@crawlee/browser 4.0.0-beta.21 → 4.0.0-beta.211
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.
- package/README.md +14 -14
- package/index.d.ts +0 -1
- package/index.js +0 -1
- package/internals/browser-crawler.d.ts +257 -176
- package/internals/browser-crawler.js +286 -246
- package/internals/browser-launcher.d.ts +71 -23
- package/internals/browser-launcher.js +73 -23
- package/package.json +10 -10
- package/index.d.ts.map +0 -1
- package/index.js.map +0 -1
- package/internals/browser-crawler.d.ts.map +0 -1
- package/internals/browser-crawler.js.map +0 -1
- package/internals/browser-launcher.d.ts.map +0 -1
- package/internals/browser-launcher.js.map +0 -1
|
@@ -1,8 +1,21 @@
|
|
|
1
|
-
import { BasicCrawler,
|
|
2
|
-
import {
|
|
3
|
-
import {
|
|
4
|
-
import
|
|
5
|
-
import { tryCancel } from '@apify/timeout';
|
|
1
|
+
import { BasicCrawler, browserPoolCookieToToughCookie, ContextPipeline, cookieStringToToughCookie, EnqueueStrategy, NavigationSkippedError, OwnedOrInjected, remainingNavigationWindowMillis, RequestState, ContextPipelineInitializationError, RequestHandlerError, RequestThrottledError, resolveBaseUrlForEnqueueLinksFiltering, SessionError, SessionRetiredError, toughCookieToBrowserPoolCookie, validators, } from '@crawlee/basic';
|
|
2
|
+
import { assertBrowserPoolNotConfigured, CLOUDFLARE_RETRY_CSS_SELECTORS, parseArgument, RETRY_CSS_SELECTORS, schemas, tryAbsoluteURL, } from '@crawlee/utils/internal';
|
|
3
|
+
import { sleep } from '@crawlee/utils';
|
|
4
|
+
import { z } from 'zod';
|
|
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)
|
|
@@ -15,118 +28,126 @@ 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
|
|
19
|
-
*
|
|
20
|
-
* constructor
|
|
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
|
-
*
|
|
25
|
-
*
|
|
26
|
-
*
|
|
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
|
-
*
|
|
35
|
-
*
|
|
36
|
-
*
|
|
37
|
-
*
|
|
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
|
-
|
|
60
|
+
/** Backs the {@link BrowserCrawler.browserPool|`browserPool`} getter. */
|
|
61
|
+
#browserPoolDep;
|
|
47
62
|
/**
|
|
48
|
-
* A reference to the underlying
|
|
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.
|
|
49
65
|
*/
|
|
50
|
-
browserPool
|
|
51
|
-
|
|
66
|
+
get browserPool() {
|
|
67
|
+
return this.#browserPoolDep.value;
|
|
68
|
+
}
|
|
52
69
|
ignoreShadowRoots;
|
|
53
70
|
ignoreIframes;
|
|
54
|
-
navigationTimeoutMillis;
|
|
55
|
-
preNavigationHooks;
|
|
56
|
-
postNavigationHooks;
|
|
57
|
-
|
|
71
|
+
#navigationTimeoutMillis;
|
|
72
|
+
#preNavigationHooks;
|
|
73
|
+
#postNavigationHooks;
|
|
74
|
+
#saveResponseCookies;
|
|
75
|
+
/**
|
|
76
|
+
* @internal
|
|
77
|
+
*/
|
|
58
78
|
static optionsShape = {
|
|
59
79
|
...BasicCrawler.optionsShape,
|
|
60
|
-
navigationTimeoutSecs:
|
|
61
|
-
|
|
62
|
-
|
|
63
|
-
|
|
64
|
-
|
|
65
|
-
|
|
66
|
-
|
|
67
|
-
|
|
68
|
-
|
|
69
|
-
proxyConfiguration:
|
|
80
|
+
navigationTimeoutSecs: schemas.anyNumber
|
|
81
|
+
.refine((value) => value > 0, 'Expected a number greater than 0')
|
|
82
|
+
.default(60),
|
|
83
|
+
preNavigationHooks: schemas.anyArray.default(() => []),
|
|
84
|
+
postNavigationHooks: schemas.anyArray.default(() => []),
|
|
85
|
+
browserPool: validators.browserPool.optional(),
|
|
86
|
+
browserPoolBuilder: schemas.anyFunction.optional(),
|
|
87
|
+
remoteBrowser: schemas.anyObject.optional(),
|
|
88
|
+
saveResponseCookies: z.boolean().default(true),
|
|
89
|
+
proxyConfiguration: validators.proxyConfiguration.optional(),
|
|
90
|
+
ignoreIframes: z.boolean().default(false),
|
|
91
|
+
ignoreShadowRoots: z.boolean().default(false),
|
|
70
92
|
};
|
|
93
|
+
/** @internal */
|
|
94
|
+
static optionsSchema = z.strictObject(BrowserCrawler.optionsShape);
|
|
71
95
|
/**
|
|
72
96
|
* All `BrowserCrawler` parameters are passed via an options object.
|
|
97
|
+
* @internal
|
|
73
98
|
*/
|
|
74
|
-
constructor(options
|
|
75
|
-
|
|
76
|
-
|
|
99
|
+
constructor(options) {
|
|
100
|
+
const { navigationTimeoutSecs, saveResponseCookies, browserPool, remoteBrowser, preNavigationHooks, postNavigationHooks, ignoreIframes, ignoreShadowRoots, contextPipelineBuilder, browserPoolBuilder, extendContext, ...basicCrawlerOptions } = parseArgument(options, BrowserCrawler.optionsSchema, 'BrowserCrawlerOptions');
|
|
101
|
+
if (browserPool) {
|
|
102
|
+
assertBrowserPoolNotConfigured(new.target.name, { remoteBrowser });
|
|
103
|
+
}
|
|
104
|
+
const skipGuard = (action) => async (ctx) => ctx.request.skipNavigation ? {} : ((await action(ctx)) ?? {});
|
|
77
105
|
super({
|
|
78
106
|
...basicCrawlerOptions,
|
|
79
|
-
contextPipelineBuilder: () =>
|
|
80
|
-
|
|
81
|
-
|
|
82
|
-
|
|
83
|
-
|
|
84
|
-
|
|
85
|
-
|
|
86
|
-
|
|
87
|
-
|
|
88
|
-
|
|
89
|
-
|
|
90
|
-
|
|
91
|
-
|
|
92
|
-
|
|
93
|
-
|
|
94
|
-
|
|
107
|
+
contextPipelineBuilder: () => {
|
|
108
|
+
// A single navigation window covers the pre-navigation hooks, the navigation, and the
|
|
109
|
+
// post-navigation hooks: the whole phase shares one `navigationTimeoutSecs` budget, so a slow
|
|
110
|
+
// hook eats into the same window the navigation uses. The navigation itself is bounded by
|
|
111
|
+
// capping its `gotoOptions.timeout` to the remaining budget.
|
|
112
|
+
const windowGuard = (step) => skipGuard(async (ctx) => {
|
|
113
|
+
const remaining = remainingNavigationWindowMillis(ctx, this.#navigationTimeoutMillis);
|
|
114
|
+
if (remaining <= 0) {
|
|
115
|
+
throw new TimeoutError(`Navigation timed out after ${this.#navigationTimeoutMillis / 1000} seconds.`);
|
|
116
|
+
}
|
|
117
|
+
return addTimeoutToPromise(async () => step(ctx), remaining, `Navigation timed out after ${this.#navigationTimeoutMillis / 1000} seconds.`);
|
|
118
|
+
});
|
|
119
|
+
let pipeline = contextPipelineBuilder().compose(this.prepareNavigation.bind(this));
|
|
120
|
+
for (const hook of this.#preNavigationHooks) {
|
|
121
|
+
pipeline = pipeline.compose(windowGuard(hook));
|
|
122
|
+
}
|
|
123
|
+
pipeline = pipeline.compose(skipGuard(this.navigate.bind(this)));
|
|
124
|
+
for (const hook of this.#postNavigationHooks) {
|
|
125
|
+
pipeline = pipeline.compose(windowGuard(hook));
|
|
126
|
+
}
|
|
127
|
+
return pipeline
|
|
128
|
+
.compose(skipGuard(this.finalizeNavigation.bind(this)))
|
|
129
|
+
.compose(this.handleBlockedRequestByContent.bind(this))
|
|
130
|
+
.compose(this.restoreRequestState.bind(this));
|
|
131
|
+
},
|
|
132
|
+
extendContext,
|
|
133
|
+
});
|
|
134
|
+
this.#navigationTimeoutMillis = navigationTimeoutSecs * 1000;
|
|
135
|
+
// The public option hooks are extension-aware; internal storage uses the base context type
|
|
136
|
+
// (the pipeline composes hooks against the concrete context, which does not statically carry
|
|
137
|
+
// `ContextExtension`). The extension members are present at runtime regardless.
|
|
138
|
+
this.#preNavigationHooks = preNavigationHooks;
|
|
139
|
+
this.#postNavigationHooks = postNavigationHooks;
|
|
95
140
|
this.ignoreIframes = ignoreIframes;
|
|
96
141
|
this.ignoreShadowRoots = ignoreShadowRoots;
|
|
97
|
-
|
|
98
|
-
|
|
99
|
-
|
|
100
|
-
|
|
101
|
-
|
|
102
|
-
|
|
103
|
-
}
|
|
104
|
-
else {
|
|
105
|
-
this.persistCookiesPerSession = false;
|
|
106
|
-
}
|
|
107
|
-
if (launchContext?.userAgent) {
|
|
108
|
-
if (browserPoolOptions.useFingerprints)
|
|
109
|
-
this.log.info('Custom user agent provided, disabling automatic browser fingerprint injection!');
|
|
110
|
-
browserPoolOptions.useFingerprints = false;
|
|
111
|
-
}
|
|
112
|
-
const { preLaunchHooks = [], postLaunchHooks = [], ...rest } = browserPoolOptions;
|
|
113
|
-
this.browserPool = new BrowserPool({
|
|
114
|
-
...rest,
|
|
115
|
-
preLaunchHooks: [this._extendLaunchContext.bind(this), ...preLaunchHooks],
|
|
116
|
-
postLaunchHooks: [this._maybeAddSessionRetiredListener.bind(this), ...postLaunchHooks],
|
|
117
|
-
});
|
|
142
|
+
this.#saveResponseCookies = saveResponseCookies;
|
|
143
|
+
this.#browserPoolDep = OwnedOrInjected.resolve(browserPool, () => browserPoolBuilder(remoteBrowser));
|
|
144
|
+
}
|
|
145
|
+
/** @internal */
|
|
146
|
+
getNavigationTimeoutMillis() {
|
|
147
|
+
return this.#navigationTimeoutMillis;
|
|
118
148
|
}
|
|
119
149
|
buildContextPipeline() {
|
|
120
|
-
return ContextPipeline.create().compose(
|
|
121
|
-
action: this.preparePage.bind(this),
|
|
122
|
-
cleanup: async (context) => {
|
|
123
|
-
context.registerDeferredCleanup(async () => {
|
|
124
|
-
await context.page
|
|
125
|
-
.close()
|
|
126
|
-
.catch((error) => this.log.debug('Error while closing page', { error }));
|
|
127
|
-
});
|
|
128
|
-
},
|
|
129
|
-
});
|
|
150
|
+
return ContextPipeline.create().compose(this.preparePage.bind(this));
|
|
130
151
|
}
|
|
131
152
|
async containsSelectors(page, selectors) {
|
|
132
153
|
const foundSelectors = (await Promise.all(selectors.map((selector) => page.$(selector))))
|
|
@@ -137,12 +158,6 @@ export class BrowserCrawler extends BasicCrawler {
|
|
|
137
158
|
}
|
|
138
159
|
async isRequestBlocked(crawlingContext) {
|
|
139
160
|
const { page, response } = crawlingContext;
|
|
140
|
-
const blockedStatusCodes =
|
|
141
|
-
// eslint-disable-next-line dot-notation
|
|
142
|
-
(this.sessionPool?.['blockedStatusCodes'].length ?? 0) > 0
|
|
143
|
-
? // eslint-disable-next-line dot-notation
|
|
144
|
-
this.sessionPool['blockedStatusCodes']
|
|
145
|
-
: DEFAULT_BLOCKED_STATUS_CODES;
|
|
146
161
|
// Cloudflare specific heuristic - wait 5 seconds if we get a 403 for the JS challenge to load / resolve.
|
|
147
162
|
if ((await this.containsSelectors(page, CLOUDFLARE_RETRY_CSS_SELECTORS)) && response?.status() === 403) {
|
|
148
163
|
await sleep(5000);
|
|
@@ -153,117 +168,185 @@ export class BrowserCrawler extends BasicCrawler {
|
|
|
153
168
|
return `Cloudflare challenge failed, found selectors: ${foundSelectors.join(', ')}`;
|
|
154
169
|
}
|
|
155
170
|
const foundSelectors = await this.containsSelectors(page, RETRY_CSS_SELECTORS);
|
|
156
|
-
const
|
|
171
|
+
const statusCode = response?.status() ?? 0;
|
|
157
172
|
if (foundSelectors)
|
|
158
173
|
return `Found selectors: ${foundSelectors.join(', ')}`;
|
|
159
|
-
if (
|
|
160
|
-
return `Received blocked status code: ${
|
|
174
|
+
if (this.blockedStatusCodes.has(statusCode))
|
|
175
|
+
return `Received blocked status code: ${statusCode}`;
|
|
161
176
|
return false;
|
|
162
177
|
}
|
|
163
|
-
async preparePage(crawlingContext) {
|
|
164
|
-
const
|
|
178
|
+
async preparePage(crawlingContext, onCleanup) {
|
|
179
|
+
const page = await this.browserPool.newPage({
|
|
165
180
|
id: crawlingContext.id,
|
|
166
|
-
|
|
167
|
-
|
|
168
|
-
if (crawlingContext.session?.proxyInfo) {
|
|
169
|
-
const proxyInfo = crawlingContext.session.proxyInfo;
|
|
170
|
-
crawlingContext.proxyInfo = proxyInfo;
|
|
171
|
-
newPageOptions.proxyUrl = proxyInfo?.url;
|
|
172
|
-
newPageOptions.proxyTier = proxyInfo?.proxyTier;
|
|
173
|
-
if (proxyInfo?.ignoreTlsErrors) {
|
|
174
|
-
/**
|
|
175
|
-
* @see https://playwright.dev/docs/api/class-browser/#browser-new-context
|
|
176
|
-
* @see https://github.com/puppeteer/puppeteer/blob/main/docs/api.md
|
|
177
|
-
*/
|
|
178
|
-
newPageOptions.pageOptions = {
|
|
179
|
-
ignoreHTTPSErrors: true,
|
|
180
|
-
acceptInsecureCerts: true,
|
|
181
|
-
};
|
|
182
|
-
}
|
|
183
|
-
}
|
|
184
|
-
const page = (await this.browserPool.newPage(newPageOptions));
|
|
181
|
+
session: crawlingContext.session,
|
|
182
|
+
});
|
|
185
183
|
tryCancel();
|
|
186
|
-
|
|
187
|
-
|
|
188
|
-
|
|
189
|
-
|
|
190
|
-
|
|
184
|
+
onCleanup((failure) => {
|
|
185
|
+
crawlingContext.registerDeferredCleanup(async () => {
|
|
186
|
+
const cause = failure instanceof RequestHandlerError || failure instanceof ContextPipelineInitializationError
|
|
187
|
+
? failure.cause
|
|
188
|
+
: failure;
|
|
189
|
+
// Only tells the pool what to do with the page's browser state - the request's failure, if any, has
|
|
190
|
+
// already been handled. A thrown `SessionError` or a blocked session counts as a block; a session
|
|
191
|
+
// unusable for any other reason is reported as retired. `retire()`/`markBad()` also bump the usage
|
|
192
|
+
// count, so the block check must come first.
|
|
193
|
+
const { session } = crawlingContext;
|
|
194
|
+
const closeReason = cause instanceof SessionError
|
|
195
|
+
? cause
|
|
196
|
+
: session.isBlocked()
|
|
197
|
+
? new SessionError()
|
|
198
|
+
: session.isUsable()
|
|
199
|
+
? undefined
|
|
200
|
+
: new SessionRetiredError();
|
|
201
|
+
await this.browserPool
|
|
202
|
+
.closePage(page, { error: closeReason })
|
|
203
|
+
.catch((closeError) => this.log.debug('Error while closing page', { error: closeError }));
|
|
204
|
+
});
|
|
205
|
+
});
|
|
206
|
+
const addRequests = crawlingContext.addRequests;
|
|
207
|
+
const extractLinks = async (options) => {
|
|
208
|
+
return extractUrlsFromPage(page, options?.selector ?? 'a', options?.baseUrl ?? crawlingContext.request.loadedUrl ?? crawlingContext.request.url);
|
|
209
|
+
};
|
|
191
210
|
return {
|
|
192
211
|
page,
|
|
193
212
|
get response() {
|
|
194
213
|
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)");
|
|
195
214
|
},
|
|
196
|
-
|
|
197
|
-
|
|
198
|
-
|
|
199
|
-
|
|
200
|
-
|
|
201
|
-
|
|
202
|
-
|
|
203
|
-
requestQueue: await this.getRequestQueue(),
|
|
204
|
-
robotsTxtFile: await this.getRobotsTxtFileForUrl(crawlingContext.request.url),
|
|
205
|
-
onSkippedRequest: this.handleSkippedRequest,
|
|
206
|
-
originalRequestUrl: crawlingContext.request.url,
|
|
215
|
+
get gotoOptions() {
|
|
216
|
+
throw new Error('The `gotoOptions` property is not available until `prepareNavigation` runs.');
|
|
217
|
+
},
|
|
218
|
+
extractLinks,
|
|
219
|
+
enqueueLinks: async (options = {}) => {
|
|
220
|
+
const baseUrl = resolveBaseUrlForEnqueueLinksFiltering({
|
|
221
|
+
enqueueStrategy: options.strategy,
|
|
207
222
|
finalRequestUrl: crawlingContext.request.loadedUrl,
|
|
208
|
-
|
|
209
|
-
|
|
223
|
+
originalRequestUrl: crawlingContext.request.url,
|
|
224
|
+
userProvidedBaseUrl: options.baseUrl,
|
|
225
|
+
});
|
|
226
|
+
const urls = await extractLinks(options);
|
|
227
|
+
return addRequests(urls, {
|
|
228
|
+
...options,
|
|
229
|
+
baseUrl,
|
|
230
|
+
strategy: options.strategy ?? EnqueueStrategy.SameHostname,
|
|
231
|
+
});
|
|
210
232
|
},
|
|
211
233
|
};
|
|
212
234
|
}
|
|
213
|
-
async
|
|
235
|
+
async prepareNavigation(crawlingContext) {
|
|
214
236
|
if (crawlingContext.request.skipNavigation) {
|
|
215
237
|
return {
|
|
216
238
|
request: new Proxy(crawlingContext.request, {
|
|
217
239
|
get(target, propertyName, receiver) {
|
|
218
240
|
if (propertyName === 'loadedUrl') {
|
|
219
|
-
throw new
|
|
241
|
+
throw new NavigationSkippedError('The `request.loadedUrl` property is not available - `skipNavigation` was used');
|
|
220
242
|
}
|
|
221
243
|
return Reflect.get(target, propertyName, receiver);
|
|
222
244
|
},
|
|
223
245
|
}),
|
|
224
246
|
get response() {
|
|
225
|
-
throw new
|
|
247
|
+
throw new NavigationSkippedError('The `response` property is not available - `skipNavigation` was used');
|
|
226
248
|
},
|
|
227
249
|
};
|
|
228
250
|
}
|
|
229
|
-
const gotoOptions = { timeout: this.navigationTimeoutMillis };
|
|
230
|
-
const preNavigationHooksCookies = this._getCookieHeaderFromRequest(crawlingContext.request);
|
|
231
251
|
crawlingContext.request.state = RequestState.BEFORE_NAV;
|
|
232
|
-
|
|
252
|
+
return {
|
|
253
|
+
// Default to the full navigation timeout so a pre-navigation hook can read it; `navigate` narrows it
|
|
254
|
+
// to the remaining shared window unless a hook overrode it (see there).
|
|
255
|
+
gotoOptions: { timeout: this.#navigationTimeoutMillis },
|
|
256
|
+
[COOKIES_BEFORE_HOOKS]: this.getCookieHeaderFromRequest(crawlingContext.request),
|
|
257
|
+
};
|
|
258
|
+
}
|
|
259
|
+
// oxlint-disable-next-line crawlee/prefer-private-fields -- patched by @crawlee/otel
|
|
260
|
+
async navigate(crawlingContext) {
|
|
233
261
|
tryCancel();
|
|
234
|
-
const
|
|
235
|
-
|
|
262
|
+
const gotoOptions = crawlingContext.gotoOptions;
|
|
263
|
+
const remaining = remainingNavigationWindowMillis(crawlingContext, this.#navigationTimeoutMillis);
|
|
264
|
+
if (remaining <= 0) {
|
|
265
|
+
throw new TimeoutError(`Navigation timed out after ${this.#navigationTimeoutMillis / 1000} seconds.`);
|
|
266
|
+
}
|
|
267
|
+
// If a hook left the default `navigationTimeoutMillis` in place, bound the goto to whatever is left of the
|
|
268
|
+
// shared navigation window. If it overrode the value - including `0`, Playwright's "no timeout" - honour
|
|
269
|
+
// that verbatim as the goto's own timeout. The driver enforces this natively (so a timed-out goto is
|
|
270
|
+
// aborted, not left lingering) and `handleNavigationTimeout` turns its error into our own message.
|
|
271
|
+
const gotoTimeout = gotoOptions;
|
|
272
|
+
if (gotoTimeout.timeout === this.#navigationTimeoutMillis) {
|
|
273
|
+
gotoTimeout.timeout = remaining;
|
|
274
|
+
}
|
|
275
|
+
const cookiesBeforeHooks = readContextField(crawlingContext, COOKIES_BEFORE_HOOKS);
|
|
276
|
+
const cookiesAfterHooks = this.getCookieHeaderFromRequest(crawlingContext.request);
|
|
277
|
+
await this.applyCookies(crawlingContext, cookiesBeforeHooks, cookiesAfterHooks);
|
|
236
278
|
let response;
|
|
237
279
|
try {
|
|
238
|
-
response = (await this.
|
|
280
|
+
response = (await this.navigationHandler(crawlingContext, gotoOptions)) ?? undefined;
|
|
239
281
|
}
|
|
240
282
|
catch (error) {
|
|
241
|
-
await this.
|
|
283
|
+
await this.handleNavigationTimeout(crawlingContext, error);
|
|
242
284
|
crawlingContext.request.state = RequestState.ERROR;
|
|
243
|
-
this.
|
|
285
|
+
this.throwIfProxyError(error);
|
|
244
286
|
throw error;
|
|
245
287
|
}
|
|
246
288
|
tryCancel();
|
|
247
289
|
crawlingContext.request.state = RequestState.AFTER_NAV;
|
|
248
|
-
|
|
290
|
+
return { response };
|
|
291
|
+
}
|
|
292
|
+
async finalizeNavigation(crawlingContext) {
|
|
293
|
+
tryCancel();
|
|
294
|
+
let response;
|
|
295
|
+
try {
|
|
296
|
+
response = crawlingContext.response;
|
|
297
|
+
}
|
|
298
|
+
catch {
|
|
299
|
+
// `preparePage` installs a throwing getter for `response`; reaching this branch means
|
|
300
|
+
// navigation produced no response and no hook overrode it. Treat as undefined.
|
|
301
|
+
}
|
|
249
302
|
await this.processResponse(response, crawlingContext);
|
|
250
303
|
tryCancel();
|
|
251
|
-
//
|
|
252
|
-
//
|
|
253
|
-
|
|
254
|
-
|
|
255
|
-
|
|
256
|
-
|
|
304
|
+
// Persist cookies from the navigation response before the user handler runs.
|
|
305
|
+
// Cookies set during `requestHandler` are saved again afterwards.
|
|
306
|
+
await this.persistCookiesFromPage(crawlingContext);
|
|
307
|
+
return { request: crawlingContext.request };
|
|
308
|
+
}
|
|
309
|
+
/**
|
|
310
|
+
* Copies cookies from the live browser page into the session cookie jar.
|
|
311
|
+
*/
|
|
312
|
+
async persistCookiesFromPage(crawlingContext) {
|
|
313
|
+
if (!this.#saveResponseCookies || !crawlingContext.session) {
|
|
314
|
+
return;
|
|
257
315
|
}
|
|
258
|
-
|
|
259
|
-
|
|
260
|
-
|
|
261
|
-
|
|
262
|
-
|
|
316
|
+
const { cookies } = await this.browserPool.extractPageState(crawlingContext.page);
|
|
317
|
+
tryCancel();
|
|
318
|
+
// Prefer the live page URL — the handler may have navigated after the initial load.
|
|
319
|
+
const url = (await crawlingContext.page.url()) || crawlingContext.request.loadedUrl || crawlingContext.request.url;
|
|
320
|
+
for (const cookie of cookies) {
|
|
321
|
+
try {
|
|
322
|
+
await crawlingContext.session.cookieJar.setCookie(browserPoolCookieToToughCookie(cookie), url, {
|
|
323
|
+
ignoreError: false,
|
|
324
|
+
});
|
|
325
|
+
}
|
|
326
|
+
catch (e) {
|
|
327
|
+
this.log.debug(`Could not set cookie: ${e.message}`);
|
|
328
|
+
}
|
|
329
|
+
}
|
|
330
|
+
}
|
|
331
|
+
/**
|
|
332
|
+
* Runs the user request handler, then re-reads browser cookies so login flows /
|
|
333
|
+
* `page.setCookie` / XHR `Set-Cookie` updates are stored for later requests.
|
|
334
|
+
* @internal
|
|
335
|
+
*/
|
|
336
|
+
async runRequestHandler(crawlingContext) {
|
|
337
|
+
try {
|
|
338
|
+
await super.runRequestHandler(crawlingContext);
|
|
339
|
+
}
|
|
340
|
+
finally {
|
|
341
|
+
if (!crawlingContext.request.skipNavigation) {
|
|
342
|
+
try {
|
|
343
|
+
await this.persistCookiesFromPage(crawlingContext);
|
|
344
|
+
}
|
|
345
|
+
catch {
|
|
346
|
+
// Page may already be closed on some failure paths; ignore.
|
|
347
|
+
}
|
|
348
|
+
}
|
|
263
349
|
}
|
|
264
|
-
return {
|
|
265
|
-
request: crawlingContext.request,
|
|
266
|
-
};
|
|
267
350
|
}
|
|
268
351
|
async handleBlockedRequestByContent(crawlingContext) {
|
|
269
352
|
if (this.retryOnBlocked) {
|
|
@@ -277,41 +360,64 @@ export class BrowserCrawler extends BasicCrawler {
|
|
|
277
360
|
crawlingContext.request.state = RequestState.REQUEST_HANDLER;
|
|
278
361
|
return {};
|
|
279
362
|
}
|
|
280
|
-
async
|
|
281
|
-
const sessionCookie = session
|
|
363
|
+
async applyCookies({ session, request, page }, preHooksCookies, postHooksCookies) {
|
|
364
|
+
const sessionCookie = session
|
|
365
|
+
? (await session.cookieJar.getCookies(request.url)).map(toughCookieToBrowserPoolCookie)
|
|
366
|
+
: [];
|
|
282
367
|
const parsedPreHooksCookies = preHooksCookies.split(/ *; */).map((c) => cookieStringToToughCookie(c));
|
|
283
368
|
const parsedPostHooksCookies = postHooksCookies.split(/ *; */).map((c) => cookieStringToToughCookie(c));
|
|
284
|
-
|
|
369
|
+
const cookies = [...sessionCookie, ...parsedPreHooksCookies, ...parsedPostHooksCookies]
|
|
285
370
|
.filter((c) => typeof c !== 'undefined' && c !== null)
|
|
286
|
-
.map((c) => ({ ...c, url: c.domain ? undefined : request.url }))
|
|
371
|
+
.map((c) => ({ ...c, url: c.domain ? undefined : request.url }));
|
|
372
|
+
await this.browserPool.injectPageState(page, { cookies });
|
|
287
373
|
}
|
|
288
374
|
/**
|
|
289
|
-
* Marks session bad in
|
|
375
|
+
* Marks session bad on navigation timeout, and stops in-flight page loading on any navigation error.
|
|
290
376
|
*/
|
|
291
|
-
async
|
|
292
|
-
const { session } = crawlingContext;
|
|
293
|
-
|
|
294
|
-
|
|
377
|
+
async handleNavigationTimeout(crawlingContext, error) {
|
|
378
|
+
const { session, page } = crawlingContext;
|
|
379
|
+
// Fire-and-forget: no user code will run on this page after a failed navigation.
|
|
380
|
+
// Swallow rejections: the page may already be detached.
|
|
381
|
+
void page.evaluate(() => window.stop()).catch(() => { });
|
|
382
|
+
if (isNavigationTimeoutError(error)) {
|
|
383
|
+
session?.markBad();
|
|
384
|
+
// The driver was handed the remaining window (usually shorter than `navigationTimeoutSecs` once the
|
|
385
|
+
// hooks have run), so it names that value in its own error; report the configured window instead.
|
|
386
|
+
throw new TimeoutError(`Navigation timed out after ${this.#navigationTimeoutMillis / 1000} seconds.`);
|
|
295
387
|
}
|
|
296
|
-
await crawlingContext.page.close();
|
|
297
388
|
}
|
|
298
389
|
/**
|
|
299
390
|
* Transforms proxy-related errors to `SessionError`.
|
|
300
391
|
*/
|
|
301
|
-
|
|
392
|
+
throwIfProxyError(error) {
|
|
302
393
|
if (this.isProxyError(error)) {
|
|
303
|
-
throw new SessionError(this.
|
|
394
|
+
throw new SessionError(this.getMessageFromError(error));
|
|
304
395
|
}
|
|
305
396
|
}
|
|
306
397
|
async processResponse(response, crawlingContext) {
|
|
307
398
|
const { session, request, page } = crawlingContext;
|
|
308
399
|
if (typeof response === 'object' && typeof response.status === 'function') {
|
|
309
400
|
const status = response.status();
|
|
310
|
-
this.
|
|
401
|
+
this.statistics.registerStatusCode(status);
|
|
402
|
+
// Ahead of the error-status throw below: a 429 the user opted into treating as an error is still a
|
|
403
|
+
// rate limit the domain should back off from.
|
|
404
|
+
if (status === 429) {
|
|
405
|
+
// Both drivers lower-case header names and join duplicates, so a plain lookup is enough.
|
|
406
|
+
const retryAfter = response.headers?.()['retry-after'];
|
|
407
|
+
if (this.recordDomainRateLimit(crawlingContext.request.url, retryAfter)) {
|
|
408
|
+
throw new RequestThrottledError(`${crawlingContext.request.url} responded with 429.`);
|
|
409
|
+
}
|
|
410
|
+
}
|
|
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
|
+
}
|
|
311
417
|
}
|
|
312
418
|
if (this.sessionPool && response && session) {
|
|
313
419
|
if (typeof response === 'object' && typeof response.status === 'function') {
|
|
314
|
-
this.
|
|
420
|
+
this.throwOnBlockedRequest(response.status());
|
|
315
421
|
}
|
|
316
422
|
else {
|
|
317
423
|
this.log.debug('Got a malformed Browser response.', { request, response });
|
|
@@ -319,83 +425,18 @@ export class BrowserCrawler extends BasicCrawler {
|
|
|
319
425
|
}
|
|
320
426
|
request.loadedUrl = await page.url();
|
|
321
427
|
}
|
|
322
|
-
async _extendLaunchContext(_pageId, launchContext) {
|
|
323
|
-
const launchContextExtends = {};
|
|
324
|
-
if (this.sessionPool) {
|
|
325
|
-
launchContextExtends.session = await this.sessionPool.newSession({
|
|
326
|
-
proxyInfo: await this.proxyConfiguration?.newProxyInfo({
|
|
327
|
-
// cannot pass a request here, since session is created on browser launch
|
|
328
|
-
}),
|
|
329
|
-
});
|
|
330
|
-
}
|
|
331
|
-
if (!launchContext.proxyUrl && launchContextExtends.session?.proxyInfo) {
|
|
332
|
-
const proxyInfo = launchContextExtends.session.proxyInfo;
|
|
333
|
-
launchContext.proxyUrl = proxyInfo?.url;
|
|
334
|
-
launchContextExtends.proxyInfo = proxyInfo;
|
|
335
|
-
// Disable SSL verification for MITM proxies
|
|
336
|
-
if (proxyInfo?.ignoreTlsErrors) {
|
|
337
|
-
/**
|
|
338
|
-
* @see https://playwright.dev/docs/api/class-browser/#browser-new-context
|
|
339
|
-
* @see https://github.com/puppeteer/puppeteer/blob/main/docs/api.md
|
|
340
|
-
*/
|
|
341
|
-
launchContext.launchOptions.ignoreHTTPSErrors = true;
|
|
342
|
-
launchContext.launchOptions.acceptInsecureCerts = true;
|
|
343
|
-
}
|
|
344
|
-
}
|
|
345
|
-
launchContext.extend(launchContextExtends);
|
|
346
|
-
}
|
|
347
|
-
_maybeAddSessionRetiredListener(_pageId, browserController) {
|
|
348
|
-
if (this.sessionPool) {
|
|
349
|
-
const listener = (session) => {
|
|
350
|
-
const { launchContext } = browserController;
|
|
351
|
-
if (session.id === launchContext.session.id) {
|
|
352
|
-
this.browserPool.retireBrowserController(browserController);
|
|
353
|
-
}
|
|
354
|
-
};
|
|
355
|
-
this.sessionPool.on(EVENT_SESSION_RETIRED, listener);
|
|
356
|
-
browserController.on("browserClosed" /* BROWSER_CONTROLLER_EVENTS.BROWSER_CLOSED */, () => {
|
|
357
|
-
return this.sessionPool.removeListener(EVENT_SESSION_RETIRED, listener);
|
|
358
|
-
});
|
|
359
|
-
}
|
|
360
|
-
}
|
|
361
428
|
/**
|
|
362
|
-
*
|
|
363
|
-
*
|
|
429
|
+
* Closes the browsers of a pool the crawler owns, so a finished run leaves none behind. The pool itself is
|
|
430
|
+
* crawler-lifetime and survives — destroying it here would hand a repeated `run()` a dead pool.
|
|
364
431
|
*/
|
|
365
432
|
async teardown() {
|
|
366
|
-
await this.
|
|
433
|
+
await this.#browserPoolDep.ifOwned((pool) => pool.releaseAllBrowsers());
|
|
367
434
|
await super.teardown();
|
|
368
435
|
}
|
|
369
|
-
|
|
370
|
-
|
|
371
|
-
|
|
372
|
-
return !!options.enqueueLinks;
|
|
373
|
-
}
|
|
374
|
-
/** @internal */
|
|
375
|
-
export async function browserCrawlerEnqueueLinks(options) {
|
|
376
|
-
const { options: enqueueLinksOptions, finalRequestUrl, originalRequestUrl, page } = options;
|
|
377
|
-
const baseUrl = resolveBaseUrlForEnqueueLinksFiltering({
|
|
378
|
-
enqueueStrategy: enqueueLinksOptions?.strategy,
|
|
379
|
-
finalRequestUrl,
|
|
380
|
-
originalRequestUrl,
|
|
381
|
-
userProvidedBaseUrl: enqueueLinksOptions?.baseUrl,
|
|
382
|
-
});
|
|
383
|
-
const urls = await extractUrlsFromPage(page, enqueueLinksOptions?.selector ?? 'a', enqueueLinksOptions?.baseUrl ?? finalRequestUrl ?? originalRequestUrl);
|
|
384
|
-
if (containsEnqueueLinks(options)) {
|
|
385
|
-
return options.enqueueLinks({
|
|
386
|
-
urls,
|
|
387
|
-
baseUrl,
|
|
388
|
-
...enqueueLinksOptions,
|
|
389
|
-
});
|
|
436
|
+
async destroy() {
|
|
437
|
+
await super.destroy();
|
|
438
|
+
await this.#browserPoolDep.ifOwned((pool) => pool.destroy());
|
|
390
439
|
}
|
|
391
|
-
return enqueueLinks({
|
|
392
|
-
requestQueue: options.requestQueue,
|
|
393
|
-
robotsTxtFile: options.robotsTxtFile,
|
|
394
|
-
onSkippedRequest: options.onSkippedRequest,
|
|
395
|
-
urls,
|
|
396
|
-
baseUrl,
|
|
397
|
-
...enqueueLinksOptions,
|
|
398
|
-
});
|
|
399
440
|
}
|
|
400
441
|
/**
|
|
401
442
|
* Extracts URLs from a given page.
|
|
@@ -422,4 +463,3 @@ page, selector, baseUrl) {
|
|
|
422
463
|
})
|
|
423
464
|
.filter((href) => !!href);
|
|
424
465
|
}
|
|
425
|
-
//# sourceMappingURL=browser-crawler.js.map
|