Instrument headless chrome/chromium instances from PHP
Graham Campbell 044f57003f Release 1.7.2 | 1 year ago | |
---|---|---|
.github | 2 years ago | |
src | 1 year ago | |
tests | 1 year ago | |
vendor-bin | 2 years ago | |
.editorconfig | 3 years ago | |
.gitattributes | 2 years ago | |
.gitignore | 3 years ago | |
.styleci.yml | 2 years ago | |
CHANGELOG.md | 1 year ago | |
LICENSE | 1 year ago | |
Makefile | 3 years ago | |
README.md | 2 years ago | |
composer.json | 2 years ago | |
phpstan.neon.dist | 3 years ago | |
phpunit.xml.dist | 2 years ago |
This library lets you start playing with chrome/chromium in headless mode from PHP.
Can be used synchronously and asynchronously!
Happy browsing!
Requires PHP 7.3-8.2 and a chrome/chromium 65+ executable.
Note that the library is only tested on Linux but is compatible with macOS and Windows.
The library can be installed with Composer and is available on packagist under chrome-php/chrome:
$ composer require chrome-php/chrome
It uses a simple and understandable API to start chrome, to open pages, take screenshots, crawl websites... and almost everything that you can do with chrome as a human.
use HeadlessChromium\BrowserFactory;
$browserFactory = new BrowserFactory();
// starts headless chrome
$browser = $browserFactory->createBrowser();
try {
// creates a new page and navigate to an URL
$page = $browser->createPage();
$page->navigate('http://example.com')->waitForNavigation();
// get page title
$pageTitle = $page->evaluate('document.title')->getReturnValue();
// screenshot - Say "Cheese"! 😄
$page->screenshot()->saveToFile('/foo/bar.png');
// pdf
$page->pdf(['printBackground' => false])->saveToFile('/foo/bar.pdf');
} finally {
// bye
$browser->close();
}
When starting, the factory will look for the environment variable "CHROME_PATH"
to use as the chrome executable.
If the variable is not found, it will try to guess the correct executable path according to your OS or use "chrome"
as the default.
You are also able to explicitly set up any executable of your choice when creating a new object. For instance "chromium-browser"
:
use HeadlessChromium\BrowserFactory;
// replace default 'chrome' with 'chromium-browser'
$browserFactory = new BrowserFactory('chromium-browser');
The following example disables headless mode to ease debugging
use HeadlessChromium\BrowserFactory;
$browserFactory = new BrowserFactory();
$browser = $browserFactory->createBrowser([
'headless' => false, // disable headless mode
]);
Other debug options:
[
'connectionDelay' => 0.8, // add 0.8 second of delay between each instruction sent to chrome,
'debugLogger' => 'php://stdout', // will enable verbose mode
]
About debugLogger
: this can be any of a resource string, a resource, or an object implementing
LoggerInterface
from Psr\Log (such as monolog
or apix/log).
Options set directly in the createBrowser
method will be used only for a single browser creation. The default options will be ignored.
use HeadlessChromium\BrowserFactory;
$browserFactory = new BrowserFactory();
$browser = $browserFactory->createBrowser([
'windowSize' => [1920, 1000],
'enableImages' => false,
]);
// this browser will be created without any options
$browser2 = $browserFactory->createBrowser();
Options set using the setOptions
and addOptions
methods will persist.
$browserFactory->setOptions([
'windowSize' => [1920, 1000],
]);
// both browser will have the same 'windowSize' option
$browser1 = $browserFactory->createBrowser();
$browser2 = $browserFactory->createBrowser();
$browserFactory->addOptions(['enableImages' => false]);
// this browser will have both the 'windowSize' and 'enableImages' options
$browser3 = $browserFactory->createBrowser();
$browserFactory->addOptions(['enableImages' => true]);
// this browser will have the previous 'windowSize', but 'enableImages' will be true
$browser4 = $browserFactory->createBrowser();
Here are the options available for the browser factory:
Option name | Default | Description |
---|---|---|
connectionDelay |
0 |
Delay to apply between each operation for debugging purposes |
customFlags |
none | An array of flags to pass to the command line. Eg: ['--option1', '--option2=someValue'] |
debugLogger |
null |
A string (e.g "php://stdout"), or resource, or PSR-3 logger instance to print debug messages |
enableImages |
true |
Toggles loading of images |
envVariables |
none | An array of environment variables to pass to the process (example DISPLAY variable) |
headers |
none | An array of custom HTTP headers |
headless |
true |
Enable or disable headless mode |
ignoreCertificateErrors |
false |
Set chrome to ignore ssl errors |
keepAlive |
false |
Set to true to keep alive the chrome instance when the script terminates |
noSandbox |
false |
Enable no sandbox mode, useful to run in a docker container |
noProxyServer |
false |
Don't use a proxy server, always make direct connections. Overrides other proxy settings. |
proxyBypassList |
none | Specifies a list of hosts for whom we bypass proxy settings and use direct connections |
proxyServer |
none | Proxy server to use. usage: 127.0.0.1:8080 (authorisation with credentials does not work) |
sendSyncDefaultTimeout |
5000 |
Default timeout (ms) for sending sync messages |
startupTimeout |
30 |
Maximum time in seconds to wait for chrome to start |
userAgent |
none | User agent to use for the whole browser (see page API for alternative) |
userDataDir |
none | Chrome user data dir (default: a new empty dir is generated temporarily) |
windowSize |
none | Size of the window. usage: $width, $height - see also Page::setViewport |
This example shows how to share a single instance of chrome for multiple scripts.
The first time the script is started we use the browser factory in order to start chrome, afterwards we save the uri to connect to this browser in the file system.
The next calls to the script will read the uri from that file in order to connect to the chrome instance instead of creating a new one. If chrome was closed or crashed, a new instance is started again.
use \HeadlessChromium\BrowserFactory;
use \HeadlessChromium\Exception\BrowserConnectionFailed;
// path to the file to store websocket's uri
$socket = \file_get_contents('/tmp/chrome-php-demo-socket');
try {
$browser = BrowserFactory::connectToBrowser($socket);
} catch (BrowserConnectionFailed $e) {
// The browser was probably closed, start it again
$factory = new BrowserFactory();
$browser = $factory->createBrowser([
'keepAlive' => true,
]);
// save the uri to be able to connect again to browser
\file_put_contents($socketFile, $browser->getSocketUri(), LOCK_EX);
}
$page = $browser->createPage();
$pages = $browser->getPages();
$browser->close();
$browser->setPagePreScript('// Simulate navigator permissions;
const originalQuery = window.navigator.permissions.query;
window.navigator.permissions.query = (parameters) => (
parameters.name === 'notifications' ?
Promise.resolve({ state: Notification.permission }) :
originalQuery(parameters)
);');
// navigate
$navigation = $page->navigate('http://example.com');
// wait for the page to be loaded
$navigation->waitForNavigation();
When using $navigation->waitForNavigation()
you will wait for 30sec until the page event "loaded" is triggered.
You can change the timeout or the event to listen for:
use HeadlessChromium\Page;
// wait 10secs for the event "DOMContentLoaded" to be triggered
$navigation->waitForNavigation(Page::DOM_CONTENT_LOADED, 10000);
Available events (in the order they trigger):
Page::DOM_CONTENT_LOADED
: dom has completely loadedPage::LOAD
: (default) page and all resources are loadedPage::NETWORK_IDLE
: page has loaded, and no network activity has occurred for at least 500msWhen you want to wait for the page to navigate 2 main issues may occur. First, the page is too long to load and second, the page you were waiting to be loaded has been replaced. The good news is that you can handle those issues using a good old try-catch:
use HeadlessChromium\Exception\OperationTimedOut;
use HeadlessChromium\Exception\NavigationExpired;
try {
$navigation->waitForNavigation()
} catch (OperationTimedOut $e) {
// too long to load
} catch (NavigationExpired $e) {
// An other page was loaded
}
Once the page has completed the navigation you can evaluate arbitrary script on this page:
// navigate
$navigation = $page->navigate('http://example.com');
// wait for the page to be loaded
$navigation->waitForNavigation();
// evaluate script in the browser
$evaluation = $page->evaluate('document.documentElement.innerHTML');
// wait for the value to return and get it
$value = $evaluation->getReturnValue();
Sometimes the script you evaluate will click a link or submit a form, in this case, the page will reload and you will want to wait for the new page to reload.
You can achieve this by using $page->evaluate('some js that will reload the page')->waitForPageReload()
.
An example is available in form-submit.php
This is an alternative to evaluate
that allows calling a given function with the given arguments in the page context:
$evaluation = $page->callFunction(
"function(a, b) {\n window.foo = a + b;\n}",
[1, 2]
);
$value = $evaluation->getReturnValue();
That's useful if you want to add jQuery (or anything else) to the page:
$page->addScriptTag([
'content' => file_get_contents('path/to/jquery.js')
])->waitForResponse();
$page->evaluate('$(".my.element").html()');
You can also use an URL to feed the src attribute:
$page->addScriptTag([
'url' => 'https://code.jquery.com/jquery-3.3.1.min.js'
])->waitForResponse();
$page->evaluate('$(".my.element").html()');
You can manually inject html to a page using the setHtml
method.
$page->setHtml('<p>text</p>');
Note that this will not append to the current page html, it will completely replace it.
You can get the page HTML as a string using the getHtml
method.
$html = $page->getHtml();
$page->addPreScript('// Simulate navigator permissions;
const originalQuery = window.navigator.permissions.query;
window.navigator.permissions.query = (parameters) => (
parameters.name === 'notifications' ?
Promise.resolve({ state: Notification.permission }) :
originalQuery(parameters)
);');
If your script needs the dom to be fully populated before it runs then you can use the option "onLoad":
$page->addPreScript($script, ['onLoad' => true]);
This feature allows changing the size of the viewport (emulation) for the current page without affecting the size of
all the browser's pages (see also option "windowSize"
of BrowserFactory::createBrowser).
$width = 600;
$height = 300;
$page->setViewport($width, $height)
->await(); // wait for the operation to complete
// navigate
$navigation = $page->navigate('http://example.com');
// wait for the page to be loaded
$navigation->waitForNavigation();
// take a screenshot
$screenshot = $page->screenshot([
'format' => 'jpeg', // default to 'png' - possible values: 'png', 'jpeg',
'quality' => 80, // only if format is 'jpeg' - default 100
]);
// save the screenshot
$screenshot->saveToFile('/some/place/file.jpg');
Screenshot an area on a page
You can use the option "clip" to choose an area on a page for the screenshot
use HeadlessChromium\Clip;
// navigate
$navigation = $page->navigate('http://example.com');
// wait for the page to be loaded
$navigation->waitForNavigation();
// create a rectangle by specifying to left corner coordinates + width and height
$x = 10;
$y = 10;
$width = 100;
$height = 100;
$clip = new Clip($x, $y, $width, $height);
// take the screenshot (in memory binaries)
$screenshot = $page->screenshot([
'clip' => $clip,
]);
// save the screenshot
$screenshot->saveToFile('/some/place/file.jpg');
Full-page screenshot
You can also take a screenshot for the full-page layout (not only the viewport) using $page->getFullPageClip
with attribute captureBeyondViewport = true
// navigate
$navigation = $page->navigate('https://example.com');
// wait for the page to be loaded
$navigation->waitForNavigation();
$screenshot = $page->screenshot([
'captureBeyondViewport' => true,
'clip' => $page->getFullPageClip(),
'format' => 'jpeg', // default to 'png' - possible values: 'png', 'jpeg',
]);
// save the screenshot
$screenshot->saveToFile('/some/place/file.jpg');
// navigate
$navigation = $page->navigate('http://example.com');
// wait for the page to be loaded
$navigation->waitForNavigation();
$options = [
'landscape' => true, // default to false
'printBackground' => true, // default to false
'displayHeaderFooter' => true, // default to false
'preferCSSPageSize' => true, // default to false (reads parameters directly from @page)
'marginTop' => 0.0, // defaults to ~0.4 (must be a float, value in inches)
'marginBottom' => 1.4, // defaults to ~0.4 (must be a float, value in inches)
'marginLeft' => 5.0, // defaults to ~0.4 (must be a float, value in inches)
'marginRight' => 1.0, // defaults to ~0.4 (must be a float, value in inches)
'paperWidth' => 6.0, // defaults to 8.5 (must be a float, value in inches)
'paperHeight' => 6.0, // defaults to 8.5 (must be a float, value in inches)
'headerTemplate' => '<div>foo</div>', // see details above
'footerTemplate' => '<div>foo</div>', // see details above
'scale' => 1.2, // defaults to 1.0 (must be a float)
];
// print as pdf (in memory binaries)
$pdf = $page->pdf($options);
// save the pdf
$pdf->saveToFile('/some/place/file.pdf');
// or directly output pdf without saving
header('Content-Description: File Transfer');
header('Content-Type: application/pdf');
header('Content-Disposition: inline; filename=filename.pdf');
header('Content-Transfer-Encoding: binary');
header('Expires: 0');
header('Cache-Control: must-revalidate, post-check=0, pre-check=0');
header('Pragma: public');
echo base64_decode($pdf->getBase64());
Options headerTemplate
and footerTemplate
:
Should be valid HTML markup with the following classes used to inject printing values into them:
You can set the path to save downloaded files.
// After creating a page.
$page->setDownloadPath('/path/to/save/downloaded/files');
The mouse API is dependent on the page instance and allows you to control the mouse's moves, clicks and scroll.
$page->mouse()
->move(10, 20) // Moves mouse to position x=10; y=20
->click() // left-click on position set above
->move(100, 200, ['steps' => 5]) // move mouse to x=100; y=200 in 5 equal steps
->click(['button' => Mouse::BUTTON_RIGHT]; // right-click on position set above
// given the last click was on a link, the next step will wait
// for the page to load after the link was clicked
$page->waitForReload();
You can emulate the mouse wheel to scroll up and down in a page, frame, or element.
$page->mouse()
->scrollDown(100) // scroll down 100px
->scrollUp(50); // scroll up 50px
The find
method will search for elements using querySelector and move the cursor to a random position over it.
try {
$page->mouse()->find('#a')->click(); // find and click at an element with id "a"
$page->mouse()->find('.a', 10); // find the 10th or last element with class "a"
} catch (ElementNotFoundException $exception) {
// element not found
}
This method will attempt to scroll right and down to bring the element to the visible screen. If the element is inside an internal scrollable section, try moving the mouse to inside that section first.
The keyboard API is dependent on the page instance and allows you to type like a real user.
$page->keyboard()
->typeRawKey('Tab') // type a raw key, such as Tab
->typeText('bar'); // type the text "bar"
To impersonate a real user you may want to add a delay between each keystroke using the setKeyInterval
method:
$page->keyboard()->setKeyInterval(10); // sets a delay of 10 milliseconds between keystrokes
The methods press
, type
, and release
can be used to send key combinations such as ctrl + v
.
// ctrl + a to select all text
$page->keyboard()
->press('control') // key names are case insensitive and trimmed
->type('a') // press and release
->release('Control');
// ctrl + c to copy and ctrl + v to paste it twice
$page->keyboard()
->press('Ctrl') // alias for Control
->type('c')
->type('V') // upper and lower cases should behave the same way
->release(); // release all
You can press the same key several times in sequence, this is the equivalent to a user pressing and holding the key. The release event, however, will be sent only once per key.
Key | Aliases |
---|---|
Control | Control , Ctrl , Ctr |
Alt | Alt , AltGr , Alt Gr |
Meta | Meta , Command , Cmd |
Shift | Shift |
You can set and get cookies for a page:
use HeadlessChromium\Cookies\Cookie;
$page = $browser->createPage();
// example 1: set cookies for a given domain
$page->setCookies([
Cookie::create('name', 'value', [
'domain' => 'example.com',
'expires' => time() + 3600 // expires in 1 hour
])
])->await();
// example 2: set cookies for the current page
$page->navigate('http://example.com')->waitForNavigation();
$page->setCookies([
Cookie::create('name', 'value', ['expires'])
])->await();
use HeadlessChromium\Cookies\Cookie;
$page = $browser->createPage();
// example 1: get all cookies for the browser
$cookies = $page->getAllCookies();
// example 2: get cookies for the current page
$page->navigate('http://example.com')->waitForNavigation();
$cookies = $page->getCookies();
// filter cookies with name == 'foo'
$cookiesFoo = $cookies->filterBy('name', 'foo');
// find first cookie with name == 'bar'
$cookieBar = $cookies->findOneBy('name', 'bar');
if ($cookieBar) {
// do something
}
You can set up a user-agent per page:
$page->setUserAgent('my user-agent');
See also BrowserFactory option userAgent
to set up it for the whole browser.
The library ships with tools that hide all the communication logic but you can use the tools used internally to communicate directly with chrome debug protocol.
Example:
use HeadlessChromium\Communication\Connection;
use HeadlessChromium\Communication\Message;
// chrome devtools URI
$webSocketUri = 'ws://127.0.0.1:9222/devtools/browser/xxx';
// create a connection
$connection = new Connection($webSocketUri);
$connection->connect();
// send method "Target.activateTarget"
$responseReader = $connection->sendMessage(new Message('Target.activateTarget', ['targetId' => 'xxx']));
// wait up to 1000ms for a response
$response = $responseReader->waitForResponse(1000);
// given a target id
$targetId = 'yyy';
// create a session for this target (attachToTarget)
$session = $connection->createSession($targetId);
// send message to this target (Target.sendMessageToTarget)
$response = $session->sendMessageSync(new Message('Page.reload'));
You can ease the debugging by setting a delay before each operation is made:
$connection->setConnectionDelay(500); // wait for 500ms between each operation to ease debugging
use HeadlessChromium\Communication\Connection;
use HeadlessChromium\Browser;
// chrome devtools URI
$webSocketUri = 'ws://127.0.0.1:9222/devtools/browser/xxx';
// create connection given a WebSocket URI
$connection = new Connection($webSocketUri);
$connection->connect();
// create browser
$browser = new Browser($connection);
Find one element on a page by CSS selector:
$page = $browser->createPage();
$page->navigate('http://example.com')->waitForNavigation();
$elem = $page->dom()->querySelector('#index_email');
Find all elements inside another element by CSS selector:
$elem = $page->dom()->querySelector('#index_email');
$elem->querySelectorAll('a.link');
Find all elements on a page by XPath selector:
$page = $browser->createPage();
$page->navigate('http://example.com')->waitForNavigation();
$elem = $page->dom()->search('//div/*/a');
You can send out a text to an element or click on it:
$elem->click();
$elem->sendKeys('Sample text');
You can upload file to file from the input:
$elem->sendFile('/path/to/file');
You can get element text or attribute:
$text = $elem->getText();
$attr = $elem->getAttribute('class');
See CONTRIBUTING.md for contribution details.
This project is licensed under the The MIT License (MIT).