scrapy-playwright 0.0.7

Playwright integration for Scrapy

Playwright integration for Scrapy

This project provides a Scrapy Download Handler which performs requests using Playwright for Python. It can be used to handle pages that require JavaScript. This package does not interfere with regular Scrapy workflows such as request scheduling or item processing.

Motivation

After the release of version 2.0, which includes partial coroutine syntax support and experimental asyncio support, Scrapy allows to integrate asyncio-based projects such as Playwright.

Requirements

• Python >= 3.7
• Scrapy >= 2.0 (!= 2.4.0)
• Playwright >= 1.8.0a1

Installation

$ pip install scrapy-playwright

Changelog

Please see the changelog.md file.

Configuration

Replace the default http and https Download Handlers through DOWNLOAD_HANDLERS:

DOWNLOAD_HANDLERS = {
    "http": "scrapy_playwright.handler.ScrapyPlaywrightDownloadHandler",
    "https": "scrapy_playwright.handler.ScrapyPlaywrightDownloadHandler",
}

Note that the ScrapyPlaywrightDownloadHandler class inherits from the default http/https handler, and it will only use Playwright for requests that are explicitly marked (see the "Basic usage" section for details).

Also, be sure to install the asyncio-based Twisted reactor:

TWISTED_REACTOR = "twisted.internet.asyncioreactor.AsyncioSelectorReactor"

Settings

scrapy-playwright accepts the following settings:

• PLAYWRIGHT_BROWSER_TYPE (type str, default chromium) The browser type to be launched. Valid values are (chromium, firefox, webkit).

• PLAYWRIGHT_LAUNCH_OPTIONS (type dict, default {})

  A dictionary with options to be passed when launching the Browser. See the docs for BrowserType.launch.

• PLAYWRIGHT_CONTEXT_ARGS (type dict, default {})

  A dictionary with default keyword arguments to be passed when creating the "default" Browser context.

  Deprecated: use PLAYWRIGHT_CONTEXTS instead

• PLAYWRIGHT_CONTEXTS (type dict[str, dict], default {})

  A dictionary which defines Browser contexts to be created on startup. It should be a mapping of (name, keyword arguments) For instance:

  {
      "first": {
          "context_arg1": "value",
          "context_arg2": "value",
      },
      "second": {
          "context_arg1": "value",
      },
  }
  

  If no contexts are defined, a default context (called default) is created. The arguments passed here take precedence over the ones defined in PLAYWRIGHT_CONTEXT_ARGS. See the docs for Browser.new_context.

• PLAYWRIGHT_DEFAULT_NAVIGATION_TIMEOUT (type Optional[int], default None)

  The timeout used when requesting pages by Playwright. If None or unset, the default value will be used (30000 ms at the time of writing this). See the docs for BrowserContext.set_default_navigation_timeout.

Basic usage

Set the playwright Request.meta key to download a request using Playwright:

import scrapy

class AwesomeSpider(scrapy.Spider):
    name = "awesome"

    def start_requests(self):
        # GET request
        yield scrapy.Request("https://httpbin.org/get", meta={"playwright": True})
        # POST request
        yield scrapy.FormRequest(
            url="https://httpbin.org/post",
            formdata={"foo": "bar"},
            meta={"playwright": True},
        )

    def parse(self, response):
        # 'response' contains the page as seen by the browser
        yield {"url": response.url}

Notes about the User-Agent header

By default, outgoing requests include the User-Agent set by Scrapy (either with the USER_AGENT or DEFAULT_REQUEST_HEADERS settings or via the Request.headers attribute). This could cause some sites to react in unexpected ways, for instance if the user agent does not match the Browser being used. If you prefer to send the User-Agent from the Browser, set the Scrapy user agent to None.

Receiving the Page object in the callback

Specifying a non-False value for the playwright_include_page meta key for a request will result in the corresponding playwright.async_api.Page object being available in the playwright_page meta key in the request callback. In order to be able to await coroutines on the provided Page object, the callback needs to be defined as a coroutine function (async def).

import scrapy
import playwright

class AwesomeSpiderWithPage(scrapy.Spider):
    name = "page"

    def start_requests(self):
        yield scrapy.Request(
            url="https://example.org",
            meta={"playwright": True, "playwright_include_page": True},
        )

    async def parse(self, response):
        page = response.meta["playwright_page"]
        title = await page.title()  # "Example Domain"
        await page.close()
        return {"title": title}

Notes:

• In order to avoid memory issues, it is recommended to manually close the page by awaiting the Page.close coroutine.
• Any network operations resulting from awaiting a coroutine on a Page object (goto, go_back, etc) will be executed directly by Playwright, bypassing the Scrapy request workflow (Scheduler, Middlewares, etc).

Multiple browser contexts

Multiple browser contexts to be launched at startup can be defined via the PLAYWRIGHT_CONTEXTS setting.

Choosing a specific context for a request

Pass the name of the desired context in the playwright_context meta key:

yield scrapy.Request(
    url="https://example.org",
    meta={"playwright": True, "playwright_context": "first"},
)

Creating a context during a crawl

If the context specified in the playwright_context meta key does not exist, it will be created. You can specify keyword arguments to be passed to Browser.new_context in the playwright_context_kwargs meta key:

yield scrapy.Request(
    url="https://example.org",
    meta={
        "playwright": True,
        "playwright_context": "new",
        "playwright_context_kwargs": {
            "java_script_enabled": False,
            "ignore_https_errors": True,
            "proxy": {
                "server": "http://myproxy.com:3128",
                "username": "user",
                "password": "pass",
            },
        },
    },
)

Please note that if a context with the specified name already exists, that context is used and playwright_context_kwargs are ignored.

Closing a context during a crawl

After receiving the Page object in your callback, you can access a context though the corresponding Page.context attribute, and await close on it.

def parse(self, response):
    yield scrapy.Request(
        url="https://example.org",
        callback=self.parse_in_new_context,
        meta={"playwright": True, "playwright_context": "new", "playwright_include_page": True},
    )

async def parse_in_new_context(self, response):
    page = response.meta["playwright_page"]
    title = await page.title()
    await page.context.close()  # close the context
    await page.close()
    return {"title": title}

Page coroutines

A sorted iterable (list, tuple or dict, for instance) could be passed in the playwright_page_coroutines Request.meta key to request coroutines to be awaited on the Page before returning the final Response to the callback.

This is useful when you need to perform certain actions on a page, like scrolling down or clicking links, and you want everything to count as a single Scrapy Response, containing the final result.

Supported actions

• scrapy_playwright.page.PageCoroutine(method: str, *args, **kwargs):

  Represents a coroutine to be awaited on a playwright.page.Page object, such as "click", "screenshot", "evaluate", etc. method should be the name of the coroutine, *args and **kwargs are passed to the function call.

  The coroutine result will be stored in the PageCoroutine.result attribute

  For instance,

  PageCoroutine("screenshot", path="quotes.png", fullPage=True)
  

  produces the same effect as:

  # 'page' is a playwright.async_api.Page object
  await page.screenshot(path="quotes.png", fullPage=True)
  

Page events

A dictionary of Page event handlers can be specified in the playwright_page_event_handlers Request.meta key. Keys are the name of the event to be handled (dialog, download, etc). Values can be either callables or strings (in which case a spider method with the name will be looked up).

Example:

from playwright.async_api import Dialog

async def handle_dialog(dialog: Dialog) -> None:
    logging.info(f"Handled dialog with message: {dialog.message}")
    await dialog.dismiss()

class EventSpider(scrapy.Spider):
    name = "event"

    def start_requests(self):
        yield scrapy.Request(
            url="https://example.org",
            meta=dict(
                playwright=True,
                playwright_page_event_handlers={
                    "dialog": handle_dialog,
                    "response": "handle_response",
                },
            ),
        )

    async def handle_response(self, response: PlaywrightResponse) -> None:
        logging.info(f"Received response with URL {response.url}")

See the upstream Page docs for a list of the accepted events and the arguments passed to their handlers.

Note: keep in mind that, unless they are removed later, these handlers will remain attached to the page and will be called for subsequent downloads using the same page. This is usually not a problem, since by default requests are performed in single-use pages.

Examples

Click on a link, save the resulting page as PDF

class ClickAndSavePdfSpider(scrapy.Spider):
    name = "pdf"

    def start_requests(self):
        yield scrapy.Request(
            url="https://example.org",
            meta=dict(
                playwright=True,
                playwright_page_coroutines={
                    "click": PageCoroutine("click", selector="a"),
                    "pdf": PageCoroutine("pdf", path="/tmp/file.pdf"),
                },
            ),
        )

    def parse(self, response):
        pdf_bytes = response.meta["playwright_page_coroutines"]["pdf"].result
        with open("iana.pdf", "wb") as fp:
            fp.write(pdf_bytes)
        yield {"url": response.url}  # response.url is "https://www.iana.org/domains/reserved"

Scroll down on an infinite scroll page, take a screenshot of the full page

class ScrollSpider(scrapy.Spider):
    name = "scroll"

    def start_requests(self):
        yield scrapy.Request(
            url="http://quotes.toscrape.com/scroll",
            meta=dict(
                playwright=True,
                playwright_include_page=True,
                playwright_page_coroutines=[
                    PageCoroutine("wait_for_selector", "div.quote"),
                    PageCoroutine("evaluate", "window.scrollBy(0, document.body.scrollHeight)"),
                    PageCoroutine("wait_for_selector", "div.quote:nth-child(11)"),  # 10 per page
                ],
            ),
        )

    async def parse(self, response):
        page = response.meta["playwright_page"]
        await page.screenshot(path="quotes.png", fullPage=True)
        await page.close()
        return {"quote_count": len(response.css("div.quote"))}  # quotes from several pages

For more examples, please see the scripts in the examples directory.