, snapshot_1.convertJobToSnapshot)({
-                    job: formattedJob,
-                    snapshot: {
-                        status: 'success',
-                        screenshot: screenshotPath === null || screenshotPath === void 0 ? void 0 : screenshotPath.replace(config_1.config.dataDir, ''),
-                        html: htmlPath === null || htmlPath === void 0 ? void 0 : htmlPath.replace(config_1.config.dataDir, ''),
-                        meta: result.meta,
-                    },
-                });
-                yield snapshot_2.Snapshot.upsert(snapshot);
+                const snapshot = yield store_1.sequelize.transaction((txn) => __awaiter(this, void 0, void 0, function* () {
+                    // delete old snapshot
+                    if (formattedJob.replace) {
+                        try {
+                            const deletedJobIds = yield (0, snapshot_1.deleteSnapshots)({
+                                url: formattedJob.url,
+                                replace: true,
+                            }, { txn });
+                            if (deletedJobIds) {
+                                config_1.logger.info('Deleted old snapshot', { deletedJobIds });
+                            }
+                        }
+                        catch (error) {
+                            config_1.logger.error('Failed to delete old snapshot', { error, formattedJob });
+                        }
+                    }
+                    // save html and screenshot to data dir
+                    const { screenshotPath, htmlPath } = yield saveSnapshotToLocal({
+                        screenshot: result.screenshot,
+                        html: result.html,
+                        format: formattedJob.format,
+                    });
+                    const snapshot = (0, snapshot_1.convertJobToSnapshot)({
+                        job: formattedJob,
+                        snapshot: {
+                            status: 'success',
+                            screenshot: screenshotPath === null || screenshotPath === void 0 ? void 0 : screenshotPath.replace(config_1.config.dataDir, ''),
+                            html: htmlPath === null || htmlPath === void 0 ? void 0 : htmlPath.replace(config_1.config.dataDir, ''),
+                            meta: result.meta,
+                        },
+                    });
+                    yield store_1.Snapshot.upsert(snapshot, { transaction: txn });
+                    return snapshot;
+                }));
                 return snapshot;
             }
             catch (error) {
@@ -100,7 +120,7 @@ function createCrawlQueue() {
                         error: 'Internal error',
                     },
                 });
-                yield snapshot_2.Snapshot.upsert(snapshot);
+                yield store_1.Snapshot.upsert(snapshot);
                 return snapshot;
             }
         }),
@@ -116,13 +136,13 @@ function getDataDir() {
     });
 }
 function saveSnapshotToLocal(_a) {
-    return __awaiter(this, arguments, void 0, function* ({ screenshot, html }) {
+    return __awaiter(this, arguments, void 0, function* ({ screenshot, html, format = 'webp', }) {
         const { htmlDir, screenshotDir } = yield getDataDir();
         let screenshotPath = null;
         let htmlPath = null;
         if (screenshot) {
             const hash = (0, utils_1.md5)(screenshot);
-            screenshotPath = path_1.default.join(screenshotDir, `${hash}.webp`);
+            screenshotPath = path_1.default.join(screenshotDir, `${hash}.${format}`);
             config_1.logger.debug('saveSnapshotToLocal.screenshot', { screenshotPath });
             yield fs_extra_1.default.writeFile(screenshotPath, screenshot);
         }
@@ -138,7 +158,7 @@ function saveSnapshotToLocal(_a) {
         };
     });
 }
-const getPageContent = (_a) => __awaiter(void 0, [_a], void 0, function* ({ url, includeScreenshot = true, includeHtml = true, width = 1440, height = 900, quality = 80, timeout = 90 * 1000, fullPage = false, headers, cookies, localStorage, }) {
+const getPageContent = (_a, handler_1) => __awaiter(void 0, [_a, handler_1], void 0, function* ({ url, includeScreenshot = true, includeHtml = true, width = 1440, height = 900, quality = 80, format = 'webp', timeout = 90 * 1000, waitTime = 0, fullPage = false, headers, cookies, localStorage, }, handler) {
     const page = yield (0, puppeteer_1.initPage)();
     if (width && height) {
         yield page.setViewport({ width, height, deviceScaleFactor: 2 });
@@ -175,9 +195,18 @@ const getPageContent = (_a) => __awaiter(void 0, [_a], void 0, function* ({ url,
         }
         // await for networkidle0
         // https://pptr.dev/api/puppeteer.page.waitfornetworkidle
-        yield page.waitForNetworkIdle({
-            idleTime: 1.5 * 1000,
-        });
+        try {
+            yield Promise.all([
+                page.waitForNetworkIdle({
+                    idleTime: 1.5 * 1000,
+                    timeout,
+                }),
+                (0, utils_1.sleep)(waitTime),
+            ]);
+        }
+        catch (err) {
+            config_1.logger.warn(`Failed to wait for network idle in ${url}:`, err);
+        }
         // get screenshot
         if (includeScreenshot) {
             // Try to find the tallest element and set the browser to the same height
@@ -193,7 +222,9 @@ const getPageContent = (_a) => __awaiter(void 0, [_a], void 0, function* ({ url,
                 }
             }
             try {
-                screenshot = yield page.screenshot({ fullPage, quality, type: 'webp' });
+                screenshot = (handler === null || handler === void 0 ? void 0 : handler.handleScreenshot)
+                    ? yield handler.handleScreenshot(page)
+                    : yield page.screenshot({ fullPage, quality, type: format });
             }
             catch (err) {
                 config_1.logger.error('Failed to get screenshot:', err);
@@ -221,12 +252,12 @@ const getPageContent = (_a) => __awaiter(void 0, [_a], void 0, function* ({ url,
             // check if the page is an error page
             const isErrorPage = ['<h2>Unexpected Application Error!</h2>', 'Current route occurred an error'].some((errorHtml) => data.html.includes(errorHtml));
             if (isErrorPage) {
-                throw new Error('Page is an error page');
+                throw new Error(`${url} is an error page`);
             }
             meta.title = data.title;
             meta.description = data.description;
             if (includeHtml) {
-                html = data.html;
+                html = (handler === null || handler === void 0 ? void 0 : handler.handleHtml) ? yield handler.handleHtml(page) : data.html;
             }
         }
         catch (err) {
@@ -254,17 +285,17 @@ exports.getPageContent = getPageContent;
  * @param callback callback when job finished
  */
 // eslint-disable-next-line require-await
-function crawlUrl(params, callback) {
+function enqueue(queue, params, callback) {
     return __awaiter(this, void 0, void 0, function* () {
         // skip duplicate job
-        const existsJob = yield job_1.Job.isExists(params);
-        if (existsJob) {
+        const existsJob = yield store_1.Job.isExists(params);
+        if (existsJob && !params.sync) {
             config_1.logger.info(`Crawl job already exists for ${params.url}, skip`);
             return existsJob.id;
         }
         config_1.logger.info('enqueue crawl job', params);
         const jobId = (0, crypto_1.randomUUID)();
-        const job = crawlQueue.push(Object.assign(Object.assign({}, params), { id: jobId }));
+        const job = queue.push(Object.assign(Object.assign({}, params), { id: jobId }));
         job.on('finished', (_a) => __awaiter(this, [_a], void 0, function* ({ result }) {
             config_1.logger.info(`Crawl completed ${params.url}, status: ${result ? 'success' : 'failed'}`, { job: params, result });
             callback === null || callback === void 0 ? void 0 : callback(result ? yield (0, snapshot_1.formatSnapshot)(result) : null);
@@ -276,3 +307,9 @@ function crawlUrl(params, callback) {
         return jobId;
     });
 }
+function crawlUrl(params, callback) {
+    return enqueue(params.sync ? syncQueue : crawlQueue, params, callback);
+}
+function crawlCode(params, callback) {
+    return enqueue(codeQueue, Object.assign({ ignoreRobots: true, includeHtml: false, includeScreenshot: true }, params), callback);
+}

package/lib/cjs/index.d.ts

@@ -3,4 +3,5 @@ export * from './crawler';
 export * from './site';
 export * from './services/snapshot';
 export * as utils from './utils';
+export { migrate } from './store/migrate';
 export declare function initCrawler(params: Pick<Config, 'puppeteerPath' | 'siteCron' | 'cookies' | 'localStorage' | 'concurrency'>): Promise<void>;

package/lib/cjs/index.js

@@ -48,28 +48,26 @@ var __importDefault = (this && this.__importDefault) || function (mod) {
     return (mod && mod.__esModule) ? mod : { "default": mod };
 };
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.utils = void 0;
+exports.migrate = exports.utils = void 0;
 exports.initCrawler = initCrawler;
 /* eslint-disable @typescript-eslint/indent */
 const merge_1 = __importDefault(require("lodash/merge"));
 const config_1 = require("./config");
-const crawler_1 = require("./crawler");
 const cron_1 = require("./cron");
 const puppeteer_1 = require("./puppeteer");
-const store_1 = require("./store");
 __exportStar(require("./crawler"), exports);
 __exportStar(require("./site"), exports);
 __exportStar(require("./services/snapshot"), exports);
 exports.utils = __importStar(require("./utils"));
+var migrate_1 = require("./store/migrate");
+Object.defineProperty(exports, "migrate", { enumerable: true, get: function () { return migrate_1.migrate; } });
 function initCrawler(params) {
     return __awaiter(this, void 0, void 0, function* () {
         var _a;
         (0, merge_1.default)(config_1.config, params);
         config_1.logger.info('Init crawler', { params, config: config_1.config });
         try {
-            yield (0, store_1.initDatabase)();
             yield (0, puppeteer_1.ensureBrowser)();
-            yield (0, crawler_1.createCrawlQueue)();
             if ((_a = config_1.config.siteCron) === null || _a === void 0 ? void 0 : _a.enabled) {
                 yield (0, cron_1.initCron)();
             }

package/lib/cjs/services/carbon.d.ts

@@ -0,0 +1,3 @@
+import { Page } from '@blocklet/puppeteer';
+import { JobState } from '../store';
+export declare function createCarbonImage(page: Page, params?: JobState): Promise<Buffer<ArrayBuffer>>;

package/lib/cjs/services/carbon.js

@@ -0,0 +1,87 @@
+"use strict";
+var __awaiter = (this && this.__awaiter) || function (thisArg, _arguments, P, generator) {
+    function adopt(value) { return value instanceof P ? value : new P(function (resolve) { resolve(value); }); }
+    return new (P || (P = Promise))(function (resolve, reject) {
+        function fulfilled(value) { try { step(generator.next(value)); } catch (e) { reject(e); } }
+        function rejected(value) { try { step(generator["throw"](value)); } catch (e) { reject(e); } }
+        function step(result) { result.done ? resolve(result.value) : adopt(result.value).then(fulfilled, rejected); }
+        step((generator = generator.apply(thisArg, _arguments || [])).next());
+    });
+};
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.createCarbonImage = createCarbonImage;
+const config_1 = require("../config");
+// TODO expose local version of dom-to-image
+const DOM_TO_IMAGE_URL = 'https://unpkg.com/dom-to-image@2.6.0/dist/dom-to-image.min.js';
+function createCarbonImage(page, params) {
+    return __awaiter(this, void 0, void 0, function* () {
+        try {
+            yield page.addScriptTag({ url: DOM_TO_IMAGE_URL });
+            yield page.waitForSelector('.export-container', { visible: true, timeout: (params === null || params === void 0 ? void 0 : params.timeout) || 120 });
+            const targetElement = yield page.$('.export-container');
+            const format = (params === null || params === void 0 ? void 0 : params.format) || 'png';
+            const dataUrl = yield page.evaluate((target = document, imageFormat = 'png') => {
+                const query = new URLSearchParams(document.location.search);
+                const EXPORT_SIZES_HASH = {
+                    '1x': '1',
+                    '2x': '2',
+                    '4x': '4',
+                };
+                const exportSize = EXPORT_SIZES_HASH[query.get('es')] || '2';
+                if (!target) {
+                    throw new Error('Target element not found');
+                }
+                target.querySelectorAll('span[role="presentation"]').forEach((node) => {
+                    var _a;
+                    const el = node;
+                    if (el && el.innerText && el.innerText.match(/%[A-Fa-f0-9]{2}/)) {
+                        (_a = el.innerText.match(/%[A-Fa-f0-9]{2}/g)) === null || _a === void 0 ? void 0 : _a.forEach((t) => {
+                            el.innerHTML = el.innerHTML.replace(t, encodeURIComponent(t));
+                        });
+                    }
+                });
+                const width = target.offsetWidth * exportSize;
+                const height = query.get('si') === 'true'
+                    ? target.offsetWidth * exportSize
+                    : target.offsetHeight * exportSize;
+                const config = {
+                    style: {
+                        transform: `scale(${exportSize})`,
+                        'transform-origin': 'center',
+                        background: query.get('si') ? query.get('bg') : 'none',
+                    },
+                    filter: (n) => {
+                        if (n.className) {
+                            return String(n.className).indexOf('eliminateOnRender') < 0;
+                        }
+                        return true;
+                    },
+                    width,
+                    height,
+                };
+                switch (imageFormat) {
+                    case 'jpeg':
+                        // @ts-ignore: domtoimage is injected by addScriptTag
+                        return domtoimage.toJpeg(target, config);
+                    case 'webp':
+                        // dom-to-image doesn't support webp directly, fall back to png
+                        // @ts-ignore: domtoimage is injected by addScriptTag
+                        return domtoimage.toPng(target, config);
+                    case 'png':
+                    default:
+                        // @ts-ignore: domtoimage is injected by addScriptTag
+                        return domtoimage.toPng(target, config);
+                }
+            }, targetElement, format);
+            const base64Data = dataUrl.split(',')[1];
+            if (!base64Data) {
+                throw new Error('Failed to extract base64 data from image');
+            }
+            return Buffer.from(base64Data, 'base64');
+        }
+        catch (e) {
+            config_1.logger.error('failed to crawl from carbon', { error: e });
+            throw e;
+        }
+    });
+}

package/lib/cjs/services/snapshot.d.ts

@@ -1,5 +1,5 @@
-import { JobState } from '../store/job';
-import { SnapshotModel } from '../store/snapshot';
+import { Transaction, WhereOptions } from '@sequelize/core';
+import { JobState, SnapshotModel } from '../store';
 export declare function convertJobToSnapshot({ job, snapshot }: {
     job: JobState;
     snapshot?: Partial<SnapshotModel>;
@@ -10,3 +10,6 @@ export declare function formatSnapshot(snapshot: SnapshotModel, columns?: Array<
  */
 export declare function getSnapshot(jobId: string): Promise<SnapshotModel | null>;
 export declare function getLatestSnapshot(url: string): Promise<SnapshotModel | null>;
+export declare function deleteSnapshots(where: WhereOptions<SnapshotModel>, { txn }?: {
+    txn?: Transaction;
+}): Promise<string[]>;

package/lib/cjs/services/snapshot.js

@@ -16,17 +16,17 @@ exports.convertJobToSnapshot = convertJobToSnapshot;
 exports.formatSnapshot = formatSnapshot;
 exports.getSnapshot = getSnapshot;
 exports.getLatestSnapshot = getLatestSnapshot;
+exports.deleteSnapshots = deleteSnapshots;
 const cloneDeep_1 = __importDefault(require("lodash/cloneDeep"));
 const pick_1 = __importDefault(require("lodash/pick"));
 const promises_1 = __importDefault(require("node:fs/promises"));
 const node_path_1 = __importDefault(require("node:path"));
 const ufo_1 = require("ufo");
 const config_1 = require("../config");
-const job_1 = require("../store/job");
-const snapshot_1 = require("../store/snapshot");
+const store_1 = require("../store");
 const utils_1 = require("../utils");
 function convertJobToSnapshot({ job, snapshot }) {
-    return Object.assign({ jobId: job.jobId || job.id, url: job.url, lastModified: job.lastModified || new Date().toISOString(), options: {
+    return Object.assign({ jobId: job.jobId || job.id, url: job.url, lastModified: job.lastModified || new Date().toISOString(), replace: job.replace, options: {
             width: job.width,
             height: job.height,
             includeScreenshot: job.includeScreenshot,
@@ -64,11 +64,11 @@ function formatSnapshot(snapshot, columns) {
  */
 function getSnapshot(jobId) {
     return __awaiter(this, void 0, void 0, function* () {
-        const snapshot = yield snapshot_1.Snapshot.findSnapshot({ where: { jobId } });
+        const snapshot = yield store_1.Snapshot.findSnapshot({ where: { jobId } });
         if (snapshot) {
             return formatSnapshot(snapshot);
         }
-        const job = yield job_1.Job.findJob({ id: jobId });
+        const job = yield store_1.Job.findJob({ id: jobId });
         if (job) {
             return {
                 jobId,
@@ -80,12 +80,42 @@ function getSnapshot(jobId) {
 }
 function getLatestSnapshot(url) {
     return __awaiter(this, void 0, void 0, function* () {
-        const snapshot = yield snapshot_1.Snapshot.findSnapshot({
+        const snapshot = yield store_1.Snapshot.findSnapshot({
             where: {
                 url: (0, utils_1.formatUrl)(url),
                 status: 'success',
             },
+            order: [
+                ['lastModified', 'DESC'],
+                ['updatedAt', 'DESC'],
+            ],
         });
         return snapshot ? formatSnapshot(snapshot) : null;
     });
 }
+function deleteSnapshots(where_1) {
+    return __awaiter(this, arguments, void 0, function* (where, { txn } = {}) {
+        const snapshots = yield store_1.Snapshot.findAll({
+            where,
+            order: [
+                ['lastModified', 'DESC'],
+                ['updatedAt', 'DESC'],
+            ],
+        });
+        const jobIds = yield Promise.all(snapshots.map((snapshot) => __awaiter(this, void 0, void 0, function* () {
+            try {
+                yield Promise.all([
+                    snapshot.html && promises_1.default.unlink(node_path_1.default.join(config_1.config.dataDir, snapshot.html)),
+                    snapshot.screenshot && promises_1.default.unlink(node_path_1.default.join(config_1.config.dataDir, snapshot.screenshot)),
+                ]);
+                yield snapshot.destroy({ transaction: txn });
+                return snapshot.jobId;
+            }
+            catch (error) {
+                config_1.logger.error('Failed to delete snapshot', { error, snapshot });
+                throw error;
+            }
+        })));
+        return jobIds.filter(Boolean);
+    });
+}

package/lib/cjs/site.d.ts

@@ -1,2 +1,2 @@
 import { Site } from './config';
-export declare const crawlSite: ({ url, pathname, interval }: Site) => Promise<(string | null)[]>;
+export declare const crawlSite: ({ url, pathname, interval }: Site) => Promise<(`${string}-${string}-${string}-${string}-${string}` | null)[]>;

package/lib/cjs/site.js

@@ -14,12 +14,14 @@ var __importDefault = (this && this.__importDefault) || function (mod) {
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.crawlSite = void 0;
 const uniq_1 = __importDefault(require("lodash/uniq"));
+const node_crypto_1 = require("node:crypto");
 const p_map_1 = __importDefault(require("p-map"));
 const config_1 = require("./config");
 const crawler_1 = require("./crawler");
-const snapshot_1 = require("./store/snapshot");
+const store_1 = require("./store");
 const utils_1 = require("./utils");
 const crawlBlockletRunningMap = new Map();
+const crawlQueue = (0, crawler_1.createCrawlQueue)('cronJobs');
 function parseSitemapUrl(sitemapItem) {
     var _a;
     const links = ((_a = sitemapItem.links) === null || _a === void 0 ? void 0 : _a.map((item) => item.url)) || [];
@@ -48,7 +50,7 @@ const crawlSite = (_a) => __awaiter(void 0, [_a], void 0, function* ({ url, path
     try {
         const jobIds = yield (0, p_map_1.default)(sitemapItems, (_a) => __awaiter(void 0, [_a], void 0, function* ({ url, sitemapItem }) {
             processCount++;
-            const snapshot = yield snapshot_1.Snapshot.findOne({ where: { url: (0, utils_1.formatUrl)(url) } });
+            const snapshot = yield store_1.Snapshot.findOne({ where: { url: (0, utils_1.formatUrl)(url) } });
             if (snapshot === null || snapshot === void 0 ? void 0 : snapshot.lastModified) {
                 const lastModified = new Date(snapshot.lastModified);
                 // skip if snapshot lastModified is greater than sitemap lastmod
@@ -67,12 +69,16 @@ const crawlSite = (_a) => __awaiter(void 0, [_a], void 0, function* ({ url, path
                 url,
             });
             crawlCount++;
-            return (0, crawler_1.crawlUrl)({
+            const jobId = (0, node_crypto_1.randomUUID)();
+            crawlQueue.push({
+                id: jobId,
                 url,
                 lastModified: sitemapItem.lastmod,
                 includeScreenshot: false,
                 includeHtml: true,
+                replace: true,
             });
+            return jobId;
         }), { concurrency: ((_b = config_1.config.siteCron) === null || _b === void 0 ? void 0 : _b.concurrency) || 30 });
         config_1.logger.info('Enqueued jobs from sitemap finished', {
             url,

package/lib/cjs/store/index.d.ts

@@ -1,3 +1,6 @@
 import { Sequelize } from '@sequelize/core';
 import { SqliteDialect } from '@sequelize/sqlite3';
-export declare function initDatabase(): Promise<Sequelize<SqliteDialect>>;
+declare const sequelize: Sequelize<SqliteDialect>;
+export { sequelize };
+export * from './job';
+export * from './snapshot';