@crawlee/browser 4.0.0-beta.11 → 4.0.0-beta.110

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -1,8 +1,21 @@
1
- import { BasicCrawler, BLOCKED_STATUS_CODES as DEFAULT_BLOCKED_STATUS_CODES, Configuration, ContextPipeline, cookieStringToToughCookie, enqueueLinks, EVENT_SESSION_RETIRED, handleRequestTimeout, RequestState, resolveBaseUrlForEnqueueLinksFiltering, SessionError, tryAbsoluteURL, validators, } from '@crawlee/basic';
2
- import { BrowserPool } from '@crawlee/browser-pool';
1
+ import { BasicCrawler, browserPoolCookieToToughCookie, ContextPipeline, cookieStringToToughCookie, enqueueLinks, NavigationSkippedError, OwnedOrInjected, remainingNavigationWindowMillis, RequestState, resolveBaseUrlForEnqueueLinksFiltering, SessionError, toughCookieToBrowserPoolCookie, tryAbsoluteURL, validators, } from '@crawlee/basic';
2
+ import { BrowserPool, RemoteBrowserPool } from '@crawlee/browser-pool';
3
3
  import { CLOUDFLARE_RETRY_CSS_SELECTORS, RETRY_CSS_SELECTORS, sleep } from '@crawlee/utils';
4
4
  import ow from 'ow';
5
- import { tryCancel } from '@apify/timeout';
5
+ import { addTimeoutToPromise, TimeoutError, tryCancel } from '@apify/timeout';
6
+ const COOKIES_BEFORE_HOOKS = Symbol('cookiesBeforeHooks');
7
+ const readContextField = (ctx, key) => ctx[key];
8
+ /**
9
+ * Whether an error thrown by `page.goto()` is a navigation timeout - either our own {@link TimeoutError}
10
+ * or the driver's, which Playwright/Puppeteer report with their own class and a `Timeout ... exceeded` message
11
+ * naming the raw millisecond value rather than the configured window.
12
+ */
13
+ function isNavigationTimeoutError(error) {
14
+ return (error instanceof TimeoutError ||
15
+ error?.name === 'TimeoutError' ||
16
+ error?.constructor?.name === 'TimeoutError' ||
17
+ /timeout.*exceeded/i.test(error?.message ?? ''));
18
+ }
6
19
  /**
7
20
  * Provides a simple framework for parallel crawling of web pages
8
21
  * using headless browsers with [Puppeteer](https://github.com/puppeteer/puppeteer)
@@ -15,51 +28,51 @@ import { tryCancel } from '@apify/timeout';
15
28
  * If the target website doesn't need JavaScript, we should consider using the {@link CheerioCrawler},
16
29
  * which downloads the pages using raw HTTP requests and is about 10x faster.
17
30
  *
18
- * The source URLs are represented by the {@link Request} objects that are fed from the {@link RequestList} or {@link RequestQueue} instances
19
- * provided by the {@link BrowserCrawlerOptions.requestList|`requestList`} or {@link BrowserCrawlerOptions.requestQueue|`requestQueue`}
20
- * constructor options, respectively. If neither `requestList` nor `requestQueue` options are provided,
31
+ * The source URLs are represented by the {@link Request} objects that are fed from the
32
+ * {@link IRequestManager|request manager} provided via the {@link BrowserCrawlerOptions.requestManager|`requestManager`}
33
+ * constructor option (a {@link RequestQueue} is itself a request manager). If no `requestManager` is provided,
21
34
  * the crawler will open the default request queue either when the {@link BrowserCrawler.addRequests|`crawler.addRequests()`} function is called,
22
35
  * or if `requests` parameter (representing the initial requests) of the {@link BrowserCrawler.run|`crawler.run()`} function is provided.
23
36
  *
24
- * If both {@link BrowserCrawlerOptions.requestList|`requestList`} and {@link BrowserCrawlerOptions.requestQueue|`requestQueue`} options are used,
25
- * the instance first processes URLs from the {@link RequestList} and automatically enqueues all of them
26
- * to the {@link RequestQueue} before it starts their processing. This ensures that a single URL is not crawled multiple times.
37
+ * To read from a read-only source such as a {@link RequestList} while still being able to enqueue new requests,
38
+ * combine it with a queue into a {@link RequestManagerTandem} via {@link IRequestLoader.toTandem|`requestLoader.toTandem()`}
39
+ * and pass the result as `requestManager`.
40
+ *
41
+ * > The {@link BrowserCrawlerOptions.requestList|`requestList`} and {@link BrowserCrawlerOptions.requestQueue|`requestQueue`}
42
+ * > options are deprecated; they are still accepted and folded into a single `requestManager` for back-compat.
27
43
  *
28
44
  * The crawler finishes when there are no more {@link Request} objects to crawl.
29
45
  *
30
46
  * `BrowserCrawler` opens a new browser page (i.e. tab or window) for each {@link Request} object to crawl
31
47
  * and then calls the function provided by user as the {@link BrowserCrawlerOptions.requestHandler|`requestHandler`} option.
32
48
  *
33
- * New pages are only opened when there is enough free CPU and memory available,
34
- * using the functionality provided by the {@link AutoscaledPool} class.
35
- * All {@link AutoscaledPool} configuration options can be passed to the {@link BrowserCrawlerOptions.autoscaledPoolOptions|`autoscaledPoolOptions`}
36
- * parameter of the `BrowserCrawler` constructor.
37
- * For user convenience, the {@link AutoscaledPoolOptions.minConcurrency|`minConcurrency`} and
38
- * {@link AutoscaledPoolOptions.maxConcurrency|`maxConcurrency`} options of the
39
- * underlying {@link AutoscaledPool} constructor are available directly in the `BrowserCrawler` constructor.
49
+ * New pages are only opened when there is enough free CPU and memory available, as judged by the crawler's
50
+ * {@link ConcurrencySystem}.
51
+ * Concurrency is tuned via the `minConcurrency`, `maxConcurrency` and `maxRequestsPerMinute` options of the
52
+ * `BrowserCrawler` constructor, or, for finer control, by injecting a pre-configured
53
+ * {@link ConcurrencySystem|`concurrencySystem`}.
40
54
  *
41
55
  * > *NOTE:* the pool of browser instances is internally managed by the {@link BrowserPool} class.
42
56
  *
43
57
  * @category Crawlers
44
58
  */
45
59
  export class BrowserCrawler extends BasicCrawler {
46
- config;
47
- /**
48
- * A reference to the underlying {@link ProxyConfiguration} class that manages the crawler's proxies.
49
- * Only available if used by the crawler.
50
- */
51
- proxyConfiguration;
60
+ /** Backs the {@link BrowserCrawler.browserPool|`browserPool`} getter. */
61
+ #browserPoolDep;
52
62
  /**
53
- * A reference to the underlying {@link BrowserPool} class that manages the crawler's browsers.
63
+ * A reference to the underlying browser pool that manages the crawler's browsers. Typed as
64
+ * {@link IBrowserPool} so custom implementations can be plugged in via the `browserPool` constructor option.
54
65
  */
55
- browserPool;
66
+ get browserPool() {
67
+ return this.#browserPoolDep.value;
68
+ }
56
69
  launchContext;
57
70
  ignoreShadowRoots;
58
71
  ignoreIframes;
59
- navigationTimeoutMillis;
60
- preNavigationHooks;
61
- postNavigationHooks;
62
- persistCookiesPerSession;
72
+ #navigationTimeoutMillis;
73
+ #preNavigationHooks;
74
+ #postNavigationHooks;
75
+ #saveResponseCookies;
63
76
  static optionsShape = {
64
77
  ...BasicCrawler.optionsShape,
65
78
  navigationTimeoutSecs: ow.optional.number.greaterThan(0),
@@ -67,68 +80,107 @@ export class BrowserCrawler extends BasicCrawler {
67
80
  postNavigationHooks: ow.optional.array,
68
81
  launchContext: ow.optional.object,
69
82
  headless: ow.optional.any(ow.boolean, ow.string),
70
- browserPoolOptions: ow.object,
71
- sessionPoolOptions: ow.optional.object,
72
- persistCookiesPerSession: ow.optional.boolean,
73
- useSessionPool: ow.optional.boolean,
83
+ browserPool: ow.optional.object.validate(validators.browserPool),
84
+ remoteBrowser: ow.optional.object,
85
+ browserPoolOptions: ow.optional.object,
86
+ saveResponseCookies: ow.optional.boolean,
74
87
  proxyConfiguration: ow.optional.object.validate(validators.proxyConfiguration),
75
88
  };
76
89
  /**
77
90
  * All `BrowserCrawler` parameters are passed via an options object.
78
91
  */
79
- constructor(options, config = Configuration.getGlobalConfig()) {
92
+ constructor(options) {
80
93
  ow(options, 'BrowserCrawlerOptions', ow.object.exactShape(BrowserCrawler.optionsShape));
81
- const { navigationTimeoutSecs = 60, persistCookiesPerSession, proxyConfiguration, launchContext = {}, browserPoolOptions, preNavigationHooks = [], postNavigationHooks = [], headless, ignoreIframes = false, ignoreShadowRoots = false, contextPipelineBuilder, extendContext, ...basicCrawlerOptions } = options;
94
+ const { navigationTimeoutSecs = 60, saveResponseCookies = true, launchContext = {}, browserPool, remoteBrowser, browserPoolOptions, preNavigationHooks = [], postNavigationHooks = [], headless, ignoreIframes = false, ignoreShadowRoots = false, contextPipelineBuilder, extendContext, ...basicCrawlerOptions } = options;
95
+ const skipGuard = (action) => ({
96
+ action: async (ctx) => (ctx.request.skipNavigation ? {} : ((await action(ctx)) ?? {})),
97
+ });
82
98
  super({
83
99
  ...basicCrawlerOptions,
84
- contextPipelineBuilder: () => contextPipelineBuilder()
85
- .compose({ action: this.performNavigation.bind(this) })
86
- .compose({ action: this.handleBlockedRequestByContent.bind(this) })
87
- .compose({ action: this.restoreRequestState.bind(this) }),
88
- extendContext: extendContext,
89
- }, config);
90
- this.config = config;
91
- // Cookies should be persisted per session only if session pool is used
92
- if (!this.useSessionPool && persistCookiesPerSession) {
93
- throw new Error('You cannot use "persistCookiesPerSession" without "useSessionPool" set to true.');
94
- }
100
+ contextPipelineBuilder: () => {
101
+ // A single navigation window covers the pre-navigation hooks, the navigation, and the
102
+ // post-navigation hooks: the whole phase shares one `navigationTimeoutSecs` budget, so a slow
103
+ // hook eats into the same window the navigation uses. The navigation itself is bounded by
104
+ // capping its `gotoOptions.timeout` to the remaining budget.
105
+ const windowGuard = (step) => skipGuard(async (ctx) => {
106
+ const remaining = remainingNavigationWindowMillis(ctx, this.#navigationTimeoutMillis);
107
+ if (remaining <= 0) {
108
+ throw new TimeoutError(`Navigation timed out after ${this.#navigationTimeoutMillis / 1000} seconds.`);
109
+ }
110
+ return addTimeoutToPromise(async () => step(ctx), remaining, `Navigation timed out after ${this.#navigationTimeoutMillis / 1000} seconds.`);
111
+ });
112
+ let pipeline = contextPipelineBuilder().compose({ action: this.prepareNavigation.bind(this) });
113
+ for (const hook of this.#preNavigationHooks) {
114
+ pipeline = pipeline.compose(windowGuard(hook));
115
+ }
116
+ pipeline = pipeline.compose(skipGuard(this.navigate.bind(this)));
117
+ for (const hook of this.#postNavigationHooks) {
118
+ pipeline = pipeline.compose(windowGuard(hook));
119
+ }
120
+ return pipeline
121
+ .compose(skipGuard(this.finalizeNavigation.bind(this)))
122
+ .compose({ action: this.handleBlockedRequestByContent.bind(this) })
123
+ .compose({ action: this.restoreRequestState.bind(this) });
124
+ },
125
+ extendContext,
126
+ });
95
127
  this.launchContext = launchContext;
96
- this.navigationTimeoutMillis = navigationTimeoutSecs * 1000;
97
- this.proxyConfiguration = proxyConfiguration;
98
- this.preNavigationHooks = preNavigationHooks;
99
- this.postNavigationHooks = postNavigationHooks;
128
+ this.#navigationTimeoutMillis = navigationTimeoutSecs * 1000;
129
+ // The public option hooks are extension-aware; internal storage uses the base context type
130
+ // (the pipeline composes hooks against the concrete context, which does not statically carry
131
+ // `ContextExtension`). The extension members are present at runtime regardless.
132
+ this.#preNavigationHooks = preNavigationHooks;
133
+ this.#postNavigationHooks = postNavigationHooks;
100
134
  this.ignoreIframes = ignoreIframes;
101
135
  this.ignoreShadowRoots = ignoreShadowRoots;
102
136
  if (headless != null) {
103
137
  this.launchContext.launchOptions ??= {};
104
138
  this.launchContext.launchOptions.headless = headless;
105
139
  }
106
- if (this.useSessionPool) {
107
- this.persistCookiesPerSession = persistCookiesPerSession !== undefined ? persistCookiesPerSession : true;
108
- }
109
- else {
110
- this.persistCookiesPerSession = false;
111
- }
112
- if (launchContext?.userAgent) {
113
- if (browserPoolOptions.useFingerprints)
114
- this.log.info('Custom user agent provided, disabling automatic browser fingerprint injection!');
115
- browserPoolOptions.useFingerprints = false;
116
- }
117
- const { preLaunchHooks = [], postLaunchHooks = [], ...rest } = browserPoolOptions;
118
- this.browserPool = new BrowserPool({
119
- ...rest,
120
- preLaunchHooks: [this._extendLaunchContext.bind(this), ...preLaunchHooks],
121
- postLaunchHooks: [this._maybeAddSessionRetiredListener.bind(this), ...postLaunchHooks],
140
+ this.#saveResponseCookies = saveResponseCookies;
141
+ // `browserPool` wins over `remoteBrowser` a passed-in pool is used as-is (borrowed), the sugar is ignored.
142
+ // The default is only built when no pool was injected, so all the option/launchContext fiddling below stays
143
+ // inside the factory.
144
+ this.#browserPoolDep = OwnedOrInjected.resolve(browserPool, () => {
145
+ const resolvedBrowserPoolOptions = browserPoolOptions ?? {};
146
+ if (launchContext?.userAgent) {
147
+ if (resolvedBrowserPoolOptions.useFingerprints)
148
+ this.log.info('Custom user agent provided, disabling automatic browser fingerprint injection!');
149
+ resolvedBrowserPoolOptions.useFingerprints = false;
150
+ }
151
+ if (remoteBrowser) {
152
+ // The crawler already built the right plugin for its browser — hand it to a RemoteBrowserPool so the
153
+ // remote connection is always for the matching browser (no plugin to construct, no way to mismatch).
154
+ const { browserPlugins, ...remoteBrowserPoolOptions } = resolvedBrowserPoolOptions;
155
+ return new RemoteBrowserPool({
156
+ browserPlugins: browserPlugins,
157
+ ...remoteBrowser,
158
+ browserPoolOptions: remoteBrowserPoolOptions,
159
+ });
160
+ }
161
+ // Double cast: `BrowserPool` implements `IBrowserPool<PageReturn>`, where `PageReturn` is derived from the
162
+ // plugin/controller generics and doesn't overlap with the crawler's free `Page` type param, so TS won't
163
+ // narrow it directly. The concrete pool does satisfy the `Page`/`destroy` contract at runtime — this is the
164
+ // long-standing `Page` variance gap, not a `destroy`-related hole.
165
+ return new BrowserPool({
166
+ ...resolvedBrowserPoolOptions,
167
+ });
122
168
  });
123
169
  }
170
+ getNavigationTimeoutMillis() {
171
+ return this.#navigationTimeoutMillis;
172
+ }
124
173
  buildContextPipeline() {
125
174
  return ContextPipeline.create().compose({
126
175
  action: this.preparePage.bind(this),
127
176
  cleanup: async (context) => {
128
177
  context.registerDeferredCleanup(async () => {
129
- await context.page
130
- .close()
131
- .catch((error) => this.log.debug('Error while closing page', { error }));
178
+ const error = !context.session.isUsable()
179
+ ? new SessionError('Session is no longer usable')
180
+ : undefined;
181
+ await this.browserPool
182
+ .closePage(context.page, { error })
183
+ .catch((closeError) => this.log.debug('Error while closing page', { error: closeError }));
132
184
  });
133
185
  },
134
186
  });
@@ -142,12 +194,6 @@ export class BrowserCrawler extends BasicCrawler {
142
194
  }
143
195
  async isRequestBlocked(crawlingContext) {
144
196
  const { page, response } = crawlingContext;
145
- const blockedStatusCodes =
146
- // eslint-disable-next-line dot-notation
147
- (this.sessionPool?.['blockedStatusCodes'].length ?? 0) > 0
148
- ? // eslint-disable-next-line dot-notation
149
- this.sessionPool['blockedStatusCodes']
150
- : DEFAULT_BLOCKED_STATUS_CODES;
151
197
  // Cloudflare specific heuristic - wait 5 seconds if we get a 403 for the JS challenge to load / resolve.
152
198
  if ((await this.containsSelectors(page, CLOUDFLARE_RETRY_CSS_SELECTORS)) && response?.status() === 403) {
153
199
  await sleep(5000);
@@ -158,117 +204,158 @@ export class BrowserCrawler extends BasicCrawler {
158
204
  return `Cloudflare challenge failed, found selectors: ${foundSelectors.join(', ')}`;
159
205
  }
160
206
  const foundSelectors = await this.containsSelectors(page, RETRY_CSS_SELECTORS);
161
- const blockedStatusCode = blockedStatusCodes.find((x) => x === (response?.status() ?? 0));
207
+ const statusCode = response?.status() ?? 0;
162
208
  if (foundSelectors)
163
209
  return `Found selectors: ${foundSelectors.join(', ')}`;
164
- if (blockedStatusCode)
165
- return `Received blocked status code: ${blockedStatusCode}`;
210
+ if (this.blockedStatusCodes.has(statusCode))
211
+ return `Received blocked status code: ${statusCode}`;
166
212
  return false;
167
213
  }
168
214
  async preparePage(crawlingContext) {
169
- const newPageOptions = {
215
+ const page = await this.browserPool.newPage({
170
216
  id: crawlingContext.id,
171
- };
172
- const useIncognitoPages = this.launchContext?.useIncognitoPages;
173
- if (this.proxyConfiguration) {
174
- const { session } = crawlingContext;
175
- const proxyInfo = await this.proxyConfiguration.newProxyInfo(session?.id, {
176
- request: crawlingContext.request,
177
- });
178
- crawlingContext.proxyInfo = proxyInfo;
179
- newPageOptions.proxyUrl = proxyInfo?.url;
180
- newPageOptions.proxyTier = proxyInfo?.proxyTier;
181
- if (this.proxyConfiguration.isManInTheMiddle) {
182
- /**
183
- * @see https://playwright.dev/docs/api/class-browser/#browser-new-context
184
- * @see https://github.com/puppeteer/puppeteer/blob/main/docs/api.md
185
- */
186
- newPageOptions.pageOptions = {
187
- ignoreHTTPSErrors: true,
188
- acceptInsecureCerts: true,
189
- };
190
- }
191
- }
192
- const page = (await this.browserPool.newPage(newPageOptions));
217
+ session: crawlingContext.session,
218
+ });
193
219
  tryCancel();
194
- const browserControllerInstance = this.browserPool.getBrowserControllerByPage(page);
220
+ const contextEnqueueLinks = crawlingContext.enqueueLinks;
195
221
  return {
196
222
  page,
197
223
  get response() {
198
224
  throw new Error("The `response` property is not available. This might mean that you're trying to access it before navigation or that navigation resulted in `null` (this should only happen with `about:` URLs)");
199
225
  },
200
- browserController: browserControllerInstance,
201
- session: useIncognitoPages
202
- ? crawlingContext.session
203
- : browserControllerInstance.launchContext.session,
204
- proxyInfo: crawlingContext.proxyInfo ?? browserControllerInstance.launchContext.proxyInfo,
226
+ get gotoOptions() {
227
+ throw new Error('The `gotoOptions` property is not available until `prepareNavigation` runs.');
228
+ },
205
229
  enqueueLinks: async (enqueueOptions = {}) => {
206
- return browserCrawlerEnqueueLinks({
207
- options: enqueueOptions,
230
+ return (await browserCrawlerEnqueueLinks({
231
+ options: {
232
+ ...enqueueOptions,
233
+ limit: await this.calculateEnqueuedRequestLimit(enqueueOptions?.limit),
234
+ },
208
235
  page,
209
- requestQueue: await this.getRequestQueue(),
236
+ requestManager: await this.getRequestManager(),
210
237
  robotsTxtFile: await this.getRobotsTxtFileForUrl(crawlingContext.request.url),
211
- onSkippedRequest: this.onSkippedRequest,
238
+ onSkippedRequest: this.handleSkippedRequest,
212
239
  originalRequestUrl: crawlingContext.request.url,
213
240
  finalRequestUrl: crawlingContext.request.loadedUrl,
214
- });
241
+ enqueueLinks: contextEnqueueLinks,
242
+ })); // TODO make this type safe
215
243
  },
216
244
  };
217
245
  }
218
- async performNavigation(crawlingContext) {
246
+ async prepareNavigation(crawlingContext) {
219
247
  if (crawlingContext.request.skipNavigation) {
220
248
  return {
221
249
  request: new Proxy(crawlingContext.request, {
222
250
  get(target, propertyName, receiver) {
223
251
  if (propertyName === 'loadedUrl') {
224
- throw new Error('The `request.loadedUrl` property is not available - `skipNavigation` was used');
252
+ throw new NavigationSkippedError('The `request.loadedUrl` property is not available - `skipNavigation` was used');
225
253
  }
226
254
  return Reflect.get(target, propertyName, receiver);
227
255
  },
228
256
  }),
229
257
  get response() {
230
- throw new Error('The `response` property is not available - `skipNavigation` was used');
258
+ throw new NavigationSkippedError('The `response` property is not available - `skipNavigation` was used');
231
259
  },
232
260
  };
233
261
  }
234
- const gotoOptions = { timeout: this.navigationTimeoutMillis };
235
- const preNavigationHooksCookies = this._getCookieHeaderFromRequest(crawlingContext.request);
236
262
  crawlingContext.request.state = RequestState.BEFORE_NAV;
237
- await this._executeHooks(this.preNavigationHooks, crawlingContext, gotoOptions);
263
+ return {
264
+ // Default to the full navigation timeout so a pre-navigation hook can read it; `navigate` narrows it
265
+ // to the remaining shared window unless a hook overrode it (see there).
266
+ gotoOptions: { timeout: this.#navigationTimeoutMillis },
267
+ [COOKIES_BEFORE_HOOKS]: this.getCookieHeaderFromRequest(crawlingContext.request),
268
+ };
269
+ }
270
+ async navigate(crawlingContext) {
238
271
  tryCancel();
239
- const postNavigationHooksCookies = this._getCookieHeaderFromRequest(crawlingContext.request);
240
- await this._applyCookies(crawlingContext, preNavigationHooksCookies, postNavigationHooksCookies);
272
+ const gotoOptions = crawlingContext.gotoOptions;
273
+ const remaining = remainingNavigationWindowMillis(crawlingContext, this.#navigationTimeoutMillis);
274
+ if (remaining <= 0) {
275
+ throw new TimeoutError(`Navigation timed out after ${this.#navigationTimeoutMillis / 1000} seconds.`);
276
+ }
277
+ // If a hook left the default `navigationTimeoutMillis` in place, bound the goto to whatever is left of the
278
+ // shared navigation window. If it overrode the value - including `0`, Playwright's "no timeout" - honour
279
+ // that verbatim as the goto's own timeout. The driver enforces this natively (so a timed-out goto is
280
+ // aborted, not left lingering) and `handleNavigationTimeout` turns its error into our own message.
281
+ const gotoTimeout = gotoOptions;
282
+ if (gotoTimeout.timeout === this.#navigationTimeoutMillis) {
283
+ gotoTimeout.timeout = remaining;
284
+ }
285
+ const cookiesBeforeHooks = readContextField(crawlingContext, COOKIES_BEFORE_HOOKS);
286
+ const cookiesAfterHooks = this.getCookieHeaderFromRequest(crawlingContext.request);
287
+ await this.applyCookies(crawlingContext, cookiesBeforeHooks, cookiesAfterHooks);
241
288
  let response;
242
289
  try {
243
- response = (await this._navigationHandler(crawlingContext, gotoOptions)) ?? undefined;
290
+ response = (await this.navigationHandler(crawlingContext, gotoOptions)) ?? undefined;
244
291
  }
245
292
  catch (error) {
246
- await this._handleNavigationTimeout(crawlingContext, error);
293
+ await this.handleNavigationTimeout(crawlingContext, error);
247
294
  crawlingContext.request.state = RequestState.ERROR;
248
- this._throwIfProxyError(error);
295
+ this.throwIfProxyError(error);
249
296
  throw error;
250
297
  }
251
298
  tryCancel();
252
299
  crawlingContext.request.state = RequestState.AFTER_NAV;
253
- await this._executeHooks(this.postNavigationHooks, crawlingContext, gotoOptions);
300
+ return { response };
301
+ }
302
+ async finalizeNavigation(crawlingContext) {
303
+ tryCancel();
304
+ let response;
305
+ try {
306
+ response = crawlingContext.response;
307
+ }
308
+ catch {
309
+ // `preparePage` installs a throwing getter for `response`; reaching this branch means
310
+ // navigation produced no response and no hook overrode it. Treat as undefined.
311
+ }
254
312
  await this.processResponse(response, crawlingContext);
255
313
  tryCancel();
256
- // save cookies
257
- // TODO: Should we save the cookies also after/only the handle page?
258
- if (this.persistCookiesPerSession) {
259
- const cookies = await crawlingContext.browserController.getCookies(crawlingContext.page);
260
- tryCancel();
261
- crawlingContext.session?.setCookies(cookies, crawlingContext.request.loadedUrl);
314
+ // Persist cookies from the navigation response before the user handler runs.
315
+ // Cookies set during `requestHandler` are saved again afterwards.
316
+ await this.persistCookiesFromPage(crawlingContext);
317
+ return { request: crawlingContext.request };
318
+ }
319
+ /**
320
+ * Copies cookies from the live browser page into the session cookie jar.
321
+ */
322
+ async persistCookiesFromPage(crawlingContext) {
323
+ if (!this.#saveResponseCookies || !crawlingContext.session) {
324
+ return;
262
325
  }
263
- if (response !== undefined) {
264
- return {
265
- request: crawlingContext.request,
266
- response,
267
- };
326
+ const { cookies } = await this.browserPool.extractPageState(crawlingContext.page);
327
+ tryCancel();
328
+ // Prefer the live page URL — the handler may have navigated after the initial load.
329
+ const url = (await crawlingContext.page.url()) || crawlingContext.request.loadedUrl || crawlingContext.request.url;
330
+ for (const cookie of cookies) {
331
+ try {
332
+ await crawlingContext.session.cookieJar.setCookie(browserPoolCookieToToughCookie(cookie), url, {
333
+ ignoreError: false,
334
+ });
335
+ }
336
+ catch (e) {
337
+ this.log.debug(`Could not set cookie: ${e.message}`);
338
+ }
339
+ }
340
+ }
341
+ /**
342
+ * Runs the user request handler, then re-reads browser cookies so login flows /
343
+ * `page.setCookie` / XHR `Set-Cookie` updates are stored for later requests.
344
+ */
345
+ async runRequestHandler(crawlingContext) {
346
+ try {
347
+ await super.runRequestHandler(crawlingContext);
348
+ }
349
+ finally {
350
+ if (!crawlingContext.request.skipNavigation) {
351
+ try {
352
+ await this.persistCookiesFromPage(crawlingContext);
353
+ }
354
+ catch {
355
+ // Page may already be closed on some failure paths; ignore.
356
+ }
357
+ }
268
358
  }
269
- return {
270
- request: crawlingContext.request,
271
- };
272
359
  }
273
360
  async handleBlockedRequestByContent(crawlingContext) {
274
361
  if (this.retryOnBlocked) {
@@ -282,30 +369,38 @@ export class BrowserCrawler extends BasicCrawler {
282
369
  crawlingContext.request.state = RequestState.REQUEST_HANDLER;
283
370
  return {};
284
371
  }
285
- async _applyCookies({ session, request, page, browserController }, preHooksCookies, postHooksCookies) {
286
- const sessionCookie = session?.getCookies(request.url) ?? [];
372
+ async applyCookies({ session, request, page }, preHooksCookies, postHooksCookies) {
373
+ const sessionCookie = session
374
+ ? (await session.cookieJar.getCookies(request.url)).map(toughCookieToBrowserPoolCookie)
375
+ : [];
287
376
  const parsedPreHooksCookies = preHooksCookies.split(/ *; */).map((c) => cookieStringToToughCookie(c));
288
377
  const parsedPostHooksCookies = postHooksCookies.split(/ *; */).map((c) => cookieStringToToughCookie(c));
289
- await browserController.setCookies(page, [...sessionCookie, ...parsedPreHooksCookies, ...parsedPostHooksCookies]
378
+ const cookies = [...sessionCookie, ...parsedPreHooksCookies, ...parsedPostHooksCookies]
290
379
  .filter((c) => typeof c !== 'undefined' && c !== null)
291
- .map((c) => ({ ...c, url: c.domain ? undefined : request.url })));
380
+ .map((c) => ({ ...c, url: c.domain ? undefined : request.url }));
381
+ await this.browserPool.injectPageState(page, { cookies });
292
382
  }
293
383
  /**
294
- * Marks session bad in case of navigation timeout.
384
+ * Marks session bad on navigation timeout, and stops in-flight page loading on any navigation error.
295
385
  */
296
- async _handleNavigationTimeout(crawlingContext, error) {
297
- const { session } = crawlingContext;
298
- if (error && error.constructor.name === 'TimeoutError') {
299
- handleRequestTimeout({ session, errorMessage: error.message });
386
+ async handleNavigationTimeout(crawlingContext, error) {
387
+ const { session, page } = crawlingContext;
388
+ // Fire-and-forget: no user code will run on this page after a failed navigation.
389
+ // Swallow rejections: the page may already be detached.
390
+ void page.evaluate(() => window.stop()).catch(() => { });
391
+ if (isNavigationTimeoutError(error)) {
392
+ session?.markBad();
393
+ // The driver was handed the remaining window (usually shorter than `navigationTimeoutSecs` once the
394
+ // hooks have run), so it names that value in its own error; report the configured window instead.
395
+ throw new TimeoutError(`Navigation timed out after ${this.#navigationTimeoutMillis / 1000} seconds.`);
300
396
  }
301
- await crawlingContext.page.close();
302
397
  }
303
398
  /**
304
399
  * Transforms proxy-related errors to `SessionError`.
305
400
  */
306
- _throwIfProxyError(error) {
401
+ throwIfProxyError(error) {
307
402
  if (this.isProxyError(error)) {
308
- throw new SessionError(this._getMessageFromError(error));
403
+ throw new SessionError(this.getMessageFromError(error));
309
404
  }
310
405
  }
311
406
  async processResponse(response, crawlingContext) {
@@ -313,10 +408,16 @@ export class BrowserCrawler extends BasicCrawler {
313
408
  if (typeof response === 'object' && typeof response.status === 'function') {
314
409
  const status = response.status();
315
410
  this.stats.registerStatusCode(status);
411
+ if (this.isErrorStatusCode(status)) {
412
+ if (this.additionalHttpErrorStatusCodes.has(status)) {
413
+ throw new Error(`${status} - Error status code was set by user.`);
414
+ }
415
+ throw new Error(`${status} - Internal Server Error`);
416
+ }
316
417
  }
317
418
  if (this.sessionPool && response && session) {
318
419
  if (typeof response === 'object' && typeof response.status === 'function') {
319
- this._throwOnBlockedRequest(session, response.status());
420
+ this.throwOnBlockedRequest(response.status());
320
421
  }
321
422
  else {
322
423
  this.log.debug('Got a malformed Browser response.', { request, response });
@@ -324,68 +425,43 @@ export class BrowserCrawler extends BasicCrawler {
324
425
  }
325
426
  request.loadedUrl = await page.url();
326
427
  }
327
- async _extendLaunchContext(_pageId, launchContext) {
328
- const launchContextExtends = {};
329
- if (this.sessionPool) {
330
- launchContextExtends.session = await this.sessionPool.getSession();
331
- }
332
- if (this.proxyConfiguration && !launchContext.proxyUrl) {
333
- const proxyInfo = await this.proxyConfiguration.newProxyInfo(launchContextExtends.session?.id, {
334
- proxyTier: launchContext.proxyTier ?? undefined,
335
- });
336
- launchContext.proxyUrl = proxyInfo?.url;
337
- launchContextExtends.proxyInfo = proxyInfo;
338
- // Disable SSL verification for MITM proxies
339
- if (this.proxyConfiguration.isManInTheMiddle) {
340
- /**
341
- * @see https://playwright.dev/docs/api/class-browser/#browser-new-context
342
- * @see https://github.com/puppeteer/puppeteer/blob/main/docs/api.md
343
- */
344
- launchContext.launchOptions.ignoreHTTPSErrors = true;
345
- launchContext.launchOptions.acceptInsecureCerts = true;
346
- }
347
- }
348
- launchContext.extend(launchContextExtends);
349
- }
350
- _maybeAddSessionRetiredListener(_pageId, browserController) {
351
- if (this.sessionPool) {
352
- const listener = (session) => {
353
- const { launchContext } = browserController;
354
- if (session.id === launchContext.session.id) {
355
- this.browserPool.retireBrowserController(browserController);
356
- }
357
- };
358
- this.sessionPool.on(EVENT_SESSION_RETIRED, listener);
359
- browserController.on("browserClosed" /* BROWSER_CONTROLLER_EVENTS.BROWSER_CLOSED */, () => {
360
- return this.sessionPool.removeListener(EVENT_SESSION_RETIRED, listener);
361
- });
362
- }
363
- }
364
428
  /**
365
429
  * Function for cleaning up after all requests are processed.
366
430
  * @ignore
367
431
  */
368
432
  async teardown() {
369
- await this.browserPool.destroy();
433
+ await this.#browserPoolDep.ifOwned((pool) => pool.destroy());
370
434
  await super.teardown();
371
435
  }
372
436
  }
373
437
  /** @internal */
374
- export async function browserCrawlerEnqueueLinks({ options, page, requestQueue, robotsTxtFile, onSkippedRequest, originalRequestUrl, finalRequestUrl, }) {
438
+ function containsEnqueueLinks(options) {
439
+ return !!options.enqueueLinks;
440
+ }
441
+ /** @internal */
442
+ export async function browserCrawlerEnqueueLinks(options) {
443
+ const { options: enqueueLinksOptions, finalRequestUrl, originalRequestUrl, page } = options;
375
444
  const baseUrl = resolveBaseUrlForEnqueueLinksFiltering({
376
- enqueueStrategy: options?.strategy,
445
+ enqueueStrategy: enqueueLinksOptions?.strategy,
377
446
  finalRequestUrl,
378
447
  originalRequestUrl,
379
- userProvidedBaseUrl: options?.baseUrl,
448
+ userProvidedBaseUrl: enqueueLinksOptions?.baseUrl,
380
449
  });
381
- const urls = await extractUrlsFromPage(page, options?.selector ?? 'a', options?.baseUrl ?? finalRequestUrl ?? originalRequestUrl);
450
+ const urls = await extractUrlsFromPage(page, enqueueLinksOptions?.selector ?? 'a', enqueueLinksOptions?.baseUrl ?? finalRequestUrl ?? originalRequestUrl);
451
+ if (containsEnqueueLinks(options)) {
452
+ return options.enqueueLinks({
453
+ urls,
454
+ baseUrl,
455
+ ...enqueueLinksOptions,
456
+ });
457
+ }
382
458
  return enqueueLinks({
383
- requestQueue,
384
- robotsTxtFile,
385
- onSkippedRequest,
459
+ requestManager: options.requestManager,
460
+ robotsTxtFile: options.robotsTxtFile,
461
+ onSkippedRequest: options.onSkippedRequest,
386
462
  urls,
387
463
  baseUrl,
388
- ...options,
464
+ ...enqueueLinksOptions,
389
465
  });
390
466
  }
391
467
  /**
@@ -413,4 +489,3 @@ page, selector, baseUrl) {
413
489
  })
414
490
  .filter((href) => !!href);
415
491
  }
416
- //# sourceMappingURL=browser-crawler.js.map