Scrapy Tips from the Pros: February 2016 Edition

Scrapy Tips from the Pros: February 2016 Edition

Welcome to the February Edition of Scrapy Tips from the Pros. Each month we’ll release a few tips and hacks that we’ve developed to help make your Scrapy workflow go more smoothly.

This month we’ll show you how to crawl websites more effectively following sitemaps and we’ll also demonstrate how to add custom settings to individual spiders in a project.

Crawling a Website with Scrapy SitemapSpider

Web crawlers feed on URLs. The more they have, the longer they live. Finding a good source of URLs for any given website is very important as it gives the crawler a strong starting point.

Sitemaps are an excellent source of seed URLs. Website developers use them to indicate which URLs are available for crawling in a machine-readable format. Sitemaps are also a good way to discover web pages that would be otherwise unreachable, since some pages may not be linked to from any other page outside of the sitemap.

Sitemaps are often available at /sitemap.xml or in a different location specified in the robots.txt file.

With Scrapy you don’t need to worry about parsing XML and making requests. It includes a SitemapSpider class you can inherit to handle all of this for you.

SitemapSpider in Action

Let’s say you want to crawl Apple’s website to price check different products. You would want to visit as many pages as possible so that you can scrape as much data as you can. Fortunately, Apple’s website provides a sitemap at apple.com/sitemap.xml, which looks like this:

<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">
    <url><loc>http://www.apple.com/</loc></url>
    <url><loc>http://www.apple.com/about/</loc></url>
    ...
    <url><loc>http://www.apple.com/about/workingwithapple.html</loc></url>
    <url><loc>http://www.apple.com/accessibility/</loc></url>
    ...
</urlset>

Scrapy’s generic SitemapSpider class implements all the logic for parsing and dispatching requests necessary to handle sitemaps. It reads and extracts URLs from the sitemap and it will dispatch a single request for each URL it finds. Here is a spider that will scrape Apple’s website using the sitemap as its seed:

from scrapy.spiders import SitemapSpider

class AppleSpider(SitemapSpider):
    name = 'apple-spider'
    sitemap_urls = ['http://www.apple.com/sitemap.xml']

    def parse(self, response):
        yield {
            'title': response.css("title ::text").extract_first(),
            'url': response.url
        }
        # ...

As you can see, you only need to subclass SitemapSpider and add the sitemap’s URL to the sitemap_urls attribute.

Now, run the spider and check the results:

$ scrapy runspider apple_spider.py -o items.jl --nolog

$ head -n 5 items.jl 
{"url": "http://www.apple.com/", "title": "Apple"}
{"url": "http://www.apple.com/ae/support/products/iphone.html", "title": "Support - AppleCare+ - iPhone - Apple (AE)"}
{"url": "http://www.apple.com/ae/support/products/ipad.html", "title": "Support - AppleCare+ - iPad - Apple (AE)"}
{"url": "http://www.apple.com/ae/support/products/", "title": "Support - AppleCare - Apple (AE)"}
{"url": "http://www.apple.com/ae/support/ipod/", "title": "iPod - Apple Support"}

Scrapy dispatches a request for each URL found by SitemapSpider in the sitemap and then it calls the parse method to handle each response it gets. However, some pages in a website may vary in structure and so you might want to use multiple callbacks for different types of pages.

For instance, you can define a specific callback to handle the Mac pages, another one for the iTunes pages and the default parse method for all the other pages:

from scrapy.spiders import SitemapSpider

class AppleSpider(SitemapSpider):
    name = 'apple-spider'
    sitemap_urls = ['http://www.apple.com/sitemap.xml']
    sitemap_rules = [
        ('/mac/', 'parse_mac'),
        ('/itunes/', 'parse_itunes'),
        ('', 'parse')
    ]

    def parse(self, response):
        self.log("default parsing method for {}".format(response.url))

    def parse_mac(self, response):
        self.log("parse_mac method for {}".format(response.url))

    def parse_itunes(self, response):
        self.log("parse_itunes method for {}".format(response.url))

To do it, you have to add a sitemap_rules attribute to your class, mapping URL patterns to callbacks. For instance, the URLs matching the ‘/mac/’ pattern will have its response handled by the parse_mac method.

So, the next time you write a crawler, make sure to use SitemapSpider if you want to have comprehensive crawls of the website.

For more features, check SitemapSpider’s documentation.

Customize Settings for Individual Spiders using custom_settings

The settings for Scrapy projects are typically stored in the project’s settings.py file. However, these are global settings that apply to each of the spiders in your project. If you want to set individual settings for each spider, all you need to do is add an attribute called custom_settings to your spider class.

This is especially useful when you need to enable or disable pipelines or middlewares for certain spiders or to specify different settings for each one. For example, some spiders in your project might require Crawlera (a smart proxy service) enabled while others don’t. You can achieve this by adding CRAWLERA_ENABLED = True in custom_settings in the specific spiders.

custom_settings in Action

Take a look at a simplified version of the spiders from a book catalog project. They need custom_settings to define where the book covers are going to be stored in the filesystem:

class AlibrisSpider(scrapy.Spider):
    name = "alibris-covers"
    allowed_domains = ["alibris.com"]
    start_urls = (
        'http://www.alibris.com/search/books/subject/Fiction-Science-Fiction',
    )
    custom_settings = {'IMAGES_STORE': 'imgs/alibris/sci-fi'}

    def parse(self, response):
        for book in response.css("div#selected-works ul.primaryList li"):
            yield {
                'image_urls': [book.css("img ::attr(src)").extract_first()],
                'title': book.css("p.bookTitle > a ::text").re_first("\s*((\w+\s?)+)\s*")
            }
class GoodreadsSpider(scrapy.Spider):
    name = "goodreads-covers"
    allowed_domains = ["goodreads.com"]
    start_urls = (
        'http://www.goodreads.com/genres/science-fiction',
    )
    custom_settings = {'IMAGES_STORE': 'imgs/goodreads/sci-fi'}

    def parse(self, response):
        for book in response.css("img.bookImage"):
            yield {
                'title': book.css("::attr(alt)").extract_first(),
                'image_urls': [book.css("::attr(src)").extract_first()]
            }

These spiders scrape metadata from books on a range of websites. Book covers are retrieved using ImagesPipeline, which can be enabled and configured in the project global settings.py file:

ITEM_PIPELINES = {
    'scrapy.pipelines.images.ImagesPipeline': 1
}
IMAGES_STORE = '/some/path/to/images'

The IMAGES_STORE setting lets you define where the downloaded images will be stored in your filesystem. So if you want to separate the images downloaded by each spider into different folders, you just need to override the global IMAGES_STORE setting via custom_settings in each spider:

custom_settings = {'IMAGES_STORE': '/a/different/path'}

Alternatively, you can pass the settings as arguments for the spider using the scrapy -s command line option, but that adds the hassle of having to pass all custom settings from the command line:

$ scrapy crawl alibris -s IMAGES_STORE=imgs/alibris/

So, if you ever need different settings for some spider in your project, include them in the custom_settings attribute in the spider. Heads up, this feature is available for Scrapy >= 1.0.0.

If you are interested in learning how to download images using your spiders, check the docs for more information.

Wrap Up

And that’s about it for our February tips. Check back in with us in March, follow us on Twitter, Facebook, and Instagram, and subscribe to our RSS feed to catch our next Scrapy Tips from the Pros.

Be the first to know. Gain insights. Make better decisions.

Use web data to do all this and more. We’ve been crawling the web since 2010 and can provide you with web data as a service.

Tell me more

Leave a Reply

Your email address will not be published. Required fields are marked *