Pythonで無限スクロールWebページをスクレイピングする方法

こんにちは、Crawlee 開発者の皆さん、Crawlee ブログの別のチュートリアルへようこそ。このチュートリアルでは、Crawlee for Python を使用して無限スクロール Web サイトをスクレイピングする方法を説明します。

コンテキストとしては、無限スクロールページは、古典的なページネーションに代わる最新の方法です。ユーザーが次のページを選択せずに Web ページの一番下までスクロールすると、ページには自動的に追加のデータが読み込まれ、ユーザーはさらにスクロールできるようになります。

大のスニーカーマニアとして、ナイキシューズの無限スクロール Web サイトを例として取り上げ、そこから何千ものスニーカーをかき集めます。

Crawlee for Python には、HTTP およびヘッドレスブラウザクロール用の統合インターフェイス、自動再試行など、いくつかの素晴らしい初期機能があります。

前提条件とプロジェクトのブートストラップ

次のコマンドを使用して Crawlee for Python をインストールして、チュートリアルを開始しましょう:

pipx run crawlee create nike-crawler

ログイン後にコピー

このブログを読む前に、GitHub で Crawlee for Python にスターを付けていただければ大変嬉しく思います!

アピファイ / クローリーパイソン

Crawlee - 信頼性の高いクローラを構築するための Python 用の Web スクレイピングおよびブラウザ自動化ライブラリ。 AI、LLM、RAG、または GPT のデータを抽出します。 HTML、PDF、JPG、PNG、その他のファイルを Web サイトからダウンロードします。 BeautifulSoup、Playwright、生の HTTP で動作します。ヘッドフルモードとヘッドレスモードの両方。プロキシローテーションあり。

Webスクレイピングとブラウザ自動化ライブラリ

How to scrape infinite scrolling webpages with Python

Crawlee はクローリングとスクレイピングをエンドツーエンドでカバーし、信頼性の高いスクレーパーの構築を支援します。早いです。

? Crawlee for Python は早期採用者に公開されています!

あなたのクローラーはほとんど人間のように見え、デフォルト設定であっても最新のボット保護のレーダーの下を飛行します。 Crawlee は、技術的な詳細を気にすることなく、Web をクロールしてリンクを探し、データを収集し、機械可読形式で永続的に保存するためのツールを提供します。また、豊富な構成オプションのおかげで、デフォルト設定では不十分な場合でも、プロジェクトのニーズに合わせて Crawlee のほぼすべての側面を調整できます。

? Crawlee プロジェクト Web サイトで完全なドキュメント、ガイド、例を表示します ?

Crawlee の TypeScript 実装もあります。これを探索してプロジェクトに活用できます。詳細については、GitHub の Crawlee for JS/TS の GitHub リポジトリにアクセスしてください。

インストール

私たちは…

GitHub で表示

We will scrape using headless browsers. Select PlaywrightCrawler in the terminal when Crawlee for Python asks for it.

After installation, Crawlee for Python will create boilerplate code for you. Redirect into the project folder and then run this command for all the dependencies installation:

poetry install

ログイン後にコピー

How to scrape infinite scrolling webpages

Handling accept cookie dialog
Adding request of all shoes links
Extract data from product details
Accept Cookies context manager
Handling infinite scroll on the listing page
Exporting data to CSV format

Handling accept cookie dialog

After all the necessary installations, we'll start looking into the files and configuring them accordingly.

When you look into the folder, you'll see many files, but for now, let’s focus on main.py and routes.py.

In __main__.py, let's change the target location to the Nike website. Then, just to see how scraping will happen, we'll add headless = False to the PlaywrightCrawler parameters. Let's also increase the maximum requests per crawl option to 100 to see the power of parallel scraping in Crawlee for Python.

The final code will look like this:

import asyncio

from crawlee.playwright_crawler import PlaywrightCrawler

from .routes import router


async def main() -> None:

    crawler = PlaywrightCrawler(
        headless=False,
        request_handler=router,
        max_requests_per_crawl=100,
    )

    await crawler.run(
        [
            'https://nike.com/,
        ]
    )


if __name__ == '__main__':
    asyncio.run(main())

ログイン後にコピー

Now coming to routes.py, let’s remove:

await context.enqueue_links()

ログイン後にコピー

As we don’t want to scrape the whole website.

Now, if you run the crawler using the command:

poetry run python -m nike-crawler

ログイン後にコピー

As the cookie dialog is blocking us from crawling more than one page's worth of shoes, let’s get it out of our way.

We can handle the cookie dialog by going to Chrome dev tools and looking at the test_id of the "accept cookies" button, which is dialog-accept-button.

Now, let’s remove the context.push_data call that was left there from the project template and add the code to accept the dialog in routes.py. The updated code will look like this:

from crawlee.router import Router
from crawlee.playwright_crawler import PlaywrightCrawlingContext

router = Router[PlaywrightCrawlingContext]()

@router.default_handler
async def default_handler(context: PlaywrightCrawlingContext) -> None:

    # Wait for the popup to be visible to ensure it has loaded on the page.
    await context.page.get_by_test_id('dialog-accept-button').click()

ログイン後にコピー

Adding request of all shoes links

Now, if you hover over the top bar and see all the sections, i.e., man, woman, and kids, you'll notice the “All shoes” section. As we want to scrape all the sneakers, this section interests us. Let’s use get_by_test_id with the filter of has_text=’All shoes’ and add all the links with the text “All shoes” to the request handler. Let’s add this code to the existing routes.py file:

    shoe_listing_links = (
        await context.page.get_by_test_id('link').filter(has_text='All shoes').all()
    )
    await context.add_requests(
        [
            Request.from_url(url, label='listing')
            for link in shoe_listing_links
            if (url := await link.get_attribute('href'))
        ]
    )

@router.handler('listing')
async def listing_handler(context: PlaywrightCrawlingContext) -> None:
    """Handler for shoe listings."""

ログイン後にコピー

Extract data from product details

Now that we have all the links to the pages with the title “All Shoes,” the next step is to scrape all the products on each page and the information provided on them.

We'll extract each shoe's URL, title, price, and description. Again, let's go to dev tools and extract each parameter's relevant test_id. After scraping each of the parameters, we'll use the context.push_data function to add it to the local storage. Now let's add the following code to the listing_handler and update it in the routes.py file:

@router.handler('listing')
async def listing_handler(context: PlaywrightCrawlingContext) -> None:
    """Handler for shoe listings."""        

    await context.enqueue_links(selector='a.product-card__link-overlay', label='detail')


@router.handler('detail')
async def detail_handler(context: PlaywrightCrawlingContext) -> None:
    """Handler for shoe details."""

    title = await context.page.get_by_test_id(
        'product_title',
    ).text_content()

    price = await context.page.get_by_test_id(
        'currentPrice-container',
    ).first.text_content()

    description = await context.page.get_by_test_id(
        'product-description',
    ).text_content()

    await context.push_data(
        {
            'url': context.request.loaded_url,
            'title': title,
            'price': price,
            'description': description,
        }
    )

ログイン後にコピー

Accept Cookies context manager

Since we're dealing with multiple browser pages with multiple links and we want to do infinite scrolling, we may encounter an accept cookie dialog on each page. This will prevent loading more shoes via infinite scroll.

We'll need to check for cookies on every page, as each one may be opened with a fresh session (no stored cookies) and we'll get the accept cookie dialog even though we already accepted it in another browser window. However, if we don't get the dialog, we want the request handler to work as usual.

To solve this problem, we'll try to deal with the dialog in a parallel task that will run in the background. A context manager is a nice abstraction that will allow us to reuse this logic in all the router handlers. So, let's build a context manager:

from playwright.async_api import TimeoutError as PlaywrightTimeoutError

@asynccontextmanager
async def accept_cookies(page: Page):
    task = asyncio.create_task(page.get_by_test_id('dialog-accept-button').click())
    try:
        yield
    finally:
        if not task.done():
            task.cancel()

        with suppress(asyncio.CancelledError, PlaywrightTimeoutError):
            await task

ログイン後にコピー

This context manager will make sure we're accepting the cookie dialog if it exists before scrolling and scraping the page. Let’s implement it in the routes.py file, and the updated code is here

Handling infinite scroll on the listing page

Now for the last and most interesting part of the tutorial! How to handle the infinite scroll of each shoe listing page and make sure our crawler is scrolling and scraping the data constantly.

To handle infinite scrolling in Crawlee for Python, we just need to make sure the page is loaded, which is done by waiting for the network_idle load state, and then use the infinite_scroll helper function which will keep scrolling to the bottom of the page as long as that makes additional items appear.

Let’s add two lines of code to the listing handler:

@router.handler('listing')
async def listing_handler(context: PlaywrightCrawlingContext) -> None:
    # Handler for shoe listings

    async with accept_cookies(context.page):
        await context.page.wait_for_load_state('networkidle')
        await context.infinite_scroll()
        await context.enqueue_links(
            selector='a.product-card__link-overlay', label='detail'
        )

ログイン後にコピー

Exporting data to CSV format

As we want to store all the shoe data into a CSV file, we can just add a call to the export_data helper into the __main__.py file just after the crawler run:

await crawler.export_data('shoes.csv')

ログイン後にコピー

Working crawler and its code

Now, we have a crawler ready that can scrape all the shoes from the Nike website while handling infinite scrolling and many other problems, like the cookies dialog.

You can find the complete working crawler code here on the GitHub repository.

If you have any doubts regarding this tutorial or using Crawlee for Python, feel free to join our discord community and ask fellow developers or the Crawlee team.

Crawlee & Apify

This is the official developer community of Apify and Crawlee. | 8365 members

How to scrape infinite scrolling webpages with Python

discord.com

This tutorial is taken from the webinar held on August 5th where Jan Buchar, Senior Python Engineer at Apify, gave a live demo about this use case. Watch the whole webinar here.

以上がPythonで無限スクロールWebページをスクレイピングする方法の詳細内容です。詳細については、PHP 中国語 Web サイトの他の関連記事を参照してください。