🕷 SEO Crawling & Scraping: Strategies & Recipes

Once you have mastered the basics of using the crawl function, you probably want to achieve more with better customization and control.

These are some code strategies that might be useful to customize how you run your crawls.

Most of these options can be set using the custom_settings parameter that the function takes. This can be set by using a dictionary, where the keys indicate the option you want to set, and the values specify how you want to set them.

How to crawl a list of pages, and those pages only (list mode)?

Simply provide that list as the first argument, for the url_list parameter, and make sure that follow_links=False, which is the default. This simply crawls the given pages, and stops when done.

>>> import advertools as adv
>>> url_list = ['https://example.com/page_1',
...             'https://example.com/page_2',
...             'https://example.com/page_3',
...             'https://example.com/page_4']

>>> adv.crawl(url_list,
...           output_file='example_crawl_1.jl',
...           follow_links=False)

How can I crawl a website including its sub-domains?

The crawl function takes an optional allowed_domains parameter. If not provided, it defaults to the domains of the URLs in url_list. When the crawler goes through the pages of example.com, it follows links to discover pages. If it finds pages on help.exmaple.com it won't crawl them (it's a different domain). The solution, therefore, is to provide a list of domains to the allowed_domains parameter. Make sure you also include the original domain, in this case example.com.

>>> adv.crawl('https://example.com',
...           'example_crawl_1.jl',
...           follow_links=True
...           allowed_domains=['help.example.com', 'example.com', 'community.example.com'])

How can I save a copy of the logs of my crawl for auditing them later?

It's usually good to keep a copy of the logs of all your crawls to check for errors, exceptions, stats, etc. Pass a path of the file where you want the logs to be saved, in a dictionary to the cutom_settings parameter. A good practice for consistency is to give the same name to the output_file and log file (with a different extension) for easier retreival. For example:

output_file: 'website_name_crawl_1.jl'
LOG_FILE: 'website_name_crawl_1.log' (.txt can also work)
output_file: 'website_name_crawl_2.jl'
LOG_FILE: 'website_name_crawl_2.log'
>>> adv.crawl('https://example.com', 'example_crawl_1.jl',
...           custom_settings={'LOG_FILE': 'example_crawl_1.log'})

How can I automatically stop my crawl based on a certain condition?

There are a few conditions that you can use to trigger the crawl to stop, and they mostly have descriptive names:

  • CLOSESPIDER_ERRORCOUNT: You don't want to wait three hours for a crawl to finish, only to discover that you had errors all over the place. Set a certain number of errors to trigger the crawler to stop, so you can investigate the issue.

  • CLOSESPIDER_ITEMCOUNT: Anything scraped from a page is an "item", h1, title , meta_desc, etc. Set the crawler to stop after getting a certain number of items if you want that.

  • CLOSESPIDER_PAGECOUNT: Stop the crawler after a certain number of pages have been crawled. This is useful as an exploratory technique, especially with very large websites. It might be good to crawl a few thousand pages, get an idea on its structure, and then run a full crawl with those insights in mind.

  • CLOSESPIDER_TIMEOUT: Stop the crawler after a certain number of seconds.

>>> adv.crawl('https://example.com', 'example_crawl_1.jl',
...           custom_settings={'CLOSESPIDER_PAGECOUNT': 500})

How can I (dis)obey robots.txt rules?

The crawler obeys robots.txt rules by default. Sometimes you might want to check the results of crawls without doing that. You can set the ROBOTSTXT_OBEY setting under custom_settings:

>>> adv.crawl('https://example.com',
...           'example_crawl_1.jl',
...           custom_settings={'ROBOTSTXT_OBEY': False})

How do I set my User-agent while crawling?

Set this parameter under custom_settings dictionary under the key USER_AGENT. The default User-agent can be found by running adv.spider.user_agent

>>> adv.spider.user_agent # to get the current User-agent
>>> adv.crawl('http://example.com',
...           'example_crawl_1.jl',
...           custom_settings={'USER_AGENT': 'YOUR_USER_AGENT'})

How can I control the number of concurrent requests while crawling?

Some servers are set for high sensitivity to automated and/or concurrent requests, that you can quickly be blocked/banned. You also want to be polite and not kill those servers, don't you?

There are several ways to set that under the custom_settings parameter. The available keys are the following:

CONCURRENT_ITEMS: default 100
CONCURRENT_REQUESTS : default 16
CONCURRENT_REQUESTS_PER_DOMAIN: default 8
CONCURRENT_REQUESTS_PER_IP: default 0
>>> adv.crawl('https://example.com',
...           'example_crawl_1.jl',
...           custom_settings={'CONCURRENT_REQUESTS_PER_DOMAIN': 1})

How can I slow down the crawling so I don't hit the websites' servers too hard?

Use the DOWNLOAD_DELAY setting and set the interval to be waited before downloading consecutive pages from the same website (in seconds).

>>> adv.crawl('https://example.com', 'example_crawl_1.jl',
...           custom_settings={'DOWNLOAD_DELAY': 3}) # wait 3 seconds between pages

How can I set multiple settings to the same crawl job?

Simply add multiple settings to the custom_settings parameter.

>>> adv.crawl('http://example.com',
...           'example_crawl_1.jl',
...           custom_settings={'CLOSESPIDER_PAGECOUNT': 400,
...                            'CONCURRENT_ITEMS': 75,
...                            'LOG_FILE': 'output_file.log'})

How do I pause/resume crawling, while making sure I don't crawl the same page twice?

There are several reasons why you might want to do this:

  • You want to mainly crawl the updates to the site (you already crawled the site).

  • The site is very big, and can't be crawled quickly.

  • You are not in a hurry, and you also don't want to hit the servers hard, so you run your crawl across days for example.

  • As an emergency measure (connection lost, battery died, etc.) you can start where you left off

Handling this is extremely simple, and all you have to do is simply provide a path to a new folder. Make sure it is new and empty, and make sure to only use it for the same crawl job reruns. That's all you have to worry about. The JOBDIR setting handles this.

>>> adv.crawl('http://example.com',
...           'example_crawl_1.jl',
...           custom_settings={'JOBDIR': '/Path/to/en/empty/folder'})

The first time you run the above code and then stop it. Stopping can happen by accident (lost connection, closed computer, etc.), manually (you hit ctrl+C) or you used a custom setting option to stop the crawl after a certain number of pages, seconds, etc.

The second time you want to run this, you simply run the exact same command again. If you check the folder that was created you can see a few files that manage the process. You don't need to worry about any of it. But make sure that folder doesn't get changed manually, rerun the same command as many times as you need, and the crawler should handle de-duplication for you.

How do I use a proxy while crawling?

This requires the following simple steps:

  • Install the 3rd party package scrapy-rotating-proxies. This package handles the proxy rotation for you, in addition to retries, so you don't need to worry about those details.

  • Get a list of proxies and save in a text file, one proxy per line

  • Set a few custom_settings in the crawl function (DOWNLOADER_MIDDLEWARES and ROTATING_PROXY_LIST_PATH)

$ pip install scrapy-rotating-proxies

Save a list of proxies in a text file with the template:

https://username:password@IPADDRESS:PORT

proxies.txt example file (randome values):

https://user123:password123@12.34.56.78:1111 https://user123:password123@12.34.56.78:1112 https://user123:password123@12.34.56.78:1113 https://user123:password123@12.34.56.78:1114

Then, you need to set a few custom_settings in the crawl function:

adv.crawl(
    'https://example.com', 'output_file.jl', follow_links=True,

    custom_settings={
        'DOWNLOADER_MIDDLEWARES': {
            'rotating_proxies.middlewares.RotatingProxyMiddleware': 610,
            'rotating_proxies.middlewares.BanDetectionMiddleware': 620
        },
        'ROTATING_PROXY_LIST_PATH': 'proxies.txt',
    }
)

You can then read the output file normally and see that the proxies are being used:

crawldf = pd.read_json('output_file.jl', lines=True)
crawldf.filter(regex='proxy').head()

proxy

_rotating_proxy

request_headers_proxy-authorization

proxy_retry_times

0

https://123.456.789.101:8893

1

Basic b3VzY214dHg6ODlld29rMGRsdfgt

nan

1

https://123.456.789.101:8894

1

Basic b3VzY214dHg6ODlld29rMGRsdfgt

nan

2

https://123.456.789.101:8895

1

Basic b3VzY214dHg6ODlld29rMGRsdfgt

nan

3

https://123.456.789.101:8896

1

Basic b3VzY214dHg6ODlld29rMGRsdfgt

nan

4

https://123.456.789.101:8897

1

Basic b3VzY214dHg6ODlld29rMGRsdfgt

nan

How can I change the default request headers?

This is a very common use case, and it is very easy to do. Simply add the DEFAULT_REQUEST_HEADERS setting as a dictionary to the custom_settings parameter:

adv.crawl(
    url_list='https://example.com',
    output_file='output.jl',
    custom_settings={
        'DEFAULT_REQUEST_HEADERS': {
            'Accept-Language': 'es',
            'Accept-Encoding': 'gzip, deflate'
        }
    }
)

You can easily check for the actual request headers that were used while crawling. In the crawl DataFrame, simply use the regex pattern request_headers_:

crawldf = pd.read_json('output.jl', lines=True)
crawldf.filter(regex='request_headers_')

request_headers_accept-language

request_headers_accept-encoding

request_headers_user-agent

0

es

gzip, deflate

advertools/0.13.2

XPath expressions for custom extraction

The following are some expressions you might find useful in your crawling, whether you use advertools or not. The first column indicates whether or not the respective expression is used by default by the advertools crawler.

Used by advertools

Suggested Name

XPath Expression

What it does

True

title

//title/text()

Extract the text of title tags

True

meta_desc

//meta[@name='description']/@content

Extract the content attribute of the meta tag which has the name 'description'

True

viewport

//meta[@name='viewport']/@content

Extract the content attribute of the meta tag which has the name 'viewport

True

charset

//meta[@charset]/@charset

Get the meta tag that has the attribute 'charset', extract the charset attribute

True

h1

//h1/text()

Get the h1 tags, extract their text

True

h2

//h2/text()

Get the h2 tags, extract their text

True

h3

//h3/text()

Get the h3 tags, extract their text

True

h4

//h4/text()

Get the h4 tags, extract their text

True

h5

//h5/text()

Get the h5 tags, extract their text

True

h6

//h6/text()

Get the h6 tags, extract their text

True

canonical

//link[@rel='canonical']/@href

Get the link elements with the rel attribute 'canonical', extract their href attributes

True

alt_href

//link[@rel='alternate']/@href

Get the link elements with the rel attribute 'alternate', extract their href attributes

True

alt_hreflang

//link[@rel='alternate']/@hreflang

Get the link elements with the rel attribute 'alternate', extract their hreflang attributes

True

og_props

//meta[starts-with(@property, 'og:')]/@property

Extract all properties of meta tags whos property attribute starts with 'og:' (OpenGraph)

True

og_content

//meta[starts-with(@property, 'og:')]/@content

Extract the content of meta tags whos property attribute starts with 'og:' (OpenGraph)

True

twtr_names

//meta[starts-with(@name, 'twitter:')]/@name

Get meta tags who's name starts with 'twitter:' and extract their name attribute

True

twtr_content

//meta[starts-with(@name, 'twitter:')]/@content

Get meta tags who's name starts with 'twitter:' and extract their content attribute

False

iframe_src

//iframe/@src

Get the iframes, and extract their src attribute

False

gtm_script

//script[contains(@src, 'googletagmanager.com/gtm.js?id=')]/@src

Get the script where the src attribute contains googletagmanager.com/gtm.js?id= and extract its src attribute

False

gtm_noscript

//iframe[contains(@src, 'googletagmanager.com/ns.html?id=')]/@src

Get the iframes where the src attribute contains googletagmanager.com/ns.html?id= and extract the src attribute

False

link_rel_rel

//link[@rel]/@rel

Get all the link elements that have a rel attribute, extract the rel attributes

False

link_rel_href

//link[@rel]/@href

Get all the link elements that have a rel attribute, extract the href attributes

False

link_rel_stylesheet

//link[@rel='stylesheet']/@href

Get all the link elements that have a stylesheet attribute, extract their href attribute

False

css_links

//link[contains(@href, '.css')]/@href

Get the link elements where the href attribute contains .css, extract their href attribute

True

nav_links_text

//nav//a/text()

From the nav element, extract the anchor text of links

True

nav_links_href

//nav//a/@href

From the nav element, extract all the links

True

header_links_text

//header//a/text()

From the header, extract the anchor text of links

True

header_links_href

//header//a/@href

From the header, extract all the links

True

footer_links_text

//footer//a/text()

From the footer, extract the anchor text of links

True

footer_links_href

//footer//a/@href

From the footer, extract all the links

False

js_script_src

//script[@type='text/javascript']/@src

From script tags where the type is text/javascript, extract the src of the script(s)

False

js_script_text

//script[@type='text/javascript']/text()

From script tags where the type is text/javascript, extract the text of the script(s)

False

script_src

//script//@src

Get the src attribute of any <script> tag

False

canonical_parent

name(//link[@rel='canonical']/..)

Get the name of the parent of the link element that has a rel attribute 'canonical'