thunderstrike/puppeteer
thunderstrike/puppeteer
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
2f1a52da8f
puppeteer/docs/api.md
Alex Rudenko 1308d9aa6a
feat: support chrome headless mode (#8260) 
* feat: support chrome headless mode

Co-authored-by: Ergün Erdoğmuş <erdogmusergun@gmail.com>
2022-04-25 13:11:23 +02:00

253 KiB
Raw Blame History

Puppeteer API Tip-Of-Tree

Table of Contents

Overview

Puppeteer is a Node library which provides a high-level API to control Chromium or Chrome over the DevTools Protocol.

The Puppeteer API is hierarchical and mirrors the browser structure.

Note

On the following diagram, faded entities are not currently represented in Puppeteer.

puppeteer overview

  • Puppeteer communicates with the browser using DevTools Protocol.
  • Browser instance can own multiple browser contexts.
  • BrowserContext instance defines a browsing session and can own multiple pages.
  • Page has at least one frame: main frame. There might be other frames created by iframe or frame tags.
  • Frame has at least one execution context - the default execution context - where the frame's JavaScript is executed. A Frame might have additional execution contexts that are associated with extensions.
  • WebWorker has a single execution context and facilitates interacting with WebWorkers.

(Diagram source: link)

puppeteer vs puppeteer-core

Every release since v1.7.0 we publish two packages:

puppeteer is a product for browser automation. When installed, it downloads a version of Chromium, which it then drives using puppeteer-core. Being an end-user product, puppeteer supports a bunch of convenient PUPPETEER_* env variables to tweak its behavior.

puppeteer-core is a library to help drive anything that supports DevTools protocol. puppeteer-core doesn't download Chromium when installed. Being a library, puppeteer-core is fully driven through its programmatic interface and disregards all the PUPPETEER_* env variables.

To sum up, the only differences between puppeteer-core and puppeteer are:

  • puppeteer-core doesn't automatically download Chromium when installed.
  • puppeteer-core ignores all PUPPETEER_* env variables.

In most cases, you'll be fine using the puppeteer package.

However, you should use puppeteer-core if:

  • you're building another end-user product or library atop of DevTools protocol. For example, one might build a PDF generator using puppeteer-core and write a custom install.js script that downloads headless_shell instead of Chromium to save disk space.
  • you're bundling Puppeteer to use in Chrome Extension / browser with the DevTools protocol where downloading an additional Chromium binary is unnecessary.
  • you're building a set of tools where puppeteer-core is one of the ingredients and you want to postpone install.js script execution until Chromium is about to be used.

When using puppeteer-core, remember to change the include line:

const puppeteer = require('puppeteer-core');

You will then need to call puppeteer.connect([options]) or puppeteer.launch([options]) with an explicit executablePath or channel option.

Environment Variables

Puppeteer looks for certain environment variables to aid its operations. If Puppeteer doesn't find them in the environment during the installation step, a lowercased variant of these variables will be used from the npm config.

  • HTTP_PROXY, HTTPS_PROXY, NO_PROXY - defines HTTP proxy settings that are used to download and run the browser.
  • PUPPETEER_SKIP_CHROMIUM_DOWNLOAD - do not download bundled Chromium during installation step.
  • PUPPETEER_TMP_DIR - defines the directory to be used by Puppeteer for creating temporary files. Defaults to os.tmpdir().
  • PUPPETEER_DOWNLOAD_HOST - overwrite URL prefix that is used to download Chromium. Note: this includes protocol and might even include path prefix. Defaults to https://storage.googleapis.com.
  • PUPPETEER_DOWNLOAD_PATH - overwrite the path for the downloads folder. Defaults to <root>/.local-chromium, where <root> is Puppeteer's package root.
  • PUPPETEER_CHROMIUM_REVISION - specify a certain version of Chromium you'd like Puppeteer to use. See puppeteer.launch([options]) on how executable path is inferred. BEWARE: Puppeteer is only guaranteed to work with the bundled Chromium, use at your own risk.
  • PUPPETEER_EXECUTABLE_PATH - specify an executable path to be used in puppeteer.launch. See puppeteer.launch([options]) on how the executable path is inferred. BEWARE: Puppeteer is only guaranteed to work with the bundled Chromium, use at your own risk.
  • PUPPETEER_PRODUCT - specify which browser you'd like Puppeteer to use. Must be one of chrome or firefox. This can also be used during installation to fetch the recommended browser binary. Setting product programmatically in puppeteer.launch([options]) supersedes this environment variable. The product is exposed in puppeteer.product

Note

PUPPETEER_* env variables are not accounted for in the puppeteer-core package.

Working with Chrome Extensions

Puppeteer can be used for testing Chrome Extensions.

Note

Extensions in Chrome / Chromium currently only work in non-headless mode.

The following is code for getting a handle to the background page of an extension whose source is located in ./my-extension:

const puppeteer = require('puppeteer');

(async () => {
  const pathToExtension = require('path').join(__dirname, 'my-extension');
  const browser = await puppeteer.launch({
    headless: false,
    args: [
      `--disable-extensions-except=${pathToExtension}`,
      `--load-extension=${pathToExtension}`,
    ],
  });
  const targets = await browser.targets();
  const backgroundPageTarget = targets.find(
    (target) => target.type() === 'background_page'
  );
  const backgroundPage = await backgroundPageTarget.page();
  // Test the background page as you would any other page.
  await browser.close();
})();

Note

It is not yet possible to test extension popups or content scripts.

class: Puppeteer

Puppeteer module provides a method to launch a Chromium instance. The following is a typical example of using Puppeteer to drive automation:

const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  await page.goto('https://www.google.com');
  // other actions...
  await browser.close();
})();

puppeteer.clearCustomQueryHandlers()

Clears all registered handlers.

puppeteer.connect(options)

  • options <Object>
    • browserWSEndpoint <?string> a browser websocket endpoint to connect to.
    • browserURL <?string> a browser URL to connect to, in format http://${host}:${port}. Use interchangeably with browserWSEndpoint to let Puppeteer fetch it from metadata endpoint.
    • ignoreHTTPSErrors <boolean> Whether to ignore HTTPS errors during navigation. Defaults to false.
    • defaultViewport <?Object> Sets a consistent viewport for each page. Defaults to an 800x600 viewport. null disables the default viewport.
      • width <number> page width in pixels.
      • height <number> page height in pixels.
      • deviceScaleFactor <number> Specify device scale factor (can be thought of as DPR). Defaults to 1.
      • isMobile <boolean> Whether the meta viewport tag is taken into account. Defaults to false.
      • hasTouch<boolean> Specifies if viewport supports touch events. Defaults to false
      • isLandscape <boolean> Specifies if viewport is in landscape mode. Defaults to false.
    • slowMo <number> Slows down Puppeteer operations by the specified amount of milliseconds. Useful so that you can see what is going on.
    • transport <ConnectionTransport> Experimental Specify a custom transport object for Puppeteer to use.
    • product <string> Possible values are: chrome, firefox. Defaults to chrome.
    • targetFilter <?function[Protocol.Target.TargetInfo]:boolean> Use this function to decide if Puppeteer should connect to the given target. If a targetFilter is provided, Puppeteer only connects to targets for which targetFilter returns true. By default, Puppeteer connects to all available targets.
  • returns: <Promise<Browser>>

This methods attaches Puppeteer to an existing browser instance.

puppeteer.createBrowserFetcher([options])

  • options <Object>
    • host <string> A download host to be used. Defaults to https://storage.googleapis.com. If the product is firefox, this defaults to https://archive.mozilla.org/pub/firefox/nightly/latest-mozilla-central.
    • path <string> A path for the downloads folder. Defaults to <root>/.local-chromium, where <root> is Puppeteer's package root. If the product is firefox, this defaults to <root>/.local-firefox.
    • platform <"linux"|"mac"|"win32"|"win64"> string for the current platform. Possible values are: mac, win32, win64, linux. Defaults to the current platform.
    • product <"chrome"|"firefox"> string for the product to run. Possible values are: chrome, firefox. Defaults to chrome.
  • returns: <BrowserFetcher>

puppeteer.customQueryHandlerNames()

  • returns: <Array> A list with the names of all registered custom query handlers.

puppeteer.defaultArgs([options])

  • options <Object> Set of configurable options to set on the browser. Can have the following fields:
    • headless <boolean|"chrome"> Whether to run browser in headless mode. Defaults to true unless the devtools option is true. "chrome" is a new experimental headless mode (use at your own risk).
    • args <Array<string>> Additional arguments to pass to the browser instance. The list of Chromium flags can be found here.
    • userDataDir <string> Path to a User Data Directory.
    • devtools <boolean> Whether to auto-open a DevTools panel for each tab. If this option is true, the headless option will be set false.
    • debuggingPort <number> Specify custom debugging port. Pass 0 to discover a random port. Defaults to 0.
  • returns: <Array<string>>

The default flags that Chromium will be launched with.

puppeteer.devices

Returns a list of devices to be used with page.emulate(options). Actual list of devices can be found in src/common/DeviceDescriptors.ts.

const puppeteer = require('puppeteer');
const iPhone = puppeteer.devices['iPhone 6'];

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  await page.emulate(iPhone);
  await page.goto('https://www.google.com');
  // other actions...
  await browser.close();
})();

puppeteer.errors

Puppeteer methods might throw errors if they are unable to fulfill a request. For example, page.waitForSelector(selector[, options]) might fail if the selector doesn't match any nodes during the given timeframe.

For certain types of errors Puppeteer uses specific error classes. These classes are available via puppeteer.errors

An example of handling a timeout error:

try {
  await page.waitForSelector('.foo');
} catch (e) {
  if (e instanceof puppeteer.errors.TimeoutError) {
    // Do something if this is a timeout.
  }
}

Note

The old way (Puppeteer versions <= v1.14.0) errors can be obtained with require('puppeteer/Errors').

puppeteer.executablePath()

Note

puppeteer.executablePath() is affected by the PUPPETEER_EXECUTABLE_PATH and PUPPETEER_CHROMIUM_REVISION env variables. See Environment Variables for details.

puppeteer.launch([options])

  • options <Object> Set of configurable options to set on the browser. Can have the following fields:
    • product <string> Which browser to launch. At this time, this is either chrome or firefox. See also PUPPETEER_PRODUCT.
    • ignoreHTTPSErrors <boolean> Whether to ignore HTTPS errors during navigation. Defaults to false.
    • headless <boolean|"chrome"> Whether to run browser in headless mode. Defaults to true unless the devtools option is true. "chrome" is a new experimental headless mode (use at your own risk).
    • channel <string> When specified, Puppeteer will search for the locally installed release channel of Google Chrome and use it to launch. Available values are chrome, chrome-beta, chrome-canary, chrome-dev. When channel is specified, executablePath cannot be specified.
    • executablePath <string> Path to a browser executable to run instead of the bundled Chromium. If executablePath is a relative path, then it is resolved relative to current working directory. BEWARE: Puppeteer is only guaranteed to work with the bundled Chromium, use at your own risk.
    • slowMo <number> Slows down Puppeteer operations by the specified amount of milliseconds. Useful so that you can see what is going on.
    • defaultViewport <?Object> Sets a consistent viewport for each page. Defaults to an 800x600 viewport. null disables the default viewport.
      • width <number> page width in pixels.
      • height <number> page height in pixels.
      • deviceScaleFactor <number> Specify device scale factor (can be thought of as DPR). Defaults to 1.
      • isMobile <boolean> Whether the meta viewport tag is taken into account. Defaults to false.
      • hasTouch<boolean> Specifies if viewport supports touch events. Defaults to false
      • isLandscape <boolean> Specifies if viewport is in landscape mode. Defaults to false.
    • args <Array<string>> Additional arguments to pass to the browser instance. The list of Chromium flags can be found here, and here is the list of Firefox flags.
    • ignoreDefaultArgs <boolean|Array<string>> If true, then do not use puppeteer.defaultArgs(). If an array is given, then filter out the given default arguments. Dangerous option; use with care. Defaults to false.
    • handleSIGINT <boolean> Close the browser process on Ctrl-C. Defaults to true.
    • handleSIGTERM <boolean> Close the browser process on SIGTERM. Defaults to true.
    • handleSIGHUP <boolean> Close the browser process on SIGHUP. Defaults to true.
    • timeout <number> Maximum time in milliseconds to wait for the browser instance to start. Defaults to 30000 (30 seconds). Pass 0 to disable timeout.
    • dumpio <boolean> Whether to pipe the browser process stdout and stderr into process.stdout and process.stderr. Defaults to false.
    • userDataDir <string> Path to a User Data Directory.
    • debuggingPort <number> Specify custom debugging port. Pass 0 to discover a random port. Defaults to 0.
    • env <Object> Specify environment variables that will be visible to the browser. Defaults to process.env.
    • devtools <boolean> Whether to auto-open a DevTools panel for each tab. If this option is true, the headless option will be set false.
    • pipe <boolean> Connects to the browser over a pipe instead of a WebSocket. Defaults to false.
    • extraPrefsFirefox <Object> Additional preferences that can be passed to Firefox (see PUPPETEER_PRODUCT)
    • targetFilter <?function[Protocol.Target.TargetInfo]:boolean> Use this function to decide if Puppeteer should connect to the given target. If a targetFilter is provided, Puppeteer only connects to targets for which targetFilter returns true. By default, Puppeteer connects to all available targets.
    • waitForInitialPage <boolean> Whether to wait for the initial page to be ready. Defaults to true.
  • returns: <Promise<Browser>> Promise which resolves to browser instance.

You can use ignoreDefaultArgs to filter out --mute-audio from default arguments:

const browser = await puppeteer.launch({
  ignoreDefaultArgs: ['--mute-audio'],
});

Note

Puppeteer can also be used to control the Chrome browser, but it works best with the version of Chromium it is bundled with. There is no guarantee it will work with any other version. Use executablePath or channel option with extreme caution.

If Google Chrome (rather than Chromium) is preferred, a Chrome Canary or Dev Channel build is suggested.

In puppeteer.launch([options]) above, any mention of Chromium also applies to Chrome.

See this article for a description of the differences between Chromium and Chrome. This article describes some differences for Linux users.

puppeteer.networkConditions

Returns a list of network conditions to be used with page.emulateNetworkConditions(networkConditions). Actual list of conditions can be found in src/common/NetworkConditions.ts.

const puppeteer = require('puppeteer');
const slow3G = puppeteer.networkConditions['Slow 3G'];

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  await page.emulateNetworkConditions(slow3G);
  await page.goto('https://www.google.com');
  // other actions...
  await browser.close();
})();

puppeteer.product

  • returns: <string> returns the name of the browser that is under automation ("chrome" or "firefox")

The product is set by the PUPPETEER_PRODUCT environment variable or the product option in puppeteer.launch([options]) and defaults to chrome. Firefox support is experimental and requires to install Puppeteer via PUPPETEER_PRODUCT=firefox npm i puppeteer.

puppeteer.registerCustomQueryHandler(name, queryHandler)

Registers a custom query handler.

Example:

puppeteer.registerCustomQueryHandler('getByClass', {
  queryOne: (element, selector) => {
    return element.querySelector(`.${selector}`);
  },
  queryAll: (element, selector) => {
    return element.querySelectorAll(`.${selector}`);
  },
});
const aHandle = await page.$('getByClass/…');

puppeteer.unregisterCustomQueryHandler(name)

  • name <string> The name of the query handler to unregister.

class: BrowserFetcher

BrowserFetcher can download and manage different versions of Chromium and Firefox.

BrowserFetcher operates on revision strings that specify a precise version of Chromium, e.g. "533271". Revision strings can be obtained from omahaproxy.appspot.com.

In the Firefox case, BrowserFetcher downloads Firefox Nightly and operates on version numbers such as "75".

An example of using BrowserFetcher to download a specific version of Chromium and running Puppeteer against it:

const browserFetcher = puppeteer.createBrowserFetcher();
const revisionInfo = await browserFetcher.download('533271');
const browser = await puppeteer.launch({
  executablePath: revisionInfo.executablePath,
});

Note

BrowserFetcher is not designed to work concurrently with other instances of BrowserFetcher that share the same downloads directory.

browserFetcher.canDownload(revision)

  • revision <string> a revision to check availability.
  • returns: <Promise<boolean>> returns true if the revision could be downloaded from the host.

The method initiates a HEAD request to check if the revision is available.

browserFetcher.download(revision[, progressCallback])

  • revision <string> a revision to download.
  • progressCallback <function(number, number)> A function that will be called with two arguments:
    • downloadedBytes <number> how many bytes have been downloaded
    • totalBytes <number> how large is the total download
  • returns: <Promise<Object>> Resolves with revision information when the revision is downloaded and extracted
    • revision <string> the revision the info was created from
    • folderPath <string> path to the extracted revision folder
    • executablePath <string> path to the revision executable
    • url <string> URL this revision can be downloaded from
    • local <boolean> whether the revision is locally available on disk

The method initiates a GET request to download the revision from the host.

browserFetcher.host()

  • returns: <string> The download host being used.

browserFetcher.localRevisions()

  • returns: <Promise<Array<string>>> A list of all revisions (for the current product) available locally on disk.

browserFetcher.platform()

  • returns: <string> One of mac, linux, win32 or win64.

browserFetcher.product()

  • returns: <string> One of chrome or firefox.

browserFetcher.remove(revision)

  • revision <string> a revision to remove for the current product. The method will throw if the revision has not been downloaded.
  • returns: <Promise> Resolves when the revision has been removed.

browserFetcher.revisionInfo(revision)

  • revision <string> a revision to get info for.
  • returns: <Object>
    • revision <string> the revision the info was created from
    • folderPath <string> path to the extracted revision folder
    • executablePath <string> path to the revision executable
    • url <string> URL this revision can be downloaded from
    • local <boolean> whether the revision is locally available on disk
    • product <string> one of chrome or firefox

Note

Many BrowserFetcher methods, like remove and revisionInfo are affected by the choice of product. See puppeteer.createBrowserFetcher([options]).

class: Browser

A Browser is created when Puppeteer connects to a Chromium instance, either through puppeteer.launch or puppeteer.connect.

An example of using a Browser to create a Page:

const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  await page.goto('https://example.com');
  await browser.close();
})();

An example of disconnecting from and reconnecting to a Browser:

const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  // Store the endpoint to be able to reconnect to Chromium
  const browserWSEndpoint = browser.wsEndpoint();
  // Disconnect puppeteer from Chromium
  browser.disconnect();

  // Use the endpoint to reestablish a connection
  const browser2 = await puppeteer.connect({ browserWSEndpoint });
  // Close Chromium
  await browser2.close();
})();

event: 'disconnected'

Emitted when Puppeteer gets disconnected from the Chromium instance. This might happen because of one of the following:

event: 'targetchanged'

Emitted when the URL of a target changes.

Note

This includes target changes in incognito browser contexts.

event: 'targetcreated'

Emitted when a target is created, for example when a new page is opened by window.open or browser.newPage.

Note

This includes target creations in incognito browser contexts.

event: 'targetdestroyed'

Emitted when a target is destroyed, for example when a page is closed.

Note

This includes target destructions in incognito browser contexts.

browser.browserContexts()

Returns an array of all open browser contexts. In a newly created browser, this will return a single instance of BrowserContext.

browser.close()

Closes Chromium and all of its pages (if any were opened). The Browser object itself is considered to be disposed and cannot be used anymore.

During the process of closing the browser, Puppeteer attempts to delete the temp folder created exclusively for this browser instance. If this fails (either because a file in the temp folder is locked by another process or because of insufficient permissions) an error is logged. This implies that: a) the folder and/or its content is not fully deleted; and b) the connection with the browser is not properly disposed (see browser.disconnect()).

browser.createIncognitoBrowserContext([options])

  • options <Object> Set of configurable options to set on the browserContext. Can have the following fields:
    • proxyServer <string> Optional proxy server with optional port to use for all requests. Username and password can be set in page.authenticate(credentials).
    • proxyBypassList <string> Optional: Bypass the proxy for the given semi-colon-separated list of hosts.
  • returns: <Promise<BrowserContext>>

Creates a new incognito browser context. This won't share cookies/cache with other browser contexts.

(async () => {
  const browser = await puppeteer.launch();
  // Create a new incognito browser context.
  const context = await browser.createIncognitoBrowserContext();
  // Create a new page in a pristine context.
  const page = await context.newPage();
  // Do stuff
  await page.goto('https://example.com');
})();

browser.defaultBrowserContext()

Returns the default browser context. The default browser context can not be closed.

browser.disconnect()

Disconnects Puppeteer from the browser but leaves the Chromium process running. After calling disconnect, the Browser object is considered disposed and cannot be used anymore.

browser.isConnected()

Indicates that the browser is connected.

browser.newPage()

Promise which resolves to a new Page object. The Page is created in a default browser context.

browser.pages()

  • returns: <Promise<Array<Page>>> Promise which resolves to an array of all open pages. Non visible pages, such as "background_page", will not be listed here. You can find them using target.page().

An array of all pages inside the Browser. In case of multiple browser contexts, the method will return an array with all the pages in all browser contexts.

browser.process()

browser.target()

A target associated with the browser.

browser.targets()

An array of all active targets inside the Browser. In case of multiple browser contexts, the method will return an array with all the targets in all browser contexts.

browser.userAgent()

  • returns: <Promise<string>> Promise which resolves to the browser's original user agent.

Note

Pages can override browser user agent with page.setUserAgent

browser.version()

  • returns: <Promise<string>> For headless Chromium, this is similar to HeadlessChrome/61.0.3153.0. For non-headless, this is similar to Chrome/61.0.3153.0.

Note

the format of browser.version() might change with future releases of Chromium.

browser.waitForTarget(predicate[, options])

  • predicate <function[Target]:boolean|[Promise]> A function to be run for every target
  • options <Object>
    • timeout <number> Maximum wait time in milliseconds. Pass 0 to disable the timeout. Defaults to 30 seconds.
  • returns: <Promise<Target>> Promise which resolves to the first target found that matches the predicate function.

This searches for a target in all browser contexts.

An example of finding a target for a page opened via window.open:

await page.evaluate(() => window.open('https://www.example.com/'));
const newWindowTarget = await browser.waitForTarget(
  (target) => target.url() === 'https://www.example.com/'
);

browser.wsEndpoint()

  • returns: <string> Browser websocket URL.

Browser websocket endpoint which can be used as an argument to puppeteer.connect. The format is ws://${host}:${port}/devtools/browser/<id>

You can find the webSocketDebuggerUrl from http://${host}:${port}/json/version. Learn more about the devtools protocol and the browser endpoint.

class: BrowserContext

BrowserContexts provide a way to operate multiple independent browser sessions. When a browser is launched, it has a single BrowserContext used by default. The method browser.newPage() creates a page in the default browser context.

If a page opens another page, e.g. with a window.open call, the popup will belong to the parent page's browser context.

Puppeteer allows creation of "incognito" browser contexts with browser.createIncognitoBrowserContext() method. "Incognito" browser contexts don't write any browsing data to disk.

// Create a new incognito browser context
const context = await browser.createIncognitoBrowserContext();
// Create a new page inside context.
const page = await context.newPage();
// ... do stuff with page ...
await page.goto('https://example.com');
// Dispose context once it's no longer needed.
await context.close();

event: 'targetchanged'

Emitted when the URL of a target inside the browser context changes.

event: 'targetcreated'

Emitted when a new target is created inside the browser context, for example when a new page is opened by window.open or browserContext.newPage.

event: 'targetdestroyed'

Emitted when a target inside the browser context is destroyed, for example when a page is closed.

browserContext.browser()

The browser this browser context belongs to.

browserContext.clearPermissionOverrides()

Clears all permission overrides for the browser context.

const context = browser.defaultBrowserContext();
context.overridePermissions('https://example.com', ['clipboard-read']);
// do stuff ..
context.clearPermissionOverrides();

browserContext.close()

Closes the browser context. All the targets that belong to the browser context will be closed.

Note

only incognito browser contexts can be closed.

browserContext.isIncognito()

Returns whether BrowserContext is incognito. The default browser context is the only non-incognito browser context.

Note

the default browser context cannot be closed.

browserContext.newPage()

Creates a new page in the browser context.

browserContext.overridePermissions(origin, permissions)

  • origin <string> The origin to grant permissions to, e.g. "https://example.com".
  • permissions <Array<string>> An array of permissions to grant. All permissions that are not listed here will be automatically denied. Permissions can be one of the following values:
    • 'geolocation'
    • 'midi'
    • 'midi-sysex' (system-exclusive midi)
    • 'notifications'
    • 'push'
    • 'camera'
    • 'microphone'
    • 'background-sync'
    • 'ambient-light-sensor'
    • 'accelerometer'
    • 'gyroscope'
    • 'magnetometer'
    • 'accessibility-events'
    • 'clipboard-read'
    • 'clipboard-write'
    • 'payment-handler'
    • 'persistent-storage'
  • returns: <Promise>
const context = browser.defaultBrowserContext();
await context.overridePermissions('https://html5demos.com', ['geolocation']);

browserContext.pages()

  • returns: <Promise<Array<Page>>> Promise which resolves to an array of all open pages. Non visible pages, such as "background_page", will not be listed here. You can find them using target.page().

An array of all pages inside the browser context.

browserContext.targets()

An array of all active targets inside the browser context.

browserContext.waitForTarget(predicate[, options])

  • predicate <function[Target]:boolean|[Promise]> A function to be run for every target
  • options <Object>
    • timeout <number> Maximum wait time in milliseconds. Pass 0 to disable the timeout. Defaults to 30 seconds.
  • returns: <Promise<Target>> Promise which resolves to the first target found that matches the predicate function.

This searches for a target in this specific browser context.

An example of finding a target for a page opened via window.open:

await page.evaluate(() => window.open('https://www.example.com/'));
const newWindowTarget = await browserContext.waitForTarget(
  (target) => target.url() === 'https://www.example.com/'
);

class: Page

Page provides methods to interact with a single tab or extension background page in Chromium. One Browser instance might have multiple Page instances.

This example creates a page, navigates it to a URL, and then saves a screenshot:

const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  await page.goto('https://example.com');
  await page.screenshot({ path: 'screenshot.png' });
  await browser.close();
})();

The Page class emits various events (described below) which can be handled using any of the EventEmitter methods, such as on, once or off.

This example logs a message for a single page load event:

page.once('load', () => console.log('Page loaded!'));

To unsubscribe from events use the off method:

function logRequest(interceptedRequest) {
  console.log('A request was made:', interceptedRequest.url());
}
page.on('request', logRequest);
// Sometime later...
page.off('request', logRequest);

event: 'close'

Emitted when the page closes.

event: 'console'

Emitted when JavaScript within the page calls one of console API methods, e.g. console.log or console.dir. Also emitted if the page throws an error or a warning.

The arguments passed into console.log appear as arguments on the event handler.

An example of handling console event:

page.on('console', (msg) => {
  for (let i = 0; i < msg.args().length; ++i)
    console.log(`${i}: ${msg.args()[i]}`);
});
page.evaluate(() => console.log('hello', 5, { foo: 'bar' }));

event: 'dialog'

Emitted when a JavaScript dialog appears, such as alert, prompt, confirm or beforeunload. Puppeteer can respond to the dialog via Dialog's accept or dismiss methods.

event: 'domcontentloaded'

Emitted when the JavaScript DOMContentLoaded event is dispatched.

event: 'error'

Emitted when the page crashes.

Note

error event has a special meaning in Node, see error events for details.

event: 'frameattached'

Emitted when a frame is attached.

event: 'framedetached'

Emitted when a frame is detached.

event: 'framenavigated'

Emitted when a frame is navigated to a new URL.

event: 'load'

Emitted when the JavaScript load event is dispatched.

event: 'metrics'

  • <Object>
    • title <string> The title passed to console.timeStamp.
    • metrics <Object> Object containing metrics as key/value pairs. The values of metrics are of <number> type.

Emitted when the JavaScript code makes a call to console.timeStamp. For the list of metrics see page.metrics.

event: 'pageerror'

  • <Error> The exception message

Emitted when an uncaught exception happens within the page.

event: 'popup'

  • <Page> Page corresponding to "popup" window

Emitted when the page opens a new tab or window.

const [popup] = await Promise.all([
  new Promise((resolve) => page.once('popup', resolve)),
  page.click('a[target=_blank]'),
]);

const [popup] = await Promise.all([
  new Promise((resolve) => page.once('popup', resolve)),
  page.evaluate(() => window.open('https://example.com')),
]);

event: 'request'

Emitted when a page issues a request. The HTTPRequest object is read-only. In order to intercept and mutate requests, see page.setRequestInterception.

event: 'requestfailed'

Emitted when a request fails, for example by timing out.

Note

HTTP Error responses, such as 404 or 503, are still successful responses from HTTP standpoint, so request will complete with 'requestfinished' event and not with 'requestfailed'.

event: 'requestfinished'

Emitted when a request finishes successfully.

event: 'response'

Emitted when a HTTPResponse is received.

event: 'workercreated'

Emitted when a dedicated WebWorker is spawned by the page.

event: 'workerdestroyed'

Emitted when a dedicated WebWorker is terminated.

page.$(selector)

The method runs document.querySelector within the page. If no element matches the selector, the return value resolves to null.

Shortcut for page.mainFrame().$(selector).

page.$$(selector)

The method runs document.querySelectorAll within the page. If no elements match the selector, the return value resolves to [].

Shortcut for page.mainFrame().$$(selector).

page.$$eval(selector, pageFunction[, ...args])

This method runs Array.from(document.querySelectorAll(selector)) within the page and passes it as the first argument to pageFunction.

If pageFunction returns a Promise, then page.$$eval would wait for the promise to resolve and return its value.

Examples:

const divCount = await page.$$eval('div', (divs) => divs.length);

const options = await page.$$eval('div > span.options', (options) =>
  options.map((option) => option.textContent)
);

page.$eval(selector, pageFunction[, ...args])

This method runs document.querySelector within the page and passes it as the first argument to pageFunction. If there's no element matching selector, the method throws an error.

If pageFunction returns a Promise, then page.$eval would wait for the promise to resolve and return its value.

Examples:

const searchValue = await page.$eval('#search', (el) => el.value);
const preloadHref = await page.$eval('link[rel=preload]', (el) => el.href);
const html = await page.$eval('.main-container', (e) => e.outerHTML);

Shortcut for page.mainFrame().$eval(selector, pageFunction).

page.$x(expression)

The method evaluates the XPath expression relative to the page document as its context node. If there are no such elements, the method resolves to an empty array.

Shortcut for page.mainFrame().$x(expression)

page.accessibility

page.addScriptTag(options)

  • options <Object>
    • url <string> URL of a script to be added.
    • path <string> Path to the JavaScript file to be injected into frame. If path is a relative path, then it is resolved relative to current working directory.
    • content <string> Raw JavaScript content to be injected into frame.
    • type <string> Script type. Use 'module' in order to load a Javascript ES6 module. See script for more details.
    • id <string> id attribute to add to the script tag.
  • returns: <Promise<ElementHandle>> which resolves to the added tag when the script's onload fires or when the script content was injected into frame.

Adds a <script> tag into the page with the desired URL or content.

Shortcut for page.mainFrame().addScriptTag(options).

page.addStyleTag(options)

  • options <Object>
    • url <string> URL of the <link> tag.
    • path <string> Path to the CSS file to be injected into frame. If path is a relative path, then it is resolved relative to current working directory.
    • content <string> Raw CSS content to be injected into frame.
  • returns: <Promise<ElementHandle>> which resolves to the added tag when the stylesheet's onload fires or when the CSS content was injected into frame.

Adds a <link rel="stylesheet"> tag into the page with the desired URL or a <style type="text/css"> tag with the content.

Shortcut for page.mainFrame().addStyleTag(options).

page.authenticate(credentials)

Provide credentials for HTTP authentication.

To disable authentication, pass null.

page.bringToFront()

Brings page to front (activates tab).

page.browser()

Get the browser the page belongs to.

page.browserContext()

Get the browser context that the page belongs to.

page.click(selector[, options])

  • selector <string> A selector to search for element to click. If there are multiple elements satisfying the selector, the first will be clicked.
  • options <Object>
    • button <"left"|"right"|"middle"> Defaults to left.
    • clickCount <number> defaults to 1. See UIEvent.detail.
    • delay <number> Time to wait between mousedown and mouseup in milliseconds. Defaults to 0.
  • returns: <Promise> Promise which resolves when the element matching selector is successfully clicked. The Promise will be rejected if there is no element matching selector.

This method fetches an element with selector, scrolls it into view if needed, and then uses page.mouse to click in the center of the element. If there's no element matching selector, the method throws an error.

Bear in mind that if click() triggers a navigation event and there's a separate page.waitForNavigation() promise to be resolved, you may end up with a race condition that yields unexpected results. The correct pattern for click and wait for navigation is the following:

const [response] = await Promise.all([
  page.waitForNavigation(waitOptions),
  page.click(selector, clickOptions),
]);

Shortcut for page.mainFrame().click(selector[, options]).

page.close([options])

By default, page.close() does not run beforeunload handlers.

Note

if runBeforeUnload is passed as true, a beforeunload dialog might be summoned and should be handled manually via page's 'dialog' event.

page.content()

Gets the full HTML contents of the page, including the doctype.

page.cookies([...urls])

If no URLs are specified, this method returns cookies for the current page URL. If URLs are specified, only cookies for those URLs are returned.

page.coverage

page.createPDFStream([options])

  • options <Object> Options object which might have the following properties:
    • path <string> The file path to save the PDF to. If path is a relative path, then it is resolved relative to current working directory. If no path is provided, the PDF won't be saved to the disk.
    • scale <number> Scale of the webpage rendering. Defaults to 1. Scale amount must be between 0.1 and 2.
    • displayHeaderFooter <boolean> Display header and footer. Defaults to false.
    • headerTemplate <string> HTML template for the print header. Should be valid HTML markup with following classes used to inject printing values into them:
      • date formatted print date
      • title document title
      • url document location
      • pageNumber current page number
      • totalPages total pages in the document
    • footerTemplate <string> HTML template for the print footer. Should use the same format as the headerTemplate.
    • printBackground <boolean> Print background graphics. Defaults to false.
    • landscape <boolean> Paper orientation. Defaults to false.
    • pageRanges <string> Paper ranges to print, e.g., '1-5, 8, 11-13'. Defaults to the empty string, which means print all pages.
    • format <string> Paper format. If set, takes priority over width or height options. Defaults to 'Letter'.
    • width <string|number> Paper width, accepts values labeled with units.
    • height <string|number> Paper height, accepts values labeled with units.
    • margin <Object> Paper margins, defaults to none.
      • top <string|number> Top margin, accepts values labeled with units.
      • right <string|number> Right margin, accepts values labeled with units.
      • bottom <string|number> Bottom margin, accepts values labeled with units.
      • left <string|number> Left margin, accepts values labeled with units.
    • preferCSSPageSize <boolean> Give any CSS @page size declared in the page priority over what is declared in width and height or format options. Defaults to false, which will scale the content to fit the paper size.
    • omitBackground <boolean> Hides default white background and allows capturing screenshots with transparency. Defaults to false.
  • returns: <Promise<[Readable]>> Promise which resolves with a Node.js stream for the PDF file.

Note

This method is identical to page.pdf, except it returns the PDF as a readable stream of binary data. If you are generating very large PDFs, it may be useful to use a stream to avoid high memory usage. This version will ignore the path option.

const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();

  // Stream a PDF into a file
  const pdfStream = await page.createPDFStream();
  const writeStream = fs.createWriteStream('test.pdf');
  pdfStream.pipe(writeStream);
  pdfStream.on('end', async () => {
    await browser.close();
  });
})();

page.deleteCookie(...cookies)

page.emulate(options)

  • options <Object>
    • viewport <Object>
      • width <number> page width in pixels.
      • height <number> page height in pixels.
      • deviceScaleFactor <number> Specify device scale factor (can be thought of as DPR). Defaults to 1.
      • isMobile <boolean> Whether the meta viewport tag is taken into account. Defaults to false.
      • hasTouch<boolean> Specifies if viewport supports touch events. Defaults to false
      • isLandscape <boolean> Specifies if viewport is in landscape mode. Defaults to false.
    • userAgent <string>
  • returns: <Promise>

Emulates given device metrics and user agent. This method is a shortcut for calling two methods:

To aid emulation, Puppeteer provides a list of device descriptors that can be obtained via the puppeteer.devices.

page.emulate will resize the page. A lot of websites don't expect phones to change size, so you should emulate before navigating to the page.

const puppeteer = require('puppeteer');
const iPhone = puppeteer.devices['iPhone 6'];

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  await page.emulate(iPhone);
  await page.goto('https://www.google.com');
  // other actions...
  await browser.close();
})();

List of all available devices is available in the source code: src/common/DeviceDescriptors.ts.

page.emulateCPUThrottling(factor)

  • factor <?number> Factor at which the CPU will be throttled (2x, 2.5x. 3x, ...). Passing null disables cpu throttling.
  • returns: <Promise>

Note

Real device CPU performance is impacted by many factors that are not trivial to emulate via the Chrome DevTools Protocol / Puppeteer. e.g core count, L1/L2 cache, thermal throttling impacting performance, architecture etc. Simulating CPU performance can be a good guideline, but ideally also verify any numbers you see on a real mobile device.

const puppeteer = require('puppeteer');
const slow3G = puppeteer.networkConditions['Slow 3G'];

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  await page.emulateCPUThrottling(2);
  await page.goto('https://www.google.com');
  // other actions...
  await browser.close();
})();

page.emulateIdleState(overrides)

  • overrides <?Object> If not set, clears emulation
  • returns: <Promise>

page.emulateMediaFeatures(features)

  • features <?Array<Object>> Given an array of media feature objects, emulates CSS media features on the page. Each media feature object must have the following properties:
    • name <string> The CSS media feature name. Supported names are 'prefers-colors-scheme', 'prefers-reduced-motion', and 'color-gamut'.
    • value <string> The value for the given CSS media feature.
  • returns: <Promise>
await page.emulateMediaFeatures([
  { name: 'prefers-color-scheme', value: 'dark' },
]);
await page.evaluate(() => matchMedia('(prefers-color-scheme: dark)').matches);
// → true
await page.evaluate(() => matchMedia('(prefers-color-scheme: light)').matches);
// → false

await page.emulateMediaFeatures([
  { name: 'prefers-reduced-motion', value: 'reduce' },
]);
await page.evaluate(
  () => matchMedia('(prefers-reduced-motion: reduce)').matches
);
// → true
await page.evaluate(
  () => matchMedia('(prefers-reduced-motion: no-preference)').matches
);
// → false

await page.emulateMediaFeatures([
  { name: 'prefers-color-scheme', value: 'dark' },
  { name: 'prefers-reduced-motion', value: 'reduce' },
]);
await page.evaluate(() => matchMedia('(prefers-color-scheme: dark)').matches);
// → true
await page.evaluate(() => matchMedia('(prefers-color-scheme: light)').matches);
// → false
await page.evaluate(
  () => matchMedia('(prefers-reduced-motion: reduce)').matches
);
// → true
await page.evaluate(
  () => matchMedia('(prefers-reduced-motion: no-preference)').matches
);
// → false

await page.emulateMediaFeatures([{ name: 'color-gamut', value: 'p3' }]);
await page.evaluate(() => matchMedia('(color-gamut: srgb)').matches);
// → true
await page.evaluate(() => matchMedia('(color-gamut: p3)').matches);
// → true
await page.evaluate(() => matchMedia('(color-gamut: rec2020)').matches);
// → false

page.emulateMediaType(type)

  • type <?string> Changes the CSS media type of the page. The only allowed values are 'screen', 'print' and null. Passing null disables CSS media emulation.
  • returns: <Promise>
await page.evaluate(() => matchMedia('screen').matches);
// → true
await page.evaluate(() => matchMedia('print').matches);
// → false

await page.emulateMediaType('print');
await page.evaluate(() => matchMedia('screen').matches);
// → false
await page.evaluate(() => matchMedia('print').matches);
// → true

await page.emulateMediaType(null);
await page.evaluate(() => matchMedia('screen').matches);
// → true
await page.evaluate(() => matchMedia('print').matches);
// → false

page.emulateNetworkConditions(networkConditions)

  • networkConditions <?Object> Passing null disables network condition emulation.
    • download <number> Download speed (bytes/s), -1 to disable
    • upload <number> Upload speed (bytes/s), -1 to disable
    • latency <number> Latency (ms), 0 to disable
  • returns: <Promise>

Note

This does not affect WebSockets and WebRTC PeerConnections (see https://crbug.com/563644). To set the page offline, you can use page.setOfflineMode(enabled).

const puppeteer = require('puppeteer');
const slow3G = puppeteer.networkConditions['Slow 3G'];

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  await page.emulateNetworkConditions(slow3G);
  await page.goto('https://www.google.com');
  // other actions...
  await browser.close();
})();

page.emulateTimezone(timezoneId)

  • timezoneId <?string> Changes the timezone of the page. See ICU’s metaZones.txt for a list of supported timezone IDs. Passing null disables timezone emulation.
  • returns: <Promise>

page.emulateVisionDeficiency(type)

  • type <?string> Simulates the given vision deficiency on the page. Supported vision deficiency types are 'achromatopsia', 'deuteranopia', 'protanopia', 'tritanopia', 'blurredVision', and 'none'.
  • returns: <Promise>
const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  await page.goto('https://v8.dev/blog/10-years');

  await page.emulateVisionDeficiency('achromatopsia');
  await page.screenshot({ path: 'achromatopsia.png' });

  await page.emulateVisionDeficiency('deuteranopia');
  await page.screenshot({ path: 'deuteranopia.png' });

  await page.emulateVisionDeficiency('blurredVision');
  await page.screenshot({ path: 'blurred-vision.png' });

  await browser.close();
})();

page.evaluate(pageFunction[, ...args])

If the function passed to the page.evaluate returns a Promise, then page.evaluate would wait for the promise to resolve and return its value.

If the function passed to the page.evaluate returns a non-Serializable value, then page.evaluate resolves to undefined. DevTools Protocol also supports transferring some additional values that are not serializable by JSON: -0, NaN, Infinity, -Infinity, and bigint literals.

Passing arguments to pageFunction:

const result = await page.evaluate((x) => {
  return Promise.resolve(8 * x);
}, 7);
console.log(result); // prints "56"

A string can also be passed in instead of a function:

console.log(await page.evaluate('1 + 2')); // prints "3"
const x = 10;
console.log(await page.evaluate(`1 + ${x}`)); // prints "11"

ElementHandle instances can be passed as arguments to the page.evaluate:

const bodyHandle = await page.$('body');
const html = await page.evaluate((body) => body.innerHTML, bodyHandle);
await bodyHandle.dispose();

Shortcut for page.mainFrame().evaluate(pageFunction, ...args).

page.evaluateHandle(pageFunction[, ...args])

The only difference between page.evaluate and page.evaluateHandle is that page.evaluateHandle returns in-page object (JSHandle).

If the function passed to the page.evaluateHandle returns a Promise, then page.evaluateHandle would wait for the promise to resolve and return its value.

A string can also be passed in instead of a function:

const aHandle = await page.evaluateHandle('document'); // Handle for the 'document'

JSHandle instances can be passed as arguments to the page.evaluateHandle:

const aHandle = await page.evaluateHandle(() => document.body);
const resultHandle = await page.evaluateHandle(
  (body) => body.innerHTML,
  aHandle
);
console.log(await resultHandle.jsonValue());
await resultHandle.dispose();

This function will return a JSHandle by default, however if your pageFunction returns an HTML element you will get back an ElementHandle:

const button = await page.evaluateHandle(() =>
  document.querySelector('button')
);
// button is an ElementHandle, so you can call methods such as click:
await button.click();

Shortcut for page.mainFrame().executionContext().evaluateHandle(pageFunction, ...args).

page.evaluateOnNewDocument(pageFunction[, ...args])

Adds a function which would be invoked in one of the following scenarios:

  • whenever the page is navigated
  • whenever the child frame is attached or navigated. In this case, the function is invoked in the context of the newly attached frame

The function is invoked after the document was created but before any of its scripts were run. This is useful to amend the JavaScript environment, e.g. to seed Math.random.

An example of overriding the navigator.languages property before the page loads:

// preload.js

// overwrite the `languages` property to use a custom getter
Object.defineProperty(navigator, 'languages', {
  get: function () {
    return ['en-US', 'en', 'bn'];
  },
});

// In your puppeteer script, assuming the preload.js file is in same folder of our script
const preloadFile = fs.readFileSync('./preload.js', 'utf8');
await page.evaluateOnNewDocument(preloadFile);

page.exposeFunction(name, puppeteerFunction)

  • name <string> Name of the function on the window object
  • puppeteerFunction <function> Callback function which will be called in Puppeteer's context. Can also be a module with a default export.
  • returns: <Promise>

The method adds a function called name on the page's window object. When called, the function executes puppeteerFunction in node.js and returns a Promise which resolves to the return value of puppeteerFunction.

If the puppeteerFunction returns a Promise, it will be awaited.

Note

Functions installed via page.exposeFunction survive navigations.

An example of adding an md5 function into the page:

const puppeteer = require('puppeteer');
const crypto = require('crypto');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  page.on('console', (msg) => console.log(msg.text()));
  await page.exposeFunction('md5', (text) =>
    crypto.createHash('md5').update(text).digest('hex')
  );
  await page.evaluate(async () => {
    // use window.md5 to compute hashes
    const myString = 'PUPPETEER';
    const myHash = await window.md5(myString);
    console.log(`md5 of ${myString} is ${myHash}`);
  });
  await browser.close();
})();

An example of adding a window.readfile function into the page:

const puppeteer = require('puppeteer');
const fs = require('fs');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  page.on('console', (msg) => console.log(msg.text()));
  await page.exposeFunction('readfile', async (filePath) => {
    return new Promise((resolve, reject) => {
      fs.readFile(filePath, 'utf8', (err, text) => {
        if (err) reject(err);
        else resolve(text);
      });
    });
  });
  await page.evaluate(async () => {
    // use window.readfile to read contents of a file
    const content = await window.readfile('/etc/hosts');
    console.log(content);
  });
  await browser.close();
})();

page.focus(selector)

  • selector <string> A selector of an element to focus. If there are multiple elements satisfying the selector, the first will be focused.
  • returns: <Promise> Promise which resolves when the element matching selector is successfully focused. The promise will be rejected if there is no element matching selector.

This method fetches an element with selector and focuses it. If there's no element matching selector, the method throws an error.

Shortcut for page.mainFrame().focus(selector).

page.frames()

  • returns: <Array<Frame>> An array of all frames attached to the page.

page.goBack([options])

  • options <Object> Navigation parameters which might have the following properties:
    • timeout <number> Maximum navigation time in milliseconds, defaults to 30 seconds, pass 0 to disable timeout. The default value can be changed by using the page.setDefaultNavigationTimeout(timeout) or page.setDefaultTimeout(timeout) methods.
    • waitUntil <"load"|"domcontentloaded"|"networkidle0"|"networkidle2"|Array> When to consider navigation succeeded, defaults to load. Given an array of event strings, navigation is considered to be successful after all events have been fired. Events can be either:
      • load - consider navigation to be finished when the load event is fired.
      • domcontentloaded - consider navigation to be finished when the DOMContentLoaded event is fired.
      • networkidle0 - consider navigation to be finished when there are no more than 0 network connections for at least 500 ms.
      • networkidle2 - consider navigation to be finished when there are no more than 2 network connections for at least 500 ms.
  • returns: <Promise<?HTTPResponse>> Promise which resolves to the main resource response. In case of multiple redirects, the navigation will resolve with the response of the last redirect. If can not go back, resolves to null.

Navigate to the previous page in history.

page.goForward([options])

  • options <Object> Navigation parameters which might have the following properties:
    • timeout <number> Maximum navigation time in milliseconds, defaults to 30 seconds, pass 0 to disable timeout. The default value can be changed by using the page.setDefaultNavigationTimeout(timeout) or page.setDefaultTimeout(timeout) methods.
    • waitUntil <"load"|"domcontentloaded"|"networkidle0"|"networkidle2"|Array> When to consider navigation succeeded, defaults to load. Given an array of event strings, navigation is considered to be successful after all events have been fired. Events can be either:
      • load - consider navigation to be finished when the load event is fired.
      • domcontentloaded - consider navigation to be finished when the DOMContentLoaded event is fired.
      • networkidle0 - consider navigation to be finished when there are no more than 0 network connections for at least 500 ms.
      • networkidle2 - consider navigation to be finished when there are no more than 2 network connections for at least 500 ms.
  • returns: <Promise<?HTTPResponse>> Promise which resolves to the main resource response. In case of multiple redirects, the navigation will resolve with the response of the last redirect. If can not go forward, resolves to null.

Navigate to the next page in history.

page.goto(url[, options])

  • url <string> URL to navigate page to. The URL should include scheme, e.g. https://.
  • options <Object> Navigation parameters which might have the following properties:
    • timeout <number> Maximum navigation time in milliseconds, defaults to 30 seconds, pass 0 to disable timeout. The default value can be changed by using the page.setDefaultNavigationTimeout(timeout) or page.setDefaultTimeout(timeout) methods.
    • waitUntil <"load"|"domcontentloaded"|"networkidle0"|"networkidle2"|Array> When to consider navigation succeeded, defaults to load. Given an array of event strings, navigation is considered to be successful after all events have been fired. Events can be either:
      • load - consider navigation to be finished when the load event is fired.
      • domcontentloaded - consider navigation to be finished when the DOMContentLoaded event is fired.
      • networkidle0 - consider navigation to be finished when there are no more than 0 network connections for at least 500 ms.
      • networkidle2 - consider navigation to be finished when there are no more than 2 network connections for at least 500 ms.
    • referer <string> Referer header value. If provided it will take preference over the referer header value set by page.setExtraHTTPHeaders().
  • returns: <Promise<?HTTPResponse>> Promise which resolves to the main resource response. In case of multiple redirects, the navigation will resolve with the response of the last redirect.

page.goto will throw an error if:

  • there's an SSL error (e.g. in case of self-signed certificates).
  • target URL is invalid.
  • the timeout is exceeded during navigation.
  • the remote server does not respond or is unreachable.
  • the main resource failed to load.

page.goto will not throw an error when any valid HTTP status code is returned by the remote server, including 404 "Not Found" and 500 "Internal Server Error". The status code for such responses can be retrieved by calling response.status().

Note

page.goto either throws an error or returns a main resource response. The only exceptions are navigation to about:blank or navigation to the same URL with a different hash, which would succeed and return null.

Note

Headless mode doesn't support navigation to a PDF document. See the upstream issue.

Shortcut for page.mainFrame().goto(url, options)

page.hover(selector)

  • selector <string> A selector to search for element to hover. If there are multiple elements satisfying the selector, the first will be hovered.
  • returns: <Promise> Promise which resolves when the element matching selector is successfully hovered. Promise gets rejected if there's no element matching selector.

This method fetches an element with selector, scrolls it into view if needed, and then uses page.mouse to hover over the center of the element. If there's no element matching selector, the method throws an error.

Shortcut for page.mainFrame().hover(selector).

page.isClosed()

Indicates that the page has been closed.

page.isDragInterceptionEnabled()

Indicates that drag events are being intercepted.

page.isJavaScriptEnabled()

Returns true if the page has JavaScript enabled, false otherwise.

page.keyboard

page.mainFrame()

  • returns: <Frame> The page's main frame.

Page is guaranteed to have a main frame which persists during navigations.

page.metrics()

  • returns: <Promise<Object>> Object containing metrics as key/value pairs.
    • Timestamp <number> The timestamp when the metrics sample was taken.
    • Documents <number> Number of documents in the page.
    • Frames <number> Number of frames in the page.
    • JSEventListeners <number> Number of events in the page.
    • Nodes <number> Number of DOM nodes in the page.
    • LayoutCount <number> Total number of full or partial page layout.
    • RecalcStyleCount <number> Total number of page style recalculations.
    • LayoutDuration <number> Combined durations of all page layouts.
    • RecalcStyleDuration <number> Combined duration of all page style recalculations.
    • ScriptDuration <number> Combined duration of JavaScript execution.
    • TaskDuration <number> Combined duration of all tasks performed by the browser.
    • JSHeapUsedSize <number> Used JavaScript heap size.
    • JSHeapTotalSize <number> Total JavaScript heap size.

Note

All timestamps are in monotonic time: monotonically increasing time in seconds since an arbitrary point in the past.

page.mouse

page.pdf([options])

  • options <Object> Options object which might have the following properties:
    • path <string> The file path to save the PDF to. If path is a relative path, then it is resolved relative to current working directory. If no path is provided, the PDF won't be saved to the disk.
    • scale <number> Scale of the webpage rendering. Defaults to 1. Scale amount must be between 0.1 and 2.
    • displayHeaderFooter <boolean> Display header and footer. Defaults to false.
    • headerTemplate <string> HTML template for the print header. Should be valid HTML markup with following classes used to inject printing values into them:
      • date formatted print date
      • title document title
      • url document location
      • pageNumber current page number
      • totalPages total pages in the document
    • footerTemplate <string> HTML template for the print footer. Should use the same format as the headerTemplate.
    • printBackground <boolean> Print background graphics. Defaults to false.
    • landscape <boolean> Paper orientation. Defaults to false.
    • pageRanges <string> Paper ranges to print, e.g., '1-5, 8, 11-13'. Defaults to the empty string, which means print all pages.
    • format <string> Paper format. If set, takes priority over width or height options. Defaults to 'Letter'.
    • width <string|number> Paper width, accepts values labeled with units.
    • height <string|number> Paper height, accepts values labeled with units.
    • margin <Object> Paper margins, defaults to none.
      • top <string|number> Top margin, accepts values labeled with units.
      • right <string|number> Right margin, accepts values labeled with units.
      • bottom <string|number> Bottom margin, accepts values labeled with units.
      • left <string|number> Left margin, accepts values labeled with units.
    • preferCSSPageSize <boolean> Give any CSS @page size declared in the page priority over what is declared in width and height or format options. Defaults to false, which will scale the content to fit the paper size.
    • omitBackground <boolean> Hides default white background and allows capturing screenshots with transparency. Defaults to false.
    • timeout <number> Maximum time in milliseconds, defaults to 30 seconds, pass 0 to disable timeout.
  • returns: <Promise<Buffer>> Promise which resolves with PDF buffer.

Note

Generating a pdf is currently only supported in Chrome headless.

page.pdf() generates a pdf of the page with print CSS media. To generate a pdf with screen media, call page.emulateMediaType('screen') before calling page.pdf():

Note

By default, page.pdf() generates a pdf with modified colors for printing. Use the -webkit-print-color-adjust property to force rendering of exact colors.

Note

If you are generating very large PDFs, it may be useful to use the streaming version of this function (page.createPDFStream) to avoid high memory usage.

// Generates a PDF with 'screen' media type.
await page.emulateMediaType('screen');
await page.pdf({ path: 'page.pdf' });

The width, height, and margin options accept values labeled with units. Unlabeled values are treated as pixels.

A few examples:

  • page.pdf({ width: 100 }) - prints with width set to 100 pixels
  • page.pdf({ width: '100px' }) - prints with width set to 100 pixels
  • page.pdf({ width: '10cm' }) - prints with width set to 10 centimeters.

All possible units are:

  • px - pixel
  • in - inch
  • cm - centimeter
  • mm - millimeter

The format options are:

  • Letter: 8.5in x 11in
  • Legal: 8.5in x 14in
  • Tabloid: 11in x 17in
  • Ledger: 17in x 11in
  • A0: 33.1in x 46.8in
  • A1: 23.4in x 33.1in
  • A2: 16.54in x 23.4in
  • A3: 11.7in x 16.54in
  • A4: 8.27in x 11.7in
  • A5: 5.83in x 8.27in
  • A6: 4.13in x 5.83in

Note

headerTemplate and footerTemplate markup have the following limitations:

  1. Script tags inside templates are not evaluated.
  2. Page styles are not visible inside templates.

page.queryObjects(prototypeHandle)

  • prototypeHandle <JSHandle> A handle to the object prototype.
  • returns: <Promise<JSHandle>> Promise which resolves to a handle to an array of objects with this prototype.

The method iterates the JavaScript heap and finds all the objects with the given prototype.

// Create a Map object
await page.evaluate(() => (window.map = new Map()));
// Get a handle to the Map object prototype
const mapPrototype = await page.evaluateHandle(() => Map.prototype);
// Query all map instances into an array
const mapInstances = await page.queryObjects(mapPrototype);
// Count amount of map objects in heap
const count = await page.evaluate((maps) => maps.length, mapInstances);
await mapInstances.dispose();
await mapPrototype.dispose();

Shortcut for page.mainFrame().executionContext().queryObjects(prototypeHandle).

page.reload([options])

  • options <Object> Navigation parameters which might have the following properties:
    • timeout <number> Maximum navigation time in milliseconds, defaults to 30 seconds, pass 0 to disable timeout. The default value can be changed by using the page.setDefaultNavigationTimeout(timeout) or page.setDefaultTimeout(timeout) methods.
    • waitUntil <"load"|"domcontentloaded"|"networkidle0"|"networkidle2"|Array> When to consider navigation succeeded, defaults to load. Given an array of event strings, navigation is considered to be successful after all events have been fired. Events can be either:
      • load - consider navigation to be finished when the load event is fired.
      • domcontentloaded - consider navigation to be finished when the DOMContentLoaded event is fired.
      • networkidle0 - consider navigation to be finished when there are no more than 0 network connections for at least 500 ms.
      • networkidle2 - consider navigation to be finished when there are no more than 2 network connections for at least 500 ms.
  • returns: <Promise<HTTPResponse>> Promise which resolves to the main resource response. In case of multiple redirects, the navigation will resolve with the response of the last redirect.

page.screenshot([options])

  • options <Object> Options object which might have the following properties:
    • path <string> The file path to save the image to. The screenshot type will be inferred from file extension. If path is a relative path, then it is resolved relative to current working directory. If no path is provided, the image won't be saved to the disk.
    • type <string> Specify screenshot type, can be either jpeg, png or webp. Defaults to 'png'.
    • quality <number> The quality of the image, between 0-100. Not applicable to png images.
    • fullPage <boolean> When true, takes a screenshot of the full scrollable page. Defaults to false.
    • clip <Object> An object which specifies clipping region of the page. Should have the following fields:
      • x <number> x-coordinate of top-left corner of clip area
      • y <number> y-coordinate of top-left corner of clip area
      • width <number> width of clipping area
      • height <number> height of clipping area
    • omitBackground <boolean> Hides default white background and allows capturing screenshots with transparency. Defaults to false.
    • encoding <string> The encoding of the image, can be either base64 or binary. Defaults to binary.
    • captureBeyondViewport <boolean> When true, captures screenshot beyond the viewport. When false, falls back to old behaviour, and cuts the screenshot by the viewport size. Defaults to true.
  • returns: <Promise<string|Buffer>> Promise which resolves to buffer or a base64 string (depending on the value of encoding) with captured screenshot.

Note

Screenshots take at least 1/6 second on OS X. See https://crbug.com/741689 for discussion.

page.select(selector, ...values)

  • selector <string> A selector to query page for
  • ...values <...string> Values of options to select. If the <select> has the multiple attribute, all values are considered, otherwise only the first one is taken into account.
  • returns: <Promise<Array<string>>> An array of option values that have been successfully selected.

Triggers a change and input event once all the provided options have been selected. If there's no <select> element matching selector, the method throws an error.

page.select('select#colors', 'blue'); // single selection
page.select('select#colors', 'red', 'green', 'blue'); // multiple selections

Shortcut for page.mainFrame().select()

page.setBypassCSP(enabled)

  • enabled <boolean> sets bypassing of page's Content-Security-Policy.
  • returns: <Promise>

Toggles bypassing page's Content-Security-Policy.

Note

CSP bypassing happens at the moment of CSP initialization rather than evaluation. Usually, this means that page.setBypassCSP should be called before navigating to the domain.

page.setCacheEnabled([enabled])

  • enabled <boolean> sets the enabled state of the cache.
  • returns: <Promise>

Toggles ignoring cache for each request based on the enabled state. By default, caching is enabled.

page.setContent(html[, options])

  • html <string> HTML markup to assign to the page.
  • options <Object> Parameters that might have the following properties:
    • timeout <number> Maximum time in milliseconds for resources to load, defaults to 30 seconds, pass 0 to disable timeout. The default value can be changed by using the page.setDefaultNavigationTimeout(timeout) or page.setDefaultTimeout(timeout) methods.
    • waitUntil <"load"|"domcontentloaded"|"networkidle0"|"networkidle2"|Array> When to consider setting markup succeeded, defaults to load. Given an array of event strings, setting content is considered to be successful after all events have been fired. Events can be either:
      • load - consider setting content to be finished when the load event is fired.
      • domcontentloaded - consider setting content to be finished when the DOMContentLoaded event is fired.
      • networkidle0 - consider setting content to be finished when there are no more than 0 network connections for at least 500 ms.
      • networkidle2 - consider setting content to be finished when there are no more than 2 network connections for at least 500 ms.
  • returns: <Promise>

page.setCookie(...cookies)

await page.setCookie(cookieObject1, cookieObject2);

page.setDefaultNavigationTimeout(timeout)

  • timeout <number> Maximum navigation time in milliseconds

This setting will change the default maximum navigation time for the following methods and related shortcuts:

Note

page.setDefaultNavigationTimeout takes priority over page.setDefaultTimeout

page.setDefaultTimeout(timeout)

  • timeout <number> Maximum time in milliseconds

This setting will change the default maximum time for the following methods and related shortcuts:

Note

page.setDefaultNavigationTimeout takes priority over page.setDefaultTimeout

page.setDragInterception(enabled)

Enables the Input.drag methods. This provides the capability to capture drag events emitted on the page, which can then be used to simulate drag-and-drop.

page.setExtraHTTPHeaders(headers)

  • headers <Object> An object containing additional HTTP headers to be sent with every request. All header values must be strings.
  • returns: <Promise>

The extra HTTP headers will be sent with every request the page initiates.

Note

All HTTP header names are lowercased. (HTTP headers are case-insensitive, so this shouldn’t impact your server code.)

Note

page.setExtraHTTPHeaders does not guarantee the order of headers in the outgoing requests.

page.setGeolocation(options)

Sets the page's geolocation.

await page.setGeolocation({ latitude: 59.95, longitude: 30.31667 });

Note

Consider using browserContext.overridePermissions to grant permissions for the page to read its geolocation.

page.setJavaScriptEnabled(enabled)

  • enabled <boolean> Whether or not to enable JavaScript on the page.
  • returns: <Promise>

Note

changing this value won't affect scripts that have already been run. It will take full effect on the next navigation.

page.setOfflineMode(enabled)

  • enabled <boolean> When true, enables offline mode for the page.
  • returns: <Promise>

Note

while this method sets the network connection to offline, it does not change the parameters used in page.emulateNetworkConditions(networkConditions).

page.setRequestInterception(value)

  • value <boolean> Whether to enable request interception.
  • returns: <Promise>

Activating request interception enables request.abort, request.continue and request.respond methods. This provides the capability to modify network requests that are made by a page.

Once request interception is enabled, every request will stall unless it's continued, responded or aborted.

An example of a naïve request interceptor that aborts all image requests:

const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  await page.setRequestInterception(true);
  page.on('request', (interceptedRequest) => {
    if (interceptedRequest.isInterceptResolutionHandled()) return;
    if (
      interceptedRequest.url().endsWith('.png') ||
      interceptedRequest.url().endsWith('.jpg')
    )
      interceptedRequest.abort();
    else interceptedRequest.continue();
  });
  await page.goto('https://example.com');
  await browser.close();
})();
Multiple Intercept Handlers and Asynchronous Resolutions

By default Puppeteer will raise a Request is already handled! exception if request.abort, request.continue, or request.respond are called after any of them have already been called.

Always assume that an unknown handler may have already called abort/continue/respond. Even if your handler is the only one you registered, 3rd party packages may register their own handlers. It is therefore important to always check the resolution status using request.isInterceptResolutionHandled before calling abort/continue/respond.

Importantly, the intercept resolution may get handled by another listener while your handler is awaiting an asynchronous operation. Therefore, the return value of request.isInterceptResolutionHandled is only safe in a synchronous code block. Always execute request.isInterceptResolutionHandled and abort/continue/respond synchronously together.

This example demonstrates two synchronous handlers working together:

/*
This first handler will succeed in calling request.continue because the request interception has never been resolved.
*/
page.on('request', (interceptedRequest) => {
  if (interceptedRequest.isInterceptResolutionHandled()) return;
  interceptedRequest.continue();
});

/*
This second handler will return before calling request.abort because request.continue was already
called by the first handler.
*/
page.on('request', (interceptedRequest) => {
  if (interceptedRequest.isInterceptResolutionHandled()) return;
  interceptedRequest.abort();
});

This example demonstrates asynchronous handlers working together:

/*
This first handler will succeed in calling request.continue because the request interception has never been resolved.
*/
page.on('request', (interceptedRequest) => {
  // The interception has not been handled yet. Control will pass through this guard.
  if (interceptedRequest.isInterceptResolutionHandled()) return;

  // It is not strictly necessary to return a promise, but doing so will allow Puppeteer to await this handler.
  return new Promise((resolve) => {
    // Continue after 500ms
    setTimeout(() => {
      // Inside, check synchronously to verify that the intercept wasn't handled already.
      // It might have been handled during the 500ms while the other handler awaited an async op of its own.
      if (interceptedRequest.isInterceptResolutionHandled()) {
        resolve();
        return;
      }
      interceptedRequest.continue();
      resolve();
    }, 500);
  });
});
page.on('request', async (interceptedRequest) => {
  // The interception has not been handled yet. Control will pass through this guard.
  if (interceptedRequest.isInterceptResolutionHandled()) return;

  await someLongAsyncOperation();
  // The interception *MIGHT* have been handled by the first handler, we can't be sure.
  // Therefore, we must check again before calling continue() or we risk Puppeteer raising an exception.
  if (interceptedRequest.isInterceptResolutionHandled()) return;
  interceptedRequest.continue();
});

For finer-grained introspection (see Cooperative Intercept Mode below), you may also call request.interceptResolutionState synchronously before using abort/continue/respond.

Here is the example above rewritten using request.interceptResolutionState

/*
This first handler will succeed in calling request.continue because the request interception has never been resolved.
*/
page.on('request', (interceptedRequest) => {
  // The interception has not been handled yet. Control will pass through this guard.
  const { action } = interceptedRequest.interceptResolutionState();
  if (action === InterceptResolutionAction.AlreadyHandled) return;

  // It is not strictly necessary to return a promise, but doing so will allow Puppeteer to await this handler.
  return new Promise((resolve) => {
    // Continue after 500ms
    setTimeout(() => {
      // Inside, check synchronously to verify that the intercept wasn't handled already.
      // It might have been handled during the 500ms while the other handler awaited an async op of its own.
      const { action } = interceptedRequest.interceptResolutionState();
      if (action === InterceptResolutionAction.AlreadyHandled) {
        resolve();
        return;
      }
      interceptedRequest.continue();
      resolve();
    }, 500);
  });
});
page.on('request', async (interceptedRequest) => {
  // The interception has not been handled yet. Control will pass through this guard.
  if (
    interceptedRequest.interceptResolutionState().action ===
    InterceptResolutionAction.AlreadyHandled
  )
    return;

  await someLongAsyncOperation();
  // The interception *MIGHT* have been handled by the first handler, we can't be sure.
  // Therefore, we must check again before calling continue() or we risk Puppeteer raising an exception.
  if (
    interceptedRequest.interceptResolutionState().action ===
    InterceptResolutionAction.AlreadyHandled
  )
    return;
  interceptedRequest.continue();
});
Cooperative Intercept Mode

request.abort, request.continue, and request.respond can accept an optional priority to work in Cooperative Intercept Mode. When all handlers are using Cooperative Intercept Mode, Puppeteer guarantees that all intercept handlers will run and be awaited in order of registration. The interception is resolved to the highest-priority resolution. Here are the rules of Cooperative Intercept Mode:

  • All resolutions must supply a numeric priority argument to abort/continue/respond.
  • If any resolution does not supply a numeric priority, Legacy Mode is active and Cooperative Intercept Mode is inactive.
  • Async handlers finish before intercept resolution is finalized.
  • The highest priority interception resolution "wins", i.e. the interception is ultimately aborted/responded/continued according to which resolution was given the highest priority.
  • In the event of a tie, abort > respond > continue.

For standardization, when specifying a Cooperative Intercept Mode priority use 0 or DEFAULT_INTERCEPT_RESOLUTION_PRIORITY (exported from HTTPRequest) unless you have a clear reason to use a higher priority. This gracefully prefers respond over continue and abort over respond and allows other handlers to work cooperatively. If you do intentionally want to use a different priority, higher priorities win over lower priorities. Negative priorities are allowed. For example, continue({}, 4) would win over continue({}, -2).

To preserve backward compatibility, any handler resolving the intercept without specifying priority (Legacy Mode) causes immediate resolution. For Cooperative Intercept Mode to work, all resolutions must use a priority. In practice, this means you must still test for request.isInterceptResolutionHandled because a handler beyond your control may have called abort/continue/respond without a priority (Legacy Mode).

In this example, Legacy Mode prevails and the request is aborted immediately because at least one handler omits priority when resolving the intercept:

// Final outcome: immediate abort()
page.setRequestInterception(true);
page.on('request', (request) => {
  if (request.isInterceptResolutionHandled()) return;

  // Legacy Mode: interception is aborted immediately.
  request.abort('failed');
});
page.on('request', (request) => {
  if (request.isInterceptResolutionHandled()) return;
  // Control will never reach this point because the request was already aborted in Legacy Mode

  // Cooperative Intercept Mode: votes for continue at priority 0.
  request.continue({}, 0);
});

In this example, Legacy Mode prevails and the request is continued because at least one handler does not specify a priority:

// Final outcome: immediate continue()
page.setRequestInterception(true);
page.on('request', (request) => {
  if (request.isInterceptResolutionHandled()) return;

  // Cooperative Intercept Mode: votes to abort at priority 0.
  request.abort('failed', 0);
});
page.on('request', (request) => {
  if (request.isInterceptResolutionHandled()) return;

  // Control reaches this point because the request was cooperatively aborted which postpones resolution.

  // { action: InterceptResolutionAction.Abort, priority: 0 }, because abort @ 0 is the current winning resolution
  console.log(request.interceptResolutionState());

  // Legacy Mode: intercept continues immediately.
  request.continue({});
});
page.on('request', (request) => {
  // { action: InterceptResolutionAction.AlreadyHandled }, because continue in Legacy Mode was called
  console.log(request.interceptResolutionState());
});

In this example, Cooperative Intercept Mode is active because all handlers specify a priority. continue() wins because it has a higher priority than abort().

// Final outcome: cooperative continue() @ 5
page.setRequestInterception(true);
page.on('request', (request) => {
  if (request.isInterceptResolutionHandled()) return;

  // Cooperative Intercept Mode: votes to abort at priority 10
  request.abort('failed', 0);
});
page.on('request', (request) => {
  if (request.isInterceptResolutionHandled()) return;

  // Cooperative Intercept Mode: votes to continue at priority 5
  request.continue(request.continueRequestOverrides(), 5);
});
page.on('request', (request) => {
  // { action: InterceptResolutionAction.Continue, priority: 5 }, because continue @ 5 > abort @ 0
  console.log(request.interceptResolutionState());
});

In this example, Cooperative Intercept Mode is active because all handlers specify priority. respond() wins because its priority ties with continue(), but respond() beats continue().

// Final outcome: cooperative respond() @ 15
page.setRequestInterception(true);
page.on('request', (request) => {
  if (request.isInterceptResolutionHandled()) return;

  // Cooperative Intercept Mode: votes to abort at priority 10
  request.abort('failed', 10);
});
page.on('request', (request) => {
  if (request.isInterceptResolutionHandled()) return;

  // Cooperative Intercept Mode: votes to continue at priority 15
  request.continue(request.continueRequestOverrides(), 15);
});
page.on('request', (request) => {
  if (request.isInterceptResolutionHandled()) return;

  // Cooperative Intercept Mode: votes to respond at priority 15
  request.respond(request.responseForRequest(), 15);
});
page.on('request', (request) => {
  if (request.isInterceptResolutionHandled()) return;

  // Cooperative Intercept Mode: votes to respond at priority 12
  request.respond(request.responseForRequest(), 12);
});
page.on('request', (request) => {
  // { action: InterceptResolutionAction.Respond, priority: 15 }, because respond @ 15 > continue @ 15 > respond @ 12 > abort @ 10
  console.log(request.interceptResolutionState());
});
Cooperative Request Continuation

Puppeteer requires request.continue() to be called explicitly or the request will hang. Even if your handler means to take no special action, or 'opt out', request.continue() must still be called.

With the introduction of Cooperative Intercept Mode, two use cases arise for cooperative request continuations: Unopinionated and Opinionated.

The first case (common) is that your handler means to opt out of doing anything special the request. It has no opinion on further action and simply intends to continue by default and/or defer to other handlers that might have an opinion. But in case there are no other handlers, we must call request.continue() to ensure that the request doesn't hang.

We call this an Unopinionated continuation because the intent is to continue the request if nobody else has a better idea. Use request.continue({...}, DEFAULT_INTERCEPT_RESOLUTION_PRIORITY) (or 0) for this type of continuation.

The second case (uncommon) is that your handler actually does have an opinion and means to force continuation by overriding a lower-priority abort() or respond() issued elsewhere. We call this an Opinionated continuation. In these rare cases where you mean to specify an overriding continuation priority, use a custom priority.

To summarize, reason through whether your use of request.continue is just meant to be default/bypass behavior vs falling within the intended use case of your handler. Consider using a custom priority for in-scope use cases, and a default priority otherwise. Be aware that your handler may have both Opinionated and Unopinionated cases.

Upgrading to Cooperative Intercept Mode for package maintainers

If you are package maintainer and your package uses intercept handlers, you can update your intercept handlers to use Cooperative Intercept Mode. Suppose you have the following existing handler:

page.on('request', (interceptedRequest) => {
  if (request.isInterceptResolutionHandled()) return;
  if (
    interceptedRequest.url().endsWith('.png') ||
    interceptedRequest.url().endsWith('.jpg')
  )
    interceptedRequest.abort();
  else interceptedRequest.continue();
});

To use Cooperative Intercept Mode, upgrade continue() and abort():

page.on('request', (interceptedRequest) => {
  if (request.isInterceptResolutionHandled()) return;
  if (
    interceptedRequest.url().endsWith('.png') ||
    interceptedRequest.url().endsWith('.jpg')
  )
    interceptedRequest.abort('failed', 0);
  else
    interceptedRequest.continue(
      interceptedRequest.continueRequestOverrides(),
      0
    );
});

With those simple upgrades, your handler now uses Cooperative Intercept Mode instead.

However, we recommend a slightly more robust solution because the above introduces several subtle issues:

  1. Backward compatibility. If any handler still uses a Legacy Mode resolution (ie, does not specify a priority), that handler will resolve the interception immediately even if your handler runs first. This could cause disconcerting behavior for your users because suddenly your handler is not resolving the interception and a different handler is taking priority when all the user did was upgrade your package.
  2. Hard-coded priority. Your package user has no ability to specify the default resolution priority for your handlers. This can become important when the user wishes to manipulate the priorities based on use case. For example, one user might want your package to take a high priority while another user might want it to take a low priority.

To resolve both of these issues, our recommended approach is to export a setInterceptResolutionConfig() from your package. The user can then call setInterceptResolutionConfig() to explicitly activate Cooperative Intercept Mode in your package so they aren't surprised by changes in how the interception is resolved. They can also optionally specify a custom priority using setInterceptResolutionConfig(priority) that works for their use case:

// Defaults to undefined which preserves Legacy Mode behavior
let _priority = undefined;

// Export a module configuration function
export const setInterceptResolutionConfig = (priority = 0) =>
  (_priority = priority);

/**
 * Note that this handler uses `DEFAULT_INTERCEPT_RESOLUTION_PRIORITY` to "pass" on this request. It is important to use
 * the default priority when your handler has no opinion on the request and the intent is to continue() by default.
 */
page.on('request', (interceptedRequest) => {
  if (request.isInterceptResolutionHandled()) return;
  if (
    interceptedRequest.url().endsWith('.png') ||
    interceptedRequest.url().endsWith('.jpg')
  )
    interceptedRequest.abort('failed', _priority);
  else
    interceptedRequest.continue(
      interceptedRequest.continueRequestOverrides(),
      DEFAULT_INTERCEPT_RESOLUTION_PRIORITY // Unopinionated continuation
    );
});

If your package calls for more fine-grained control over resolution priorities, use a config pattern like this:

interface InterceptResolutionConfig {
  abortPriority?: number;
  continuePriority?: number;
}

// This approach supports multiple priorities based on situational
// differences. You could, for example, create a config that
// allowed separate priorities for PNG vs JPG.
const DEFAULT_CONFIG: InterceptResolutionConfig = {
  abortPriority: undefined, // Default to Legacy Mode
  continuePriority: undefined, // Default to Legacy Mode
};

// Defaults to undefined which preserves Legacy Mode behavior
let _config: Partial<InterceptResolutionConfig> = {};

export const setInterceptResolutionConfig = (
  config: InterceptResolutionConfig
) => (_config = { ...DEFAULT_CONFIG, ...config });

page.on('request', (interceptedRequest) => {
  if (request.isInterceptResolutionHandled()) return;
  if (
    interceptedRequest.url().endsWith('.png') ||
    interceptedRequest.url().endsWith('.jpg')
  ) {
    interceptedRequest.abort('failed', _config.abortPriority);
  } else {
    // Here we use a custom-configured priority to allow for Opinionated
    // continuation.
    // We would only want to allow this if we had a very clear reason why
    // some use cases required Opinionated continuation.
    interceptedRequest.continue(
      interceptedRequest.continueRequestOverrides(),
      _config.continuePriority // Why would we ever want priority!==0 here?
    );
  }
});

The above solutions ensure backward compatibility while also allowing the user to adjust the importance of your package in the resolution chain when Cooperative Intercept Mode is being used. Your package continues to work as expected until the user has fully upgraded their code and all third party packages to use Cooperative Intercept Mode. If any handler or package still uses Legacy Mode, your package can still operate in Legacy Mode too.

page.setUserAgent(userAgent[, userAgentMetadata])

  • userAgent <string> Specific user agent to use in this page
  • userAgentMetadata <Object> Optional user agent data to use in this page. Any values not provided will use the client's default.
    • brands <Array<Object>> Optional brand information
      • brand <string> Browser or client brand name.
      • version <string> Browser or client major version.
    • fullVersion <string> Optional browser or client full version.
    • platform <string> Operating system name.
    • platformVersion <string> Operating system version.
    • architecture <string> CPU architecture.
    • model <string> Device model.
    • mobile <boolean> Indicate if this is a mobile device.
  • returns: <Promise> Promise which resolves when the user agent is set.

Note

support for userAgentMetadata is experimental in the DevTools protocol and more properties will be added.

Providing the optional userAgentMetadata header will update the related entries in navigator.userAgentData and associated Sec-CH-UA* headers.

const page = await browser.newPage();
await page.setUserAgent('MyBrowser', {
  architecture: 'My1',
  mobile: false,
  model: 'Mybook',
  platform: 'MyOS',
  platformVersion: '3.1',
});

page.setViewport(viewport)

  • viewport <Object>
    • width <number> page width in pixels. required
    • height <number> page height in pixels. required
    • deviceScaleFactor <number> Specify device scale factor (can be thought of as DPR). Defaults to 1.
    • isMobile <boolean> Whether the meta viewport tag is taken into account. Defaults to false.
    • hasTouch<boolean> Specifies if viewport supports touch events. Defaults to false
    • isLandscape <boolean> Specifies if viewport is in landscape mode. Defaults to false.
  • returns: <Promise>

Note

in certain cases, setting viewport will reload the page in order to set the isMobile or hasTouch properties.

In the case of multiple pages in a single browser, each page can have its own viewport size.

page.setViewport will resize the page. A lot of websites don't expect phones to change size, so you should set the viewport before navigating to the page.

const page = await browser.newPage();
await page.setViewport({
  width: 640,
  height: 480,
  deviceScaleFactor: 1,
});
await page.goto('https://example.com');

page.tap(selector)

  • selector <string> A selector to search for element to tap. If there are multiple elements satisfying the selector, the first will be tapped.
  • returns: <Promise>

This method fetches an element with selector, scrolls it into view if needed, and then uses page.touchscreen to tap in the center of the element. If there's no element matching selector, the method throws an error.

Shortcut for page.mainFrame().tap(selector).

page.target()

  • returns: <Target> a target this page was created from.

page.title()

Shortcut for page.mainFrame().title().

page.touchscreen

page.tracing

page.type(selector, text[, options])

  • selector <string> A selector of an element to type into. If there are multiple elements satisfying the selector, the first will be used.
  • text <string> A text to type into a focused element.
  • options <Object>
    • delay <number> Time to wait between key presses in milliseconds. Defaults to 0.
  • returns: <Promise>

Sends a keydown, keypress/input, and keyup event for each character in the text.

To press a special key, like Control or ArrowDown, use keyboard.press.

await page.type('#mytextarea', 'Hello'); // Types instantly
await page.type('#mytextarea', 'World', { delay: 100 }); // Types slower, like a user

Shortcut for page.mainFrame().type(selector, text[, options]).

page.url()

This is a shortcut for page.mainFrame().url()

page.viewport()

  • returns: <?Object>
    • width <number> page width in pixels.
    • height <number> page height in pixels.
    • deviceScaleFactor <number> Specify device scale factor (can be though of as dpr). Defaults to 1.
    • isMobile <boolean> Whether the meta viewport tag is taken into account. Defaults to false.
    • hasTouch<boolean> Specifies if viewport supports touch events. Defaults to false
    • isLandscape <boolean> Specifies if viewport is in landscape mode. Defaults to false.

page.waitFor(selectorOrFunctionOrTimeout[, options[, ...args]])

  • selectorOrFunctionOrTimeout <string|number|function> A selector, predicate or timeout to wait for
  • options <Object> Optional waiting parameters
    • visible <boolean> wait for element to be present in DOM and to be visible. Defaults to false.
    • timeout <number> maximum time to wait for in milliseconds. Defaults to 30000 (30 seconds). Pass 0 to disable timeout. The default value can be changed by using the page.setDefaultTimeout(timeout) method.
    • hidden <boolean> wait for element to not be found in the DOM or to be hidden. Defaults to false.
    • polling <string|number> An interval at which the pageFunction is executed, defaults to raf. If polling is a number, then it is treated as an interval in milliseconds at which the function would be executed. If polling is a string, then it can be one of the following values:
      • raf - to constantly execute pageFunction in requestAnimationFrame callback. This is the tightest polling mode which is suitable to observe styling changes.
      • mutation - to execute pageFunction on every DOM mutation.
  • ...args <...Serializable|JSHandle> Arguments to pass to pageFunction
  • returns: <Promise<JSHandle>> Promise which resolves to a JSHandle of the success value

This method is deprecated. You should use the more explicit API methods available:

  • page.waitForSelector
  • page.waitForXPath
  • page.waitForFunction
  • page.waitForTimeout

This method behaves differently with respect to the type of the first parameter:

  • if selectorOrFunctionOrTimeout is a string, then the first argument is treated as a selector or xpath, depending on whether or not it starts with '//', and the method is a shortcut for page.waitForSelector or page.waitForXPath
  • if selectorOrFunctionOrTimeout is a function, then the first argument is treated as a predicate to wait for and the method is a shortcut for page.waitForFunction().
  • if selectorOrFunctionOrTimeout is a number, then the first argument is treated as a timeout in milliseconds and the method returns a promise which resolves after the timeout
  • otherwise, an exception is thrown
// wait for selector
await page.waitFor('.foo');
// wait for 1 second
await page.waitFor(1000);
// wait for predicate
await page.waitFor(() => !!document.querySelector('.foo'));

To pass arguments from node.js to the predicate of page.waitFor function:

const selector = '.foo';
await page.waitFor(
  (selector) => !!document.querySelector(selector),
  {},
  selector
);

Shortcut for page.mainFrame().waitFor(selectorOrFunctionOrTimeout[, options[, ...args]]).

page.waitForFileChooser([options])

Note

In non-headless Chromium, this method results in the native file picker dialog not showing up for the user.

This method is typically coupled with an action that triggers file choosing. The following example clicks a button that issues a file chooser and then responds with /tmp/myfile.pdf as if a user has selected this file.

const [fileChooser] = await Promise.all([
  page.waitForFileChooser(),
  page.click('#upload-file-button'), // some button that triggers file selection
]);
await fileChooser.accept(['/tmp/myfile.pdf']);

Note

This must be called before the file chooser is launched. It will not return a currently active file chooser.

Note

“File picker” refers to the operating system’s file selection UI that lets you browse to a folder and select file(s) to be shared with the web app. It’s not the “Save file” dialog.

page.waitForFrame(urlOrPredicate[, options])

  • urlOrPredicate <string|Function> A URL or predicate to wait for.
  • options <Object> Optional waiting parameters
    • timeout <number> Maximum wait time in milliseconds, defaults to 30 seconds, pass 0 to disable the timeout. The default value can be changed by using the page.setDefaultTimeout(timeout) method.
  • returns: <Promise<Frame>> Promise which resolves to the matched frame.
const frame = await page.waitForFrame(async (frame) => {
  return frame.name() === 'Test';
});

page.waitForFunction(pageFunction[, options[, ...args]])

  • pageFunction <function|string> Function to be evaluated in browser context
  • options <Object> Optional waiting parameters
    • polling <string|number> An interval at which the pageFunction is executed, defaults to raf. If polling is a number, then it is treated as an interval in milliseconds at which the function would be executed. If polling is a string, then it can be one of the following values:
      • raf - to constantly execute pageFunction in requestAnimationFrame callback. This is the tightest polling mode which is suitable to observe styling changes.
      • mutation - to execute pageFunction on every DOM mutation.
    • timeout <number> maximum time to wait for in milliseconds. Defaults to 30000 (30 seconds). Pass 0 to disable timeout. The default value can be changed by using the page.setDefaultTimeout(timeout) method.
  • ...args <...Serializable|JSHandle> Arguments to pass to pageFunction
  • returns: <Promise<JSHandle>> Promise which resolves when the pageFunction returns a truthy value. It resolves to a JSHandle of the truthy value.

The waitForFunction can be used to observe viewport size change:

const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  const watchDog = page.waitForFunction('window.innerWidth < 100');
  await page.setViewport({ width: 50, height: 50 });
  await watchDog;
  await browser.close();
})();

To pass arguments from node.js to the predicate of page.waitForFunction function:

const selector = '.foo';
await page.waitForFunction(
  (selector) => !!document.querySelector(selector),
  {},
  selector
);

The predicate of page.waitForFunction can be asynchronous too:

const username = 'github-username';
await page.waitForFunction(
  async (username) => {
    const githubResponse = await fetch(
      `https://api.github.com/users/${username}`
    );
    const githubUser = await githubResponse.json();
    // show the avatar
    const img = document.createElement('img');
    img.src = githubUser.avatar_url;
    // wait 3 seconds
    await new Promise((resolve, reject) => setTimeout(resolve, 3000));
    img.remove();
  },
  {},
  username
);

Shortcut for page.mainFrame().waitForFunction(pageFunction[, options[, ...args]]).

page.waitForNavigation([options])

  • options <Object> Navigation parameters which might have the following properties:
    • timeout <number> Maximum navigation time in milliseconds, defaults to 30 seconds, pass 0 to disable timeout. The default value can be changed by using the page.setDefaultNavigationTimeout(timeout) or page.setDefaultTimeout(timeout) methods.
    • waitUntil <"load"|"domcontentloaded"|"networkidle0"|"networkidle2"|Array> When to consider navigation succeeded, defaults to load. Given an array of event strings, navigation is considered to be successful after all events have been fired. Events can be either:
      • load - consider navigation to be finished when the load event is fired.
      • domcontentloaded - consider navigation to be finished when the DOMContentLoaded event is fired.
      • networkidle0 - consider navigation to be finished when there are no more than 0 network connections for at least 500 ms.
      • networkidle2 - consider navigation to be finished when there are no more than 2 network connections for at least 500 ms.
  • returns: <Promise<?HTTPResponse>> Promise which resolves to the main resource response. In case of multiple redirects, the navigation will resolve with the response of the last redirect. In case of navigation to a different anchor or navigation due to History API usage, the navigation will resolve with null.

This resolves when the page navigates to a new URL or reloads. It is useful when you run code that will indirectly cause the page to navigate. Consider this example:

const [response] = await Promise.all([
  page.waitForNavigation(), // The promise resolves after navigation has finished
  page.click('a.my-link'), // Clicking the link will indirectly cause a navigation
]);

NOTE Usage of the History API to change the URL is considered a navigation.

Shortcut for page.mainFrame().waitForNavigation(options).

page.waitForNetworkIdle([options])

  • options <Object> Optional waiting parameters
    • timeout <number> Maximum wait time in milliseconds, defaults to 30 seconds, pass 0 to disable the timeout. The default value can be changed by using the page.setDefaultTimeout(timeout) method.
    • idleTime <number> How long to wait for no network requests in milliseconds, defaults to 500 milliseconds.
  • returns: <Promise> Promise which resolves when network is idle.
page.evaluate(() => fetch('some-url'));
page.waitForNetworkIdle(); // The promise resolves after fetch above finishes

page.waitForRequest(urlOrPredicate[, options])

  • urlOrPredicate <string|Function> A URL or predicate to wait for.
  • options <Object> Optional waiting parameters
    • timeout <number> Maximum wait time in milliseconds, defaults to 30 seconds, pass 0 to disable the timeout. The default value can be changed by using the page.setDefaultTimeout(timeout) method.
  • returns: <Promise<HTTPRequest>> Promise which resolves to the matched request.
const firstRequest = await page.waitForRequest('http://example.com/resource');
const finalRequest = await page.waitForRequest(
  (request) =>
    request.url() === 'http://example.com' && request.method() === 'GET'
);
return firstRequest.url();

page.waitForResponse(urlOrPredicate[, options])

  • urlOrPredicate <string|Function> A URL or predicate to wait for.
  • options <Object> Optional waiting parameters
    • timeout <number> Maximum wait time in milliseconds, defaults to 30 seconds, pass 0 to disable the timeout. The default value can be changed by using the page.setDefaultTimeout(timeout) method.
  • returns: <Promise<HTTPResponse>> Promise which resolves to the matched response.
const firstResponse = await page.waitForResponse(
  'https://example.com/resource'
);
const finalResponse = await page.waitForResponse(
  (response) =>
    response.url() === 'https://example.com' && response.status() === 200
);
const finalResponse = await page.waitForResponse(async (response) => {
  return (await response.text()).includes('<html>');
});
return finalResponse.ok();

page.waitForSelector(selector[, options])

  • selector <string> A selector of an element to wait for
  • options <Object> Optional waiting parameters
    • visible <boolean> wait for element to be present in DOM and to be visible, i.e. to not have display: none or visibility: hidden CSS properties. Defaults to false.
    • hidden <boolean> wait for element to not be found in the DOM or to be hidden, i.e. have display: none or visibility: hidden CSS properties. Defaults to false.
    • timeout <number> maximum time to wait for in milliseconds. Defaults to 30000 (30 seconds). Pass 0 to disable timeout. The default value can be changed by using the page.setDefaultTimeout(timeout) method.
  • returns: <Promise<?ElementHandle>> Promise which resolves when element specified by selector string is added to DOM. Resolves to null if waiting for hidden: true and selector is not found in DOM.

Wait for the selector to appear in page. If at the moment of calling the method the selector already exists, the method will return immediately. If the selector doesn't appear after the timeout milliseconds of waiting, the function will throw.

This method works across navigations:

const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  let currentURL;
  page
    .waitForSelector('img')
    .then(() => console.log('First URL with image: ' + currentURL));
  for (currentURL of [
    'https://example.com',
    'https://google.com',
    'https://bbc.com',
  ]) {
    await page.goto(currentURL);
  }
  await browser.close();
})();

Shortcut for page.mainFrame().waitForSelector(selector[, options]).

page.waitForTimeout(milliseconds)

  • milliseconds <number> The number of milliseconds to wait for.
  • returns: <Promise> Promise which resolves after the timeout has been completed.

Pauses script execution for the given number of milliseconds before continuing:

const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  page.waitForTimeout(1000).then(() => console.log('Waited a second!'));

  await browser.close();
})();

page.waitForXPath(xpath[, options])

  • xpath <string> A xpath of an element to wait for
  • options <Object> Optional waiting parameters
    • visible <boolean> wait for element to be present in DOM and to be visible, i.e. to not have display: none or visibility: hidden CSS properties. Defaults to false.
    • hidden <boolean> wait for element to not be found in the DOM or to be hidden, i.e. have display: none or visibility: hidden CSS properties. Defaults to false.
    • timeout <number> maximum time to wait for in milliseconds. Defaults to 30000 (30 seconds). Pass 0 to disable timeout. The default value can be changed by using the page.setDefaultTimeout(timeout) method.
  • returns: <Promise<?ElementHandle>> Promise which resolves when element specified by xpath string is added to DOM. Resolves to null if waiting for hidden: true and xpath is not found in DOM.

Wait for the xpath to appear in page. If at the moment of calling the method the xpath already exists, the method will return immediately. If the xpath doesn't appear after the timeout milliseconds of waiting, the function will throw.

This method works across navigations:

const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  let currentURL;
  page
    .waitForXPath('//img')
    .then(() => console.log('First URL with image: ' + currentURL));
  for (currentURL of [
    'https://example.com',
    'https://google.com',
    'https://bbc.com',
  ]) {
    await page.goto(currentURL);
  }
  await browser.close();
})();

Shortcut for page.mainFrame().waitForXPath(xpath[, options]).

page.workers()

Note

This does not contain ServiceWorkers

GeolocationOptions

  • latitude <number> Latitude between -90 and 90.
  • longitude <number> Longitude between -180 and 180.
  • accuracy <number> Optional non-negative accuracy value.

WaitTimeoutOptions

  • timeout <number> Maximum wait time in milliseconds, defaults to 30 seconds, pass 0 to disable the timeout. The default value can be changed by using the page.setDefaultTimeout(timeout) method.

class: WebWorker

The WebWorker class represents a WebWorker. The events workercreated and workerdestroyed are emitted on the page object to signal the worker lifecycle.

page.on('workercreated', (worker) =>
  console.log('Worker created: ' + worker.url())
);
page.on('workerdestroyed', (worker) =>
  console.log('Worker destroyed: ' + worker.url())
);

console.log('Current workers:');
for (const worker of page.workers()) console.log('  ' + worker.url());

webWorker.evaluate(pageFunction[, ...args])

If the function passed to the worker.evaluate returns a Promise, then worker.evaluate would wait for the promise to resolve and return its value.

If the function passed to the worker.evaluate returns a non-Serializable value, then worker.evaluate resolves to undefined. DevTools Protocol also supports transferring some additional values that are not serializable by JSON: -0, NaN, Infinity, -Infinity, and bigint literals.

Shortcut for (await worker.executionContext()).evaluate(pageFunction, ...args).

webWorker.evaluateHandle(pageFunction[, ...args])

The only difference between worker.evaluate and worker.evaluateHandle is that worker.evaluateHandle returns in-page object (JSHandle).

If the function passed to the worker.evaluateHandle returns a Promise, then worker.evaluateHandle would wait for the promise to resolve and return its value.

If the function returns an element, the returned handle is an ElementHandle.

Shortcut for (await worker.executionContext()).evaluateHandle(pageFunction, ...args).

webWorker.executionContext()

webWorker.url()

class: Accessibility

The Accessibility class provides methods for inspecting Chromium's accessibility tree. The accessibility tree is used by assistive technology such as screen readers or switches.

Accessibility is a very platform-specific thing. On different platforms, there are different screen readers that might have wildly different output.

Blink - Chrome's rendering engine - has a concept of "accessibility tree", which is then translated into different platform-specific APIs. Accessibility namespace gives users access to the Blink Accessibility Tree.

Most of the accessibility tree gets filtered out when converting from Blink AX Tree to Platform-specific AX-Tree or by assistive technologies themselves. By default, Puppeteer tries to approximate this filtering, exposing only the "interesting" nodes of the tree.

accessibility.snapshot([options])

  • options <Object>
    • interestingOnly <boolean> Prune uninteresting nodes from the tree. Defaults to true.
    • root <ElementHandle> The root DOM element for the snapshot. Defaults to the whole page.
  • returns: <Promise<Object>> An AXNode object with the following properties:
    • role <string> The role.
    • name <string> A human readable name for the node.
    • value <string|number> The current value of the node.
    • description <string> An additional human readable description of the node.
    • keyshortcuts <string> Keyboard shortcuts associated with this node.
    • roledescription <string> A human readable alternative to the role.
    • valuetext <string> A description of the current value.
    • disabled <boolean> Whether the node is disabled.
    • expanded <boolean> Whether the node is expanded or collapsed.
    • focused <boolean> Whether the node is focused.
    • modal <boolean> Whether the node is modal.
    • multiline <boolean> Whether the node text input supports multiline.
    • multiselectable <boolean> Whether more than one child can be selected.
    • readonly <boolean> Whether the node is read-only.
    • required <boolean> Whether the node is required.
    • selected <boolean> Whether the node is selected in its parent node.
    • checked <boolean|"mixed"> Whether the checkbox is checked, or "mixed".
    • pressed <boolean|"mixed"> Whether the toggle button is checked, or "mixed".
    • level <number> The level of a heading.
    • valuemin <number> The minimum value in a node.
    • valuemax <number> The maximum value in a node.
    • autocomplete <string> What kind of autocomplete is supported by a control.
    • haspopup <string> What kind of popup is currently being shown for a node.
    • invalid <string> Whether and in what way this node's value is invalid.
    • orientation <string> Whether the node is oriented horizontally or vertically.
    • children <Array<Object>> Child AXNodes of this node, if any.

Captures the current state of the accessibility tree. The returned object represents the root accessible node of the page.

Note

The Chromium accessibility tree contains nodes that go unused on most platforms and by most screen readers. Puppeteer will discard them as well for an easier to process tree, unless interestingOnly is set to false.

An example of dumping the entire accessibility tree:

const snapshot = await page.accessibility.snapshot();
console.log(snapshot);

An example of logging the focused node's name:

const snapshot = await page.accessibility.snapshot();
const node = findFocusedNode(snapshot);
console.log(node && node.name);

function findFocusedNode(node) {
  if (node.focused) return node;
  for (const child of node.children || []) {
    const foundNode = findFocusedNode(child);
    return foundNode;
  }
  return null;
}

class: Keyboard

Keyboard provides an API for managing a virtual keyboard. The high-level API is keyboard.type, which takes raw characters and generates proper keydown, keypress/input, and keyup events on your page.

For finer control, you can use keyboard.down, keyboard.up, and keyboard.sendCharacter to manually fire events as if they were generated from a real keyboard.

An example of holding down Shift in order to select and delete some text:

await page.keyboard.type('Hello World!');
await page.keyboard.press('ArrowLeft');

await page.keyboard.down('Shift');
for (let i = 0; i < ' World'.length; i++)
  await page.keyboard.press('ArrowLeft');
await page.keyboard.up('Shift');

await page.keyboard.press('Backspace');
// Result text will end up saying 'Hello!'

An example of pressing A

await page.keyboard.down('Shift');
await page.keyboard.press('KeyA');
await page.keyboard.up('Shift');

Note

On macOS, keyboard shortcuts like ⌘ A -> Select All does not work. See #1313

keyboard.down(key[, options])

  • key <string> Name of key to press, such as ArrowLeft. See USKeyboardLayout for a list of all key names.
  • options <Object>
    • text <string> If specified, generates an input event with this text.
  • returns: <Promise>

Dispatches a keydown event.

If key is a single character and no modifier keys besides Shift are being held down, a keypress/input event will also be generated. The text option can be specified to force an input event to be generated.

If key is a modifier key, Shift, Meta, Control, or Alt, subsequent key presses will be sent with that modifier active. To release the modifier key, use keyboard.up.

After the key is pressed once, subsequent calls to keyboard.down will have repeat set to true. To release the key, use keyboard.up.

Note

Modifier keys DO influence keyboard.down. Holding down Shift will type the text in upper case.

keyboard.press(key[, options])

  • key <string> Name of key to press, such as ArrowLeft. See USKeyboardLayout for a list of all key names.
  • options <Object>
    • text <string> If specified, generates an input event with this text.
    • delay <number> Time to wait between keydown and keyup in milliseconds. Defaults to 0.
  • returns: <Promise>

If key is a single character and no modifier keys besides Shift are being held down, a keypress/input event will also be generated. The text option can be specified to force an input event to be generated.

Note

Modifier keys DO affect keyboard.press. Holding down Shift will type the text in upper case.

Shortcut for keyboard.down and keyboard.up.

keyboard.sendCharacter(char)

  • char <string> Character to send into the page.
  • returns: <Promise>

Dispatches a keypress and input event. This does not send a keydown or keyup event.

page.keyboard.sendCharacter('嗨');

Note

Modifier keys DO NOT affect keyboard.sendCharacter. Holding down Shift will not type the text in upper case.

keyboard.type(text[, options])

  • text <string> A text to type into a focused element.
  • options <Object>
    • delay <number> Time to wait between key presses in milliseconds. Defaults to 0.
  • returns: <Promise>

Sends a keydown, keypress/input, and keyup event for each character in the text.

To press a special key, like Control or ArrowDown, use keyboard.press.

await page.keyboard.type('Hello'); // Types instantly
await page.keyboard.type('World', { delay: 100 }); // Types slower, like a user

Note

Modifier keys DO NOT affect keyboard.type. Holding down Shift will not type the text in upper case.

keyboard.up(key)

Dispatches a keyup event.

class: Mouse

The Mouse class operates in main-frame CSS pixels relative to the top-left corner of the viewport.

Every page object has its own Mouse, accessible with page.mouse.

// Using ‘page.mouse’ to trace a 100x100 square.
await page.mouse.move(0, 0);
await page.mouse.down();
await page.mouse.move(0, 100);
await page.mouse.move(100, 100);
await page.mouse.move(100, 0);
await page.mouse.move(0, 0);
await page.mouse.up();

Note that the mouse events trigger synthetic MouseEvents. This means that it does not fully replicate the functionality of what a normal user would be able to do with their mouse.

For example, dragging and selecting text is not possible using page.mouse. Instead, you can use the DocumentOrShadowRoot.getSelection() functionality implemented in the platform.

For example, if you want to select all content between nodes:

await page.evaluate(
  (from, to) => {
    const selection = from.getRootNode().getSelection();
    const range = document.createRange();
    range.setStartBefore(from);
    range.setEndAfter(to);
    selection.removeAllRanges();
    selection.addRange(range);
  },
  fromJSHandle,
  toJSHandle
);

If you then would want to copy-paste your selection, you can use the clipboard API:

// The clipboard API does not allow you to copy, unless the tab is focused.
await page.bringToFront();
await page.evaluate(() => {
  // Copy the selected content to the clipboard
  document.execCommand('copy');
  // Obtain the content of the clipboard as a string
  return navigator.clipboard.readText();
});

Note that if you want access to the clipboard API, you have to give it permission to do so:

await browser
  .defaultBrowserContext()
  .overridePermissions('<your origin>', ['clipboard-read', 'clipboard-write']);

mouse.click(x, y[, options])

  • x <number>
  • y <number>
  • options <Object>
    • button <"left"|"right"|"middle"> Defaults to left.
    • clickCount <number> defaults to 1. See UIEvent.detail.
    • delay <number> Time to wait between mousedown and mouseup in milliseconds. Defaults to 0.
  • returns: <Promise>

Shortcut for mouse.move, mouse.down and mouse.up.

mouse.down([options])

Dispatches a mousedown event.

mouse.drag(start, target)

  • start <Object> the position to start dragging from
  • target <Object> the position to drag to
  • returns: <[Promise<[DragData]>]>

This method creates and captures a dragevent from a given point.

mouse.dragAndDrop(start, target[, options])

  • start <Object>
  • target <Object>
  • options <Object>
    • delay <number> how long to delay before dropping onto the target point
  • returns: <[Promise<[DragData]>]>

This method drags from a given start point and drops onto a target point.

mouse.dragEnter(target, data)

This method triggers a dragenter event from the target point.

mouse.dragOver(target, data)

This method triggers a dragover event from the target point.

mouse.drop(target, data)

This method triggers a drop event from the target point.

mouse.move(x, y[, options])

Dispatches a mousemove event.

mouse.up([options])

Dispatches a mouseup event.

mouse.wheel([options])

  • options <Object>
    • deltaX X delta in CSS pixels for mouse wheel event (default: 0). Positive values emulate a scroll right and negative values a scroll left event.
    • deltaY Y delta in CSS pixels for mouse wheel event (default: 0). Positive values emulate a scroll down and negative values a scroll up event.
  • returns: <Promise>

Dispatches a mousewheel event.

Examples:

await page.goto(
  'https://mdn.mozillademos.org/en-US/docs/Web/API/Element/wheel_event$samples/Scaling_an_element_via_the_wheel?revision=1587366'
);

const elem = await page.$('div');
const boundingBox = await elem.boundingBox();
await page.mouse.move(
  boundingBox.x + boundingBox.width / 2,
  boundingBox.y + boundingBox.height / 2
);

await page.mouse.wheel({ deltaY: -100 });

class: Touchscreen

touchscreen.tap(x, y)

Dispatches a touchstart and touchend event.

class: Tracing

You can use tracing.start and tracing.stop to create a trace file which can be opened in Chrome DevTools or timeline viewer.

await page.tracing.start({ path: 'trace.json' });
await page.goto('https://www.google.com');
await page.tracing.stop();

tracing.start([options])

  • options <Object>
    • path <string> A path to write the trace file to.
    • screenshots <boolean> captures screenshots in the trace.
    • categories <Array<string>> specify custom categories to use instead of default.
  • returns: <Promise>

Only one trace can be active at a time per browser.

tracing.stop()

  • returns: <Promise<Buffer>> Promise which resolves to buffer with trace data.

class: FileChooser

FileChooser objects are returned via the 'page.waitForFileChooser' method.

File choosers let you react to the page requesting for file(s) to be loaded by the web app. (This file chooser does not cover the “Save file” dialog.)

An example of using FileChooser:

const [fileChooser] = await Promise.all([
  page.waitForFileChooser(),
  page.click('#upload-file-button'), // some button that triggers file selection
]);
await fileChooser.accept(['/tmp/myfile.pdf']);

Note

In browsers, only one file chooser can be opened at a time. All file choosers must be accepted or canceled. Not doing so will prevent subsequent file choosers from appearing.

fileChooser.accept(filePaths)

fileChooser.cancel()

Closes the file chooser without selecting any files.

fileChooser.isMultiple()

  • returns: <boolean> Whether file chooser allow for multiple file selection.

class: Dialog

Dialog objects are dispatched by page via the 'dialog' event.

An example of using Dialog class:

const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  page.on('dialog', async (dialog) => {
    console.log(dialog.message());
    await dialog.dismiss();
    await browser.close();
  });
  page.evaluate(() => alert('1'));
})();

dialog.accept([promptText])

  • promptText <string> A text to enter in prompt. Does not cause any effects if the dialog's type is not prompt.
  • returns: <Promise> Promise which resolves when the dialog has been accepted.

dialog.defaultValue()

  • returns: <string> If dialog is prompt, returns default prompt value. Otherwise, returns empty string.

dialog.dismiss()

  • returns: <Promise> Promise which resolves when the dialog has been dismissed.

dialog.message()

  • returns: <string> A message displayed in the dialog.

dialog.type()

  • returns: <string> Dialog's type, can be one of alert, beforeunload, confirm or prompt.

class: ConsoleMessage

ConsoleMessage objects are dispatched by page via the 'console' event.

consoleMessage.args()

consoleMessage.location()

  • returns: <Object>
    • url <string> URL of the resource if known or undefined otherwise.
    • lineNumber <number> 0-based line number in the resource if known or undefined otherwise.
    • columnNumber <number> 0-based column number in the resource if known or undefined otherwise.

consoleMessage.stackTrace()

  • returns: <Array<Object>>
    • url <string> URL of the resource if known or undefined otherwise.
    • lineNumber <number> 0-based line number in the resource if known or undefined otherwise.
    • columnNumber <number> 0-based column number in the resource if known or undefined otherwise.

consoleMessage.text()

consoleMessage.type()

One of the following values: 'log', 'debug', 'info', 'error', 'warning', 'dir', 'dirxml', 'table', 'trace', 'clear', 'startGroup', 'startGroupCollapsed', 'endGroup', 'assert', 'profile', 'profileEnd', 'count', 'timeEnd'.

class: Frame

At every point of time, page exposes its current frame tree via the page.mainFrame() and frame.childFrames() methods.

Frame object's lifecycle is controlled by three events, dispatched on the page object:

  • 'frameattached' - fired when the frame gets attached to the page. A Frame can be attached to the page only once.
  • 'framenavigated' - fired when the frame commits navigation to a different URL.
  • 'framedetached' - fired when the frame gets detached from the page. A Frame can be detached from the page only once.

An example of dumping frame tree:

const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  await page.goto('https://www.google.com/chrome/browser/canary.html');
  dumpFrameTree(page.mainFrame(), '');
  await browser.close();

  function dumpFrameTree(frame, indent) {
    console.log(indent + frame.url());
    for (const child of frame.childFrames()) {
      dumpFrameTree(child, indent + '  ');
    }
  }
})();

An example of getting text from an iframe element:

const frame = page.frames().find((frame) => frame.name() === 'myframe');
const text = await frame.$eval('.selector', (element) => element.textContent);
console.log(text);

frame.$(selector)

The method queries frame for the selector. If there's no such element within the frame, the method resolves to null.

frame.$$(selector)

The method runs document.querySelectorAll within the frame. If no elements match the selector, the return value resolves to [].

frame.$$eval(selector, pageFunction[, ...args])

This method runs Array.from(document.querySelectorAll(selector)) within the frame and passes it as the first argument to pageFunction.

If pageFunction returns a Promise, then frame.$$eval would wait for the promise to resolve and return its value.

Examples:

const divsCounts = await frame.$$eval('div', (divs) => divs.length);

frame.$eval(selector, pageFunction[, ...args])

This method runs document.querySelector within the frame and passes it as the first argument to pageFunction. If there's no element matching selector, the method throws an error.

If pageFunction returns a Promise, then frame.$eval would wait for the promise to resolve and return its value.

Examples:

const searchValue = await frame.$eval('#search', (el) => el.value);
const preloadHref = await frame.$eval('link[rel=preload]', (el) => el.href);
const html = await frame.$eval('.main-container', (e) => e.outerHTML);

frame.$x(expression)

The method evaluates the XPath expression relative to the frame document as its context node. If there are no such elements, the method resolves to an empty array.

frame.addScriptTag(options)

  • options <Object>
    • url <string> URL of a script to be added.
    • path <string> Path to the JavaScript file to be injected into frame. If path is a relative path, then it is resolved relative to current working directory.
    • content <string> Raw JavaScript content to be injected into frame.
    • type <string> Script type. Use 'module' in order to load a Javascript ES6 module. See script for more details.
    • id <string> id attribute to add to the script tag.
  • returns: <Promise<ElementHandle>> which resolves to the added tag when the script's onload fires or when the script content was injected into frame.

Adds a <script> tag into the page with the desired URL or content.

frame.addStyleTag(options)

  • options <Object>
    • url <string> URL of the <link> tag.
    • path <string> Path to the CSS file to be injected into frame. If path is a relative path, then it is resolved relative to current working directory.
    • content <string> Raw CSS content to be injected into frame.
  • returns: <Promise<ElementHandle>> which resolves to the added tag when the stylesheet's onload fires or when the CSS content was injected into frame.

Adds a <link rel="stylesheet"> tag into the page with the desired URL or a <style type="text/css"> tag with the content.

frame.childFrames()

frame.click(selector[, options])

  • selector <string> A selector to search for element to click. If there are multiple elements satisfying the selector, the first will be clicked.
  • options <Object>
    • button <"left"|"right"|"middle"> Defaults to left.
    • clickCount <number> defaults to 1. See UIEvent.detail.
    • delay <number> Time to wait between mousedown and mouseup in milliseconds. Defaults to 0.
  • returns: <Promise> Promise which resolves when the element matching selector is successfully clicked. The Promise will be rejected if there is no element matching selector.

This method fetches an element with selector, scrolls it into view if needed, and then uses page.mouse to click in the center of the element. If there's no element matching selector, the method throws an error.

Bear in mind that if click() triggers a navigation event and there's a separate page.waitForNavigation() promise to be resolved, you may end up with a race condition that yields unexpected results. The correct pattern for click and wait for navigation is the following:

const [response] = await Promise.all([
  page.waitForNavigation(waitOptions),
  frame.click(selector, clickOptions),
]);

frame.content()

Gets the full HTML contents of the frame, including the doctype.

frame.evaluate(pageFunction[, ...args])

If the function passed to the frame.evaluate returns a Promise, then frame.evaluate would wait for the promise to resolve and return its value.

If the function passed to the frame.evaluate returns a non-Serializable value, then frame.evaluate resolves to undefined. DevTools Protocol also supports transferring some additional values that are not serializable by JSON: -0, NaN, Infinity, -Infinity, and bigint literals.

const result = await frame.evaluate(() => {
  return Promise.resolve(8 * 7);
});
console.log(result); // prints "56"

A string can also be passed in instead of a function.

console.log(await frame.evaluate('1 + 2')); // prints "3"

ElementHandle instances can be passed as arguments to the frame.evaluate:

const bodyHandle = await frame.$('body');
const html = await frame.evaluate((body) => body.innerHTML, bodyHandle);
await bodyHandle.dispose();

frame.evaluateHandle(pageFunction[, ...args])

The only difference between frame.evaluate and frame.evaluateHandle is that frame.evaluateHandle returns in-page object (JSHandle).

If the function, passed to the frame.evaluateHandle, returns a Promise, then frame.evaluateHandle would wait for the promise to resolve and return its value.

If the function returns an element, the returned handle is an ElementHandle.

const aWindowHandle = await frame.evaluateHandle(() => Promise.resolve(window));
aWindowHandle; // Handle for the window object.

A string can also be passed in instead of a function.

const aHandle = await frame.evaluateHandle('document'); // Handle for the 'document'.

JSHandle instances can be passed as arguments to the frame.evaluateHandle:

const aHandle = await frame.evaluateHandle(() => document.body);
const resultHandle = await frame.evaluateHandle(
  (body) => body.innerHTML,
  aHandle
);
console.log(await resultHandle.jsonValue());
await resultHandle.dispose();

frame.executionContext()

Returns promise that resolves to the frame's default execution context.

frame.focus(selector)

  • selector <string> A selector of an element to focus. If there are multiple elements satisfying the selector, the first will be focused.
  • returns: <Promise> Promise which resolves when the element matching selector is successfully focused. The promise will be rejected if there is no element matching selector.

This method fetches an element with selector and focuses it. If there's no element matching selector, the method throws an error.

frame.goto(url[, options])

  • url <string> URL to navigate frame to. The URL should include scheme, e.g. https://.
  • options <Object> Navigation parameters which might have the following properties:
    • timeout <number> Maximum navigation time in milliseconds, defaults to 30 seconds, pass 0 to disable timeout. The default value can be changed by using the page.setDefaultNavigationTimeout(timeout) or page.setDefaultTimeout(timeout) methods.
    • waitUntil <"load"|"domcontentloaded"|"networkidle0"|"networkidle2"|Array> When to consider navigation succeeded, defaults to load. Given an array of event strings, navigation is considered to be successful after all events have been fired. Events can be either:
      • load - consider navigation to be finished when the load event is fired.
      • domcontentloaded - consider navigation to be finished when the DOMContentLoaded event is fired.
      • networkidle0 - consider navigation to be finished when there are no more than 0 network connections for at least 500 ms.
      • networkidle2 - consider navigation to be finished when there are no more than 2 network connections for at least 500 ms.
    • referer <string> Referer header value. If provided it will take preference over the referer header value set by page.setExtraHTTPHeaders().
  • returns: <Promise<?HTTPResponse>> Promise which resolves to the main resource response. In case of multiple redirects, the navigation will resolve with the response of the last redirect.

frame.goto will throw an error if:

  • there's an SSL error (e.g. in case of self-signed certificates).
  • target URL is invalid.
  • the timeout is exceeded during navigation.
  • the remote server does not respond or is unreachable.
  • the main resource failed to load.

frame.goto will not throw an error when any valid HTTP status code is returned by the remote server, including 404 "Not Found" and 500 "Internal Server Error". The status code for such responses can be retrieved by calling response.status().

Note

frame.goto either throws an error or returns a main resource response. The only exceptions are navigation to about:blank or navigation to the same URL with a different hash, which would succeed and return null.

Note

Headless mode doesn't support navigation to a PDF document. See the upstream issue.

frame.hover(selector)

  • selector <string> A selector to search for element to hover. If there are multiple elements satisfying the selector, the first will be hovered.
  • returns: <Promise> Promise which resolves when the element matching selector is successfully hovered. Promise gets rejected if there's no element matching selector.

This method fetches an element with selector, scrolls it into view if needed, and then uses page.mouse to hover over the center of the element. If there's no element matching selector, the method throws an error.

frame.isDetached()

Returns true if the frame has been detached, or false otherwise.

frame.isOOPFrame()

Returns true if the frame is an OOP frame, or false otherwise.

frame.name()

Returns frame's name attribute as specified in the tag.

If the name is empty, returns the id attribute instead.

Note

This value is calculated once when the frame is created, and will not update if the attribute is changed later.

frame.parentFrame()

  • returns: <?Frame> Parent frame, if any. Detached frames and main frames return null.

frame.select(selector, ...values)

  • selector <string> A selector to query frame for
  • ...values <...string> Values of options to select. If the <select> has the multiple attribute, all values are considered, otherwise only the first one is taken into account.
  • returns: <Promise<Array<string>>> An array of option values that have been successfully selected.

Triggers a change and input event once all the provided options have been selected. If there's no <select> element matching selector, the method throws an error.

frame.select('select#colors', 'blue'); // single selection
frame.select('select#colors', 'red', 'green', 'blue'); // multiple selections

frame.setContent(html[, options])

  • html <string> HTML markup to assign to the page.
  • options <Object> Parameters which might have the following properties:
    • timeout <number> Maximum time in milliseconds for resources to load, defaults to 30 seconds, pass 0 to disable timeout. The default value can be changed by using the page.setDefaultNavigationTimeout(timeout) or page.setDefaultTimeout(timeout) methods.
    • waitUntil <"load"|"domcontentloaded"|"networkidle0"|"networkidle2"|Array> When to consider setting markup succeeded, defaults to load. Given an array of event strings, setting content is considered to be successful after all events have been fired. Events can be either:
      • load - consider setting content to be finished when the load event is fired.
      • domcontentloaded - consider setting content to be finished when the DOMContentLoaded event is fired.
      • networkidle0 - consider setting content to be finished when there are no more than 0 network connections for at least 500 ms.
      • networkidle2 - consider setting content to be finished when there are no more than 2 network connections for at least 500 ms.
  • returns: <Promise>

frame.tap(selector)

  • selector <string> A selector to search for element to tap. If there are multiple elements satisfying the selector, the first will be tapped.
  • returns: <Promise>

This method fetches an element with selector, scrolls it into view if needed, and then uses page.touchscreen to tap in the center of the element. If there's no element matching selector, the method throws an error.

frame.title()

frame.type(selector, text[, options])

  • selector <string> A selector of an element to type into. If there are multiple elements satisfying the selector, the first will be used.
  • text <string> A text to type into a focused element.
  • options <Object>
    • delay <number> Time to wait between key presses in milliseconds. Defaults to 0.
  • returns: <Promise>

Sends a keydown, keypress/input, and keyup event for each character in the text.

To press a special key, like Control or ArrowDown, use keyboard.press.

await frame.type('#mytextarea', 'Hello'); // Types instantly
await frame.type('#mytextarea', 'World', { delay: 100 }); // Types slower, like a user

frame.url()

Returns frame's URL.

frame.waitFor(selectorOrFunctionOrTimeout[, options[, ...args]])

This method is deprecated. You should use the more explicit API methods available:

  • frame.waitForSelector
  • frame.waitForXPath
  • frame.waitForFunction
  • frame.waitForTimeout

This method behaves differently with respect to the type of the first parameter:

  • if selectorOrFunctionOrTimeout is a string, then the first argument is treated as a selector or xpath, depending on whether or not it starts with '//', and the method is a shortcut for frame.waitForSelector or frame.waitForXPath
  • if selectorOrFunctionOrTimeout is a function, then the first argument is treated as a predicate to wait for and the method is a shortcut for frame.waitForFunction().
  • if selectorOrFunctionOrTimeout is a number, then the first argument is treated as a timeout in milliseconds and the method returns a promise which resolves after the timeout
  • otherwise, an exception is thrown
// wait for selector
await page.waitFor('.foo');
// wait for 1 second
await page.waitFor(1000);
// wait for predicate
await page.waitFor(() => !!document.querySelector('.foo'));

To pass arguments from node.js to the predicate of page.waitFor function:

const selector = '.foo';
await page.waitFor(
  (selector) => !!document.querySelector(selector),
  {},
  selector
);

frame.waitForFunction(pageFunction[, options[, ...args]])

  • pageFunction <function|string> Function to be evaluated in browser context
  • options <Object> Optional waiting parameters
    • polling <string|number> An interval at which the pageFunction is executed, defaults to raf. If polling is a number, then it is treated as an interval in milliseconds at which the function would be executed. If polling is a string, then it can be one of the following values:
      • raf - to constantly execute pageFunction in requestAnimationFrame callback. This is the tightest polling mode which is suitable to observe styling changes.
      • mutation - to execute pageFunction on every DOM mutation.
    • timeout <number> maximum time to wait for in milliseconds. Defaults to 30000 (30 seconds). Pass 0 to disable timeout. The default value can be changed by using the page.setDefaultTimeout(timeout) method.
  • ...args <...Serializable|JSHandle> Arguments to pass to pageFunction
  • returns: <Promise<JSHandle>> Promise which resolves when the pageFunction returns a truthy value. It resolves to a JSHandle of the truthy value.

The waitForFunction can be used to observe viewport size change:

const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  const watchDog = page.mainFrame().waitForFunction('window.innerWidth < 100');
  page.setViewport({ width: 50, height: 50 });
  await watchDog;
  await browser.close();
})();

To pass arguments from node.js to the predicate of page.waitForFunction function:

const selector = '.foo';
await page.waitForFunction(
  (selector) => !!document.querySelector(selector),
  {},
  selector
);

frame.waitForNavigation([options])

  • options <Object> Navigation parameters which might have the following properties:
    • timeout <number> Maximum navigation time in milliseconds, defaults to 30 seconds, pass 0 to disable timeout. The default value can be changed by using the page.setDefaultNavigationTimeout(timeout) or page.setDefaultTimeout(timeout) methods.
    • waitUntil <"load"|"domcontentloaded"|"networkidle0"|"networkidle2"|Array> When to consider navigation succeeded, defaults to load. Given an array of event strings, navigation is considered to be successful after all events have been fired. Events can be either:
      • load - consider navigation to be finished when the load event is fired.
      • domcontentloaded - consider navigation to be finished when the DOMContentLoaded event is fired.
      • networkidle0 - consider navigation to be finished when there are no more than 0 network connections for at least 500 ms.
      • networkidle2 - consider navigation to be finished when there are no more than 2 network connections for at least 500 ms.
  • returns: <Promise<?HTTPResponse>> Promise which resolves to the main resource response. In case of multiple redirects, the navigation will resolve with the response of the last redirect. In case of navigation to a different anchor or navigation due to History API usage, the navigation will resolve with null.

This resolves when the frame navigates to a new URL. It is useful when you run code that will indirectly cause the frame to navigate. Consider this example:

const [response] = await Promise.all([
  frame.waitForNavigation(), // The navigation promise resolves after navigation has finished
  frame.click('a.my-link'), // Clicking the link will indirectly cause a navigation
]);

NOTE Usage of the History API to change the URL is considered a navigation.

frame.waitForSelector(selector[, options])

  • selector <string> A selector of an element to wait for
  • options <Object> Optional waiting parameters
    • visible <boolean> wait for element to be present in DOM and to be visible, i.e. to not have display: none or visibility: hidden CSS properties. Defaults to false.
    • hidden <boolean> wait for element to not be found in the DOM or to be hidden, i.e. have display: none or visibility: hidden CSS properties. Defaults to false.
    • timeout <number> maximum time to wait for in milliseconds. Defaults to 30000 (30 seconds). Pass 0 to disable timeout. The default value can be changed by using the page.setDefaultTimeout(timeout) method.
  • returns: <Promise<?ElementHandle>> Promise which resolves when element specified by selector string is added to DOM. Resolves to null if waiting for hidden: true and selector is not found in DOM.

Wait for the selector to appear in page. If at the moment of calling the method the selector already exists, the method will return immediately. If the selector doesn't appear after the timeout milliseconds of waiting, the function will throw.

This method works across navigations:

const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  let currentURL;
  page
    .mainFrame()
    .waitForSelector('img')
    .then(() => console.log('First URL with image: ' + currentURL));
  for (currentURL of [
    'https://example.com',
    'https://google.com',
    'https://bbc.com',
  ]) {
    await page.goto(currentURL);
  }
  await browser.close();
})();

frame.waitForTimeout(milliseconds)

  • milliseconds <number> The number of milliseconds to wait for.
  • returns: <Promise> Promise which resolves after the timeout has been completed.

Pauses script execution for the given number of seconds before continuing:

const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  page
    .mainFrame()
    .waitForTimeout(1000)
    .then(() => console.log('Waited a second!'));

  await browser.close();
})();

frame.waitForXPath(xpath[, options])

  • xpath <string> A xpath of an element to wait for
  • options <Object> Optional waiting parameters
    • visible <boolean> wait for element to be present in DOM and to be visible, i.e. to not have display: none or visibility: hidden CSS properties. Defaults to false.
    • hidden <boolean> wait for element to not be found in the DOM or to be hidden, i.e. have display: none or visibility: hidden CSS properties. Defaults to false.
    • timeout <number> maximum time to wait for in milliseconds. Defaults to 30000 (30 seconds). Pass 0 to disable timeout. The default value can be changed by using the page.setDefaultTimeout(timeout) method.
  • returns: <Promise<?ElementHandle>> Promise which resolves when element specified by xpath string is added to DOM. Resolves to null if waiting for hidden: true and xpath is not found in DOM.

Wait for the xpath to appear in page. If at the moment of calling the method the xpath already exists, the method will return immediately. If the xpath doesn't appear after the timeout milliseconds of waiting, the function will throw.

This method works across navigations:

const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  let currentURL;
  page
    .mainFrame()
    .waitForXPath('//img')
    .then(() => console.log('First URL with image: ' + currentURL));
  for (currentURL of [
    'https://example.com',
    'https://google.com',
    'https://bbc.com',
  ]) {
    await page.goto(currentURL);
  }
  await browser.close();
})();

class: ExecutionContext

The class represents a context for JavaScript execution. A Page might have many execution contexts:

  • each frame has "default" execution context that is always created after frame is attached to DOM. This context is returned by the frame.executionContext() method.
  • Extensions's content scripts create additional execution contexts.

Besides pages, execution contexts can be found in workers.

executionContext.evaluate(pageFunction[, ...args])

If the function passed to the executionContext.evaluate returns a Promise, then executionContext.evaluate would wait for the promise to resolve and return its value.

If the function passed to the executionContext.evaluate returns a non-Serializable value, then executionContext.evaluate resolves to undefined. DevTools Protocol also supports transferring some additional values that are not serializable by JSON: -0, NaN, Infinity, -Infinity, and bigint literals.

const executionContext = await page.mainFrame().executionContext();
const result = await executionContext.evaluate(() => Promise.resolve(8 * 7));
console.log(result); // prints "56"

A string can also be passed in instead of a function.

console.log(await executionContext.evaluate('1 + 2')); // prints "3"

JSHandle instances can be passed as arguments to the executionContext.evaluate:

const oneHandle = await executionContext.evaluateHandle(() => 1);
const twoHandle = await executionContext.evaluateHandle(() => 2);
const result = await executionContext.evaluate(
  (a, b) => a + b,
  oneHandle,
  twoHandle
);
await oneHandle.dispose();
await twoHandle.dispose();
console.log(result); // prints '3'.

executionContext.evaluateHandle(pageFunction[, ...args])

The only difference between executionContext.evaluate and executionContext.evaluateHandle is that executionContext.evaluateHandle returns in-page object (JSHandle).

If the function returns an element, the returned handle is an ElementHandle.

If the function passed to the executionContext.evaluateHandle returns a Promise, then executionContext.evaluateHandle would wait for the promise to resolve and return its value.

const context = await page.mainFrame().executionContext();
const aHandle = await context.evaluateHandle(() => Promise.resolve(self));
aHandle; // Handle for the global object.

A string can also be passed in instead of a function.

const aHandle = await context.evaluateHandle('1 + 2'); // Handle for the '3' object.

JSHandle instances can be passed as arguments to the executionContext.evaluateHandle:

const aHandle = await context.evaluateHandle(() => document.body);
const resultHandle = await context.evaluateHandle(
  (body) => body.innerHTML,
  aHandle
);
console.log(await resultHandle.jsonValue()); // prints body's innerHTML
await aHandle.dispose();
await resultHandle.dispose();

executionContext.frame()

  • returns: <?Frame> Frame associated with this execution context.

Note

Not every execution context is associated with a frame. For example, workers and extensions have execution contexts that are not associated with frames.

executionContext.queryObjects(prototypeHandle)

  • prototypeHandle <JSHandle> A handle to the object prototype.
  • returns: <Promise<JSHandle>> A handle to an array of objects with this prototype

The method iterates the JavaScript heap and finds all the objects with the given prototype.

// Create a Map object
await page.evaluate(() => (window.map = new Map()));
// Get a handle to the Map object prototype
const mapPrototype = await page.evaluateHandle(() => Map.prototype);
// Query all map instances into an array
const mapInstances = await page.queryObjects(mapPrototype);
// Count amount of map objects in heap
const count = await page.evaluate((maps) => maps.length, mapInstances);
await mapInstances.dispose();
await mapPrototype.dispose();

class: JSHandle

JSHandle represents an in-page JavaScript object. JSHandles can be created with the page.evaluateHandle method.

const windowHandle = await page.evaluateHandle(() => window);
// ...

JSHandle prevents the referenced JavaScript object from garbage collection unless the handle is disposed. JSHandles are auto-disposed when their origin frame gets navigated or the parent context gets destroyed.

JSHandle instances can be used as arguments in page.$eval(), page.evaluate() and page.evaluateHandle methods.

jsHandle.asElement()

Returns either null or the object handle itself if the object handle is an instance of ElementHandle.

jsHandle.dispose()

  • returns: <Promise> Promise which resolves when the object handle is successfully disposed.

The jsHandle.dispose method stops referencing the element handle.

jsHandle.evaluate(pageFunction[, ...args])

This method passes this handle as the first argument to pageFunction.

If pageFunction returns a Promise, then handle.evaluate would wait for the promise to resolve and return its value.

Examples:

const tweetHandle = await page.$('.tweet .retweets');
expect(await tweetHandle.evaluate((node) => node.innerText)).toBe('10');

jsHandle.evaluateHandle(pageFunction[, ...args])

This method passes this handle as the first argument to pageFunction.

The only difference between jsHandle.evaluate and jsHandle.evaluateHandle is that executionContext.evaluateHandle returns in-page object (JSHandle).

If the function returns an element, the returned handle is an ElementHandle.

If the function passed to the jsHandle.evaluateHandle returns a Promise, then jsHandle.evaluateHandle would wait for the promise to resolve and return its value.

See Page.evaluateHandle for more details.

jsHandle.executionContext()

Returns execution context the handle belongs to.

jsHandle.getProperties()

The method returns a map with property names as keys and JSHandle instances for the property values.

const handle = await page.evaluateHandle(() => ({ window, document }));
const properties = await handle.getProperties();
const windowHandle = properties.get('window');
const documentHandle = properties.get('document');
await handle.dispose();

jsHandle.getProperty(propertyName)

Fetches a single property from the referenced object.

jsHandle.jsonValue()

Returns a JSON representation of the object. If the object has a toJSON function, it will not be called.

Note

The method will return an empty JSON object if the referenced object is not stringifiable. It will throw an error if the object has circular references.

class: ElementHandle

ElementHandle represents an in-page DOM element. ElementHandles can be created with the page.$ method.

const puppeteer = require('puppeteer');

(async () => {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  await page.goto('https://example.com');
  const hrefElement = await page.$('a');
  await hrefElement.click();
  // ...
})();

ElementHandle prevents DOM element from garbage collection unless the handle is disposed. ElementHandles are auto-disposed when their origin frame gets navigated.

ElementHandle instances can be used as arguments in page.$eval() and page.evaluate() methods.

elementHandle.$(selector)

The method runs element.querySelector within the page. If no element matches the selector, the return value resolves to null.

elementHandle.$$(selector)

The method runs element.querySelectorAll within the page. If no elements match the selector, the return value resolves to [].

elementHandle.$$eval(selector, pageFunction[, ...args])

This method runs document.querySelectorAll within the element and passes it as the first argument to pageFunction. If there's no element matching selector, the method throws an error.

If pageFunction returns a Promise, then frame.$$eval would wait for the promise to resolve and return its value.

Examples:

<div class="feed">
  <div class="tweet">Hello!</div>
  <div class="tweet">Hi!</div>
</div>

const feedHandle = await page.$('.feed');
expect(
  await feedHandle.$$eval('.tweet', (nodes) => nodes.map((n) => n.innerText))
).toEqual(['Hello!', 'Hi!']);

elementHandle.$eval(selector, pageFunction[, ...args])

This method runs document.querySelector within the element and passes it as the first argument to pageFunction. If there's no element matching selector, the method throws an error.

If pageFunction returns a Promise, then frame.$eval would wait for the promise to resolve and return its value.

Examples:

const tweetHandle = await page.$('.tweet');
expect(await tweetHandle.$eval('.like', (node) => node.innerText)).toBe('100');
expect(await tweetHandle.$eval('.retweets', (node) => node.innerText)).toBe(
  '10'
);

elementHandle.$x(expression)

The method evaluates the XPath expression relative to the elementHandle as its context node. If there are no such elements, the method resolves to an empty array.

The expression should have the context node to be evaluated properly:

const [childHandle] = await parentHandle.$x('./div');

elementHandle.asElement()

elementHandle.boundingBox()

  • returns: <Promise<?Object>>
    • x <number> the x coordinate of the element in pixels.
    • y <number> the y coordinate of the element in pixels.
    • width <number> the width of the element in pixels.
    • height <number> the height of the element in pixels.

This method returns the bounding box of the element (relative to the main frame), or null if the element is not visible.

elementHandle.boxModel()

This method returns boxes of the element, or null if the element is not visible. Boxes are represented as an array of points; each Point is an object {x, y}. Box points are sorted clock-wise.

elementHandle.click([options])

  • options <Object>
    • button <"left"|"right"|"middle"> Defaults to left.
    • clickCount <number> defaults to 1. See UIEvent.detail.
    • delay <number> Time to wait between mousedown and mouseup in milliseconds. Defaults to 0.
    • offset <Object> Offset in pixels relative to the top-left corner of the border box of the element.
      • x x-offset in pixels relative to the top-left corner of the border box of the element.
      • y y-offset in pixels relative to the top-left corner of the border box of the element.
  • returns: <Promise> Promise which resolves when the element is successfully clicked. Promise gets rejected if the element is detached from DOM.

This method scrolls element into view if needed, and then uses page.mouse to click in the center of the element. If the element is detached from DOM, the method throws an error.

elementHandle.clickablePoint([offset])

  • offset <Object>
    • x x-offset in pixels relative to the top-left corner of the border box of the element.
    • y y-offset in pixels relative to the top-left corner of the border box of the element.
  • returns: <[Promise<[Point]>]> Resolves to the x, y point that describes the element's position.

elementHandle.contentFrame()

  • returns: <Promise<?Frame>> Resolves to the content frame for element handles referencing iframe nodes, or null otherwise

elementHandle.dispose()

  • returns: <Promise> Promise which resolves when the element handle is successfully disposed.

The elementHandle.dispose method stops referencing the element handle.

elementHandle.drag(target)

  • target <Object>
  • returns: <[Promise<[DragData]>]>

This method creates and captures a drag event from the element.

elementHandle.dragAndDrop(target[, options])

This method will drag a given element and drop it onto a target element.

elementHandle.dragEnter([data])

  • data <Object> drag data created from element.drag
  • returns: <Promise>

This method will trigger a dragenter event from the given element.

elementHandle.dragOver([data])

  • data <Object> drag data created from element.drag
  • returns: <Promise>

This method will trigger a dragover event from the given element.

elementHandle.drop([data])

  • data <Object> drag data created from element.drag
  • returns: <Promise>

This method will trigger a drop event from the given element.

elementHandle.evaluate(pageFunction[, ...args])

This method passes this handle as the first argument to pageFunction.

If pageFunction returns a Promise, then handle.evaluate would wait for the promise to resolve and return its value.

Examples:

const tweetHandle = await page.$('.tweet .retweets');
expect(await tweetHandle.evaluate((node) => node.innerText)).toBe('10');

elementHandle.evaluateHandle(pageFunction[, ...args])

This method passes this handle as the first argument to pageFunction.

The only difference between elementHandle.evaluate and elementHandle.evaluateHandle is that executionContext.evaluateHandle returns in-page object (JSHandle).

If the function returns an element, the returned handle is an ElementHandle.

If the function passed to the elementHandle.evaluateHandle returns a Promise, then elementHandle.evaluateHandle would wait for the promise to resolve and return its value.

See Page.evaluateHandle for more details.

elementHandle.executionContext()

elementHandle.focus()

Calls focus on the element.

elementHandle.getProperties()

The method returns a map with property names as keys and JSHandle instances for the property values.

const listHandle = await page.evaluateHandle(() => document.body.children);
const properties = await listHandle.getProperties();
const children = [];
for (const property of properties.values()) {
  const element = property.asElement();
  if (element) children.push(element);
}
children; // holds elementHandles to all children of document.body

elementHandle.getProperty(propertyName)

Fetches a single property from the objectHandle.

elementHandle.hover()

  • returns: <Promise> Promise which resolves when the element is successfully hovered.

This method scrolls element into view if needed, and then uses page.mouse to hover over the center of the element. If the element is detached from DOM, the method throws an error.

elementHandle.isIntersectingViewport([options])

  • options <Object>
    • threshold <number> threshold for the intersection between 0 (no intersection) and 1 (full intersection). Defaults to 1.
  • returns: <Promise<boolean>> Resolves to true if the element is visible in the current viewport.

elementHandle.jsonValue()

Returns a JSON representation of the object. The JSON is generated by running JSON.stringify on the object in page and consequent JSON.parse in puppeteer.

Note

The method will throw if the referenced object is not stringifiable.

elementHandle.press(key[, options])

  • key <string> Name of key to press, such as ArrowLeft. See USKeyboardLayout for a list of all key names.
  • options <Object>
    • text <string> If specified, generates an input event with this text.
    • delay <number> Time to wait between keydown and keyup in milliseconds. Defaults to 0.
  • returns: <Promise>

Focuses the element, and then uses keyboard.down and keyboard.up.

If key is a single character and no modifier keys besides Shift are being held down, a keypress/input event will also be generated. The text option can be specified to force an input event to be generated.

Note

Modifier keys DO affect elementHandle.press. Holding down Shift will type the text in upper case.

elementHandle.screenshot([options])

  • options <Object> Same options as in page.screenshot.
  • returns: <Promise<string|Buffer>> Promise which resolves to buffer or a base64 string (depending on the value of options.encoding) with captured screenshot.

This method scrolls element into view if needed, and then uses page.screenshot to take a screenshot of the element. If the element is detached from DOM, the method throws an error.

elementHandle.select(...values)

  • ...values <...string> Values of options to select. If the <select> has the multiple attribute, all values are considered, otherwise only the first one is taken into account.
  • returns: <Promise<Array<string>>> An array of option values that have been successfully selected.

Triggers a change and input event once all the provided options have been selected. If there's no <select> element matching selector, the method throws an error.

handle.select('blue'); // single selection
handle.select('red', 'green', 'blue'); // multiple selections

elementHandle.tap()

  • returns: <Promise> Promise which resolves when the element is successfully tapped. Promise gets rejected if the element is detached from DOM.

This method scrolls element into view if needed, and then uses touchscreen.tap to tap in the center of the element. If the element is detached from DOM, the method throws an error.

elementHandle.toString()

elementHandle.type(text[, options])

  • text <string> A text to type into a focused element.
  • options <Object>
    • delay <number> Time to wait between key presses in milliseconds. Defaults to 0.
  • returns: <Promise>

Focuses the element, and then sends a keydown, keypress/input, and keyup event for each character in the text.

To press a special key, like Control or ArrowDown, use elementHandle.press.

await elementHandle.type('Hello'); // Types instantly
await elementHandle.type('World', { delay: 100 }); // Types slower, like a user

An example of typing into a text field and then submitting the form:

const elementHandle = await page.$('input');
await elementHandle.type('some text');
await elementHandle.press('Enter');

elementHandle.uploadFile(...filePaths)

  • ...filePaths <...string> Sets the value of the file input to these paths. If some of the filePaths are relative paths, then they are resolved relative to the current working directory.
  • returns: <Promise>

This method expects elementHandle to point to an input element.

elementHandle.waitForSelector(selector[, options])

  • selector <string> A selector of an element to wait for
  • options <Object> Optional waiting parameters
    • visible <boolean> wait for element to be present in DOM and to be visible, i.e. to not have display: none or visibility: hidden CSS properties. Defaults to false.
    • hidden <boolean> wait for element to not be found in the DOM or to be hidden, i.e. have display: none or visibility: hidden CSS properties. Defaults to false.
    • timeout <number> maximum time to wait for in milliseconds. Defaults to 30000 (30 seconds). Pass 0 to disable timeout. The default value can be changed by using the page.setDefaultTimeout(timeout) method.
  • returns: <Promise<?ElementHandle>> Promise which resolves when element specified by selector string is added to DOM. Resolves to null if waiting for hidden: true and selector is not found in DOM.

Wait for an element matching selector to appear within the elementHandle’s subtree. If the selector already matches an element at the moment of calling the method, the promise returned by the method resolves immediately. If the selector doesn’t appear after timeout milliseconds of waiting, the promise rejects.

This method does not work across navigations or if the element is detached from DOM.

class: HTTPRequest

Whenever the page sends a request, such as for a network resource, the following events are emitted by Puppeteer's page:

  • 'request' emitted when the request is issued by the page.
  • 'response' emitted when/if the response is received for the request.
  • 'requestfinished' emitted when the response body is downloaded and the request is complete.

If request fails at some point, then instead of 'requestfinished' event (and possibly instead of 'response' event), the 'requestfailed' event is emitted.

Note

HTTP Error responses, such as 404 or 503, are still successful responses from HTTP standpoint, so request will complete with 'requestfinished' event.

If request gets a 'redirect' response, the request is successfully finished with the 'requestfinished' event, and a new request is issued to a redirected URL.

httpRequest.abort([errorCode], [priority])

  • errorCode <string> Optional error code. Defaults to failed, could be one of the following:
    • aborted - An operation was aborted (due to user action).
    • accessdenied - Permission to access a resource, other than the network, was denied.
    • addressunreachable - The IP address is unreachable. This usually means that there is no route to the specified host or network.
    • blockedbyclient - The client chose to block the request.
    • blockedbyresponse - The request failed because the response was delivered along with requirements which are not met ('X-Frame-Options' and 'Content-Security-Policy' ancestor checks, for instance).
    • connectionaborted - A connection timed out as a result of not receiving an ACK for data sent.
    • connectionclosed - A connection was closed (corresponding to a TCP FIN).
    • connectionfailed - A connection attempt failed.
    • connectionrefused - A connection attempt was refused.
    • connectionreset - A connection was reset (corresponding to a TCP RST).
    • internetdisconnected - The Internet connection has been lost.
    • namenotresolved - The host name could not be resolved.
    • timedout - An operation timed out.
    • failed - A generic failure occurred.
  • priority <number> - Optional intercept abort priority. If provided, intercept will be resolved using cooperative handling rules. Otherwise, intercept will be resolved immediately.
  • returns: <Promise>

Aborts request. To use this, request interception should be enabled with page.setRequestInterception. Exception is immediately thrown if the request interception is not enabled.

httpRequest.abortErrorReason()

Returns the most recent reason for aborting set by the previous call to abort() in Cooperative Intercept Mode.

httpRequest.continue([overrides], [priority])

  • overrides <Object> Optional request overwrites, which can be one of the following:
    • url <string> If set changes the request URL. This is not a redirect. The request will be silently forwarded to the new URL. For example, the address bar will show the original URL.
    • method <string> If set changes the request method (e.g. GET or POST).
    • postData <string> If set changes the post data of request.
    • headers <Object> If set changes the request HTTP headers. Header values will be converted to a string.
  • priority <number> - Optional intercept abort priority. If provided, intercept will be resolved using coopeative handling rules. Otherwise, intercept will be resolved immediately.
  • returns: <Promise>

Continues request with optional request overrides. To use this, request interception should be enabled with page.setRequestInterception. Exception is immediately thrown if the request interception is not enabled.

Note: Pass DEFAULT_INTERCEPT_RESOLUTION_PRIORITY to continue a request if your intent is to bypass/defer the request because your handler has no opinion about it.

await page.setRequestInterception(true);
page.on('request', (request) => {
  if (request.isInterceptResolutionHandled()) return;

  // Override headers
  const headers = Object.assign({}, request.headers(), {
    foo: 'bar', // set "foo" header
    origin: undefined, // remove "origin" header
  });
  request.continue({ headers });
});

httpRequest.continueRequestOverrides()

  • returns: <Object> Optional request overwrites, which can be one of the following:
    • url <string> If set changes the request URL. This is not a redirect. The request will be silently forwarded to the new URL. For example, the address bar will show the original URL.
    • method <string> If set changes the request method (e.g. GET or POST).
    • postData <string> If set changes the post data of request.
    • headers <Object> If set changes the request HTTP headers. Header values will be converted to a string.

Returns the most recent set of request overrides set with a previous call to continue() in Cooperative Intercept Mode.

httpRequest.enqueueInterceptAction(pendingHandler)

  • pendingHandler <function> The request interception handler to enqueue

Enqueues a request handler for processing. This facilitates proper execution of async handlers.

httpRequest.failure()

  • returns: <?Object> Object describing request failure, if any
    • errorText <string> Human-readable error message, e.g. 'net::ERR_FAILED'.

The method returns null unless this request was failed, as reported by requestfailed event.

Example of logging all failed requests:

page.on('requestfailed', (request) => {
  console.log(request.url() + ' ' + request.failure().errorText);
});

httpRequest.finalizeInterceptions()

  • returns: <[Promise]>

When in Cooperative Intercept Mode, awaits pending interception handlers and then decides how to fulfill the request interception.

httpRequest.frame()

  • returns: <?Frame> A Frame that initiated this request, or null if navigating to error pages.

httpRequest.headers()

  • returns: <Object> An object with HTTP headers associated with the request. All header names are lower-case.

httpRequest.initiator()

  • returns: <Object> An object describing the initiator of the request
    • type <string> Type of this initiator. Possible values: parser, script, preload, SignedExchange and other.
    • stack <?Object> JavaScript stack trace for the initiator, set for script only.
    • url <?string> Initiator URL, set for parser, script and SignedExchange type.
    • lineNumber <?number> 0 based initiator line number, set for parser and script.

httpRequest.interceptResolutionState()

  • returns: <[InterceptResolutionState]>
    • action <[InterceptResolutionAction]> Current resolution action.
    • priority <?number> The current priority of the winning action.

InterceptResolutionAction is one of:

  • abort - The request will be aborted if no higher priority arises.
  • respond - The request will be responded if no higher priority arises.
  • continue - The request will be continued if no higher priority arises.
  • disabled - Request interception is not currently enabled (see page.setRequestInterception).
  • none - abort/continue/respond have not been called yet.
  • already-handled - The interception has already been handled in Legacy Mode by a call to abort/continue/respond with a priority of undefined. Subsequent calls to abort/continue/respond will throw an exception.

This example will continue() a request at a slightly higher priority than the current action if the interception has not already handled and is not already being continued.

page.on('request', (interceptedRequest) => {
  const { action, priority } = interceptedRequest.interceptResolutionState();
  if (action === InterceptResolutionAction.AlreadyHandled) return;
  if (action === InterceptResolutionAction.Continue) return;

  // Change the action to `continue` and bump the priority so `continue` becomes the new winner
  interceptedRequest.continue(
    interceptedRequest.continueRequestOverrides(),
    priority + 1
  );
});

httpRequest.isInterceptResolutionHandled()

Whether this request's interception has been handled (i.e., abort, continue, or respond has already been called with a priority of undefined).

httpRequest.isNavigationRequest()

Whether this request is driving frame's navigation.

httpRequest.method()

  • returns: <string> Request's method (GET, POST, etc.)

httpRequest.postData()

  • returns: <string> Request's post body, if any.

httpRequest.redirectChain()

A redirectChain is a chain of requests initiated to fetch a resource.

  • If there are no redirects and the request was successful, the chain will be empty.
  • If a server responds with at least a single redirect, then the chain will contain all the requests that were redirected.

redirectChain is shared between all the requests of the same chain.

For example, if the website http://example.com has a single redirect to https://example.com, then the chain will contain one request:

const response = await page.goto('http://example.com');
const chain = response.request().redirectChain();
console.log(chain.length); // 1
console.log(chain[0].url()); // 'http://example.com'

If the website https://google.com has no redirects, then the chain will be empty:

const response = await page.goto('https://google.com');
const chain = response.request().redirectChain();
console.log(chain.length); // 0

httpRequest.resourceType()

Contains the request's resource type as it was perceived by the rendering engine. ResourceType will be one of the following: document, stylesheet, image, media, font, script, texttrack, xhr, fetch, eventsource, websocket, manifest, other.

httpRequest.respond(response, [priority])

  • response <Object> Response that will fulfill this request
    • status <number> Response status code, defaults to 200.
    • headers <Object> Optional response headers. Header values should be a string or a string array.
    • contentType <string> If set, equals to setting Content-Type response header
    • body <string|Buffer> Optional response body
  • priority <number> - Optional intercept abort priority. If provided, intercept will be resolved using coopeative handling rules. Otherwise, intercept will be resolved immediately.
  • returns: <Promise>

Fulfills request with given response. To use this, request interception should be enabled with page.setRequestInterception. Exception is thrown if request interception is not enabled.

An example of fulfilling all requests with 404 responses:

await page.setRequestInterception(true);
page.on('request', (request) => {
  if (request.isInterceptResolutionHandled()) return;

  request.respond({
    status: 404,
    contentType: 'text/plain',
    body: 'Not Found!',
  });
});

Note

Mocking responses for dataURL requests is not supported. Calling request.respond for a dataURL request is a noop.

httpRequest.response()

httpRequest.responseForRequest()

Returns the current response object set by the previous call to respond() in Cooperative Intercept Mode.

httpRequest.url()

  • returns: <string> URL of the request.

class: HTTPResponse

HTTPResponse class represents responses which are received by page.

httpResponse.buffer()

  • returns: <Promise<Buffer>> Promise which resolves to a buffer with response body.

httpResponse.frame()

  • returns: <?Frame> A Frame that initiated this response, or null if navigating to error pages.

httpResponse.fromCache()

True if the response was served from either the browser's disk cache or memory cache.

httpResponse.fromServiceWorker()

True if the response was served by a service worker.

httpResponse.headers()

  • returns: <Object> An object with HTTP headers associated with the response. All header names are lower-case.

httpResponse.json()

  • returns: <Promise<Object>> Promise which resolves to a JSON representation of response body.

This method will throw if the response body is not parsable via JSON.parse.

httpResponse.ok()

Contains a boolean stating whether the response was successful (status in the range 200-299) or not.

httpResponse.remoteAddress()

  • returns: <Object>
    • ip <string> the IP address of the remote server
    • port <number> the port used to connect to the remote server

httpResponse.request()

httpResponse.securityDetails()

  • returns: <?SecurityDetails> Security details if the response was received over the secure connection, or null otherwise.

httpResponse.status()

Contains the status code of the response (e.g., 200 for a success).

httpResponse.statusText()

Contains the status text of the response (e.g. usually an "OK" for a success).

httpResponse.text()

  • returns: <Promise<string>> Promise which resolves to a text representation of response body.

httpResponse.timing()

  • returns: <?Object>
    • requestTime <number> baseline in seconds
    • proxyStart <number> started resolving proxy (milliseconds since requestTime)
    • proxyEnd <number> finished resolving proxy (milliseconds since requestTime)
    • dnsStart <number> started DNS address resolve (milliseconds since requestTime)
    • dnsEnd <number> finished DNS address resolve (milliseconds since requestTime)
    • connectStart <number> started connecting to the remote host (milliseconds since requestTime)
    • connectEnd <number> connected to the remote host (milliseconds since requestTime)
    • sslStart <number> started SSL handshake (milliseconds since requestTime)
    • sslEnd <number> finished SSL handshake (milliseconds since requestTime)
    • workerStart <number> started running ServiceWorker (milliseconds since requestTime)
    • workerReady <number> finished Starting ServiceWorker (milliseconds since requestTime)
    • workerFetchStart <number> started fetch event (milliseconds since requestTime)
    • workerRespondWithSettled <number> settled fetch event respondWith promise (milliseconds since requestTime)
    • sendStart <number> started sending request (milliseconds since requestTime)
    • sendEnd <number> finished sending request (milliseconds since requestTime)
    • pushStart <number> time the server started pushing request (milliseconds since requestTime)
    • pushEnd <number> time the server finished pushing request (milliseconds since requestTime)
    • receiveHeadersEnd <number> finished receiving response headers (milliseconds since requestTime)

Timing information related to the response.

httpResponse.url()

Contains the URL of the response.

class: SecurityDetails

SecurityDetails class represents the security details when response was received over the secure connection.

securityDetails.issuer()

  • returns: <string> A string with the name of issuer of the certificate.

securityDetails.protocol()

  • returns: <string> String with the security protocol, eg. "TLS 1.2".

securityDetails.subjectAlternativeNames()

  • returns: <Array<string>> Returns the list of SANs (subject alternative names) of the certificate.

securityDetails.subjectName()

  • returns: <string> Name of the subject to which the certificate was issued to.

securityDetails.validFrom()

  • returns: <number> UnixTime stating the start of validity of the certificate.

securityDetails.validTo()

  • returns: <number> UnixTime stating the end of validity of the certificate.

class: Target

target.browser()

Get the browser the target belongs to.

target.browserContext()

The browser context the target belongs to.

target.createCDPSession()

Creates a Chrome Devtools Protocol session attached to the target.

target.opener()

Get the target that opened this target. Top-level targets return null.

target.page()

If the target is not of type "page" or "background_page", returns null.

target.type()

  • returns: <"page"|"background_page"|"service_worker"|"shared_worker"|"other"|"browser">

Identifies what kind of target this is. Can be "page", "background_page", "service_worker", "shared_worker", "browser" or "other".

target.url()

target.worker()

If the target is not of type "service_worker" or "shared_worker", returns null.

class: CDPSession

The CDPSession instances are used to talk raw Chrome Devtools Protocol:

  • protocol methods can be called with session.send method.
  • protocol events can be subscribed to with session.on method.

Useful links:

const client = await page.target().createCDPSession();
await client.send('Animation.enable');
client.on('Animation.animationCreated', () =>
  console.log('Animation created!')
);
const response = await client.send('Animation.getPlaybackRate');
console.log('playback rate is ' + response.playbackRate);
await client.send('Animation.setPlaybackRate', {
  playbackRate: response.playbackRate / 2,
});

cdpSession.connection()

  • returns: <[Connection]>

Returns the underlying connection associated with the session. Can be used to obtain other related sessions.

cdpSession.detach()

Detaches the cdpSession from the target. Once detached, the cdpSession object won't emit any events and can't be used to send messages.

cdpSession.id()

Returns the session's id.

cdpSession.send(method[, ...paramArgs])

class: Coverage

Coverage gathers information about parts of JavaScript and CSS that were used by the page.

An example of using JavaScript and CSS coverage to get percentage of initially executed code:

// Enable both JavaScript and CSS coverage
await Promise.all([
  page.coverage.startJSCoverage(),
  page.coverage.startCSSCoverage(),
]);
// Navigate to page
await page.goto('https://example.com');
// Disable both JavaScript and CSS coverage
const [jsCoverage, cssCoverage] = await Promise.all([
  page.coverage.stopJSCoverage(),
  page.coverage.stopCSSCoverage(),
]);
let totalBytes = 0;
let usedBytes = 0;
const coverage = [...jsCoverage, ...cssCoverage];
for (const entry of coverage) {
  totalBytes += entry.text.length;
  for (const range of entry.ranges) usedBytes += range.end - range.start - 1;
}
console.log(`Bytes used: ${(usedBytes / totalBytes) * 100}%`);

To output coverage in a form consumable by Istanbul, see puppeteer-to-istanbul.

coverage.startCSSCoverage([options])

  • options <Object> Set of configurable options for coverage
    • resetOnNavigation <boolean> Whether to reset coverage on every navigation. Defaults to true.
  • returns: <Promise> Promise that resolves when coverage is started

coverage.startJSCoverage([options])

  • options <Object> Set of configurable options for coverage
    • resetOnNavigation <boolean> Whether to reset coverage on every navigation. Defaults to true.
    • reportAnonymousScripts <boolean> Whether anonymous scripts generated by the page should be reported. Defaults to false.
  • returns: <Promise> Promise that resolves when coverage is started

Note

Anonymous scripts are ones that don't have an associated URL. These are scripts that are dynamically created on the page using eval or new Function. If reportAnonymousScripts is set to true, anonymous scripts will have __puppeteer_evaluation_script__ as their URL.

coverage.stopCSSCoverage()

  • returns: <Promise<Array<Object>>> Promise that resolves to the array of coverage reports for all stylesheets
    • url <string> StyleSheet URL
    • text <string> StyleSheet content
    • ranges <Array<Object>> StyleSheet ranges that were used. Ranges are sorted and non-overlapping.
      • start <number> A start offset in text, inclusive
      • end <number> An end offset in text, exclusive

Note

CSS Coverage doesn't include dynamically injected style tags without sourceURLs.

coverage.stopJSCoverage()

  • returns: <Promise<Array<Object>>> Promise that resolves to the array of coverage reports for all scripts
    • url <string> Script URL
    • text <string> Script content
    • ranges <Array<Object>> Script ranges that were executed. Ranges are sorted and non-overlapping.
      • start <number> A start offset in text, inclusive
      • end <number> An end offset in text, exclusive

Note

JavaScript Coverage doesn't include anonymous scripts by default. However, scripts with sourceURLs are reported.

class: TimeoutError

TimeoutError is emitted whenever certain operations are terminated due to timeout, e.g. page.waitForSelector(selector[, options]) or puppeteer.launch([options]).

class: EventEmitter

A small EventEmitter class backed by Mitt.

eventEmitter.addListener(event, handler)

  • event <string|symbol> the event to add the handler to.
  • handler <Function> the event listener that will be added.
  • returns: this so you can chain method calls

This method is identical to on and maintained for compatibility with Node's EventEmitter. We recommend using on by default.

eventEmitter.emit(event, [eventData])

  • event <string|symbol> the event to trigger.
  • eventData <Object> additional data to send with the event.
  • returns: boolean; true if there are any listeners for the event, false if there are none.

eventEmitter.listenerCount(event)

  • event <string|symbol> the event to check for listeners.
  • returns: <number> the number of listeners for the given event.

eventEmitter.off(event, handler)

  • event <string|symbol> the event to remove the handler from.
  • handler <Function> the event listener that will be removed.
  • returns: this so you can chain method calls

eventEmitter.on(event, handler)

  • event <string|symbol> the event to add the handler to.
  • handler <Function> the event listener that will be added.
  • returns: this so you can chain method calls

eventEmitter.once(event, handler)

  • event <string|symbol> the event to add the handler to.
  • handler <Function> the event listener that will be added.
  • returns: this so you can chain method calls

eventEmitter.removeAllListeners([event])

  • event <string|symbol> optional argument to remove all listeners for the given event. If it's not given this method will remove all listeners for all events.
  • returns: this so you can chain method calls

eventEmitter.removeListener(event, handler)

  • event <string|symbol> the event to remove the handler from.
  • handler <Function> the event listener that will be removed.
  • returns: this so you can chain method calls

This method is identical to off and maintained for compatibility with Node's EventEmitter. We recommend using off by default.

interface: CustomQueryHandler

After registration, the handler can be used everywhere where a selector is expected by prepending the selection string with <name>/. The name is only allowed to consist of lower and upper case Latin letters.

Contains two functions queryOne and queryAll that are executed in the page context. queryOne should take an Element and a selector string as argument and return a single Element or null if no element is found. queryAll takes the same arguments but should instead return a NodeList<Element> or Array<Element> with all the elements that match the given query selector.

NOTE: Because the function code needs to be serialized, it is not possible to access anything outside its scope or use imports. See e.g. the pierce handler which has redundancies because of this limitation.

interface: Selector

A selector is a string for querying elements in the page. The default behavior is to regard the string as a CSS selector and query using querySelector or querySelectorAll. If a selector string contains a forward slash / the selector is instead regarded as custom selector where everything before the slash is the custom handler name and everything after is the selector: <handler>/<selector>. Puppeteer ships with two such custom handlers pre-registered:

  • aria: Queries the accessibilty tree for computed accessibility properties. The selectors consist of an accessible name to query for and optionally further aria attributes on the form [<attribute>=<value>]. Currently, we only support the name and role attribute. The following examples showcase how the syntax works wrt. querying:

    • title[role="heading"] queries for elements with name 'title' and role 'heading'.
    • [role="img"] queries for elements with role 'img' and any name.
    • label queries for elements with name 'label' and any role.
    • [name=""][role="button"] queries for elements with no name and role 'button'.

  • pierce: Takes a CSS selector and queries the page using querySelector or querySeletorAll but pierces through shadow roots.