5 Advanced Techniques
5.1 Changing User-Agents
Part of what shows up in web traffic logs is the user-agent, which typically says something about your operating system and browser. If you Google “my user-agent”, you’ll see yours. Here is mine:
“Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/93.0.4577.63 Safari/537.36”
The default user-agent with
bow() is “polite R package - https://github.com/dmi3kno/polite”.
It is recommended that you change this to give your contact information, so that the website admin can contact you if they see your scrapes in their logs:
bow(url, user_agent = "I am scraping your site for a research project. You can contact me at email@example.com")
But some sites may check your user-agent to see if the request is coming from a browser, and not allow content to load. In this case, neither
bow()’s default nor our friendly message will register as a real request from a browser. (For an example, compare what you see when you visit http://www.scrapethissite.com/pages/advanced/?gotcha=headers in a browser versus when you scrape it.)
To get around this issue, you can simply use a normal user-agent when scraping, such as,
bow(url, user_agent = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/93.0.4577.63 Safari/537.36")
Another strategy is user-agent rotation, where we use a different user-agent for each page we scrape. To implement this, copy (or scrape!) agents from this list of agents, place them in a vector called
agents and then sample from this vector each scrape:
dat <- map(url, ~ bow(.x, user_agent = sample(agents, 1)) |> scrape())
Even if you do this, your IP address is still in the web traffic logs, so you are not anonymous. The way around this would be to use proxy servers, which is out of scope for this introduction.
At this point, though, if you are spoofing and rotating your user-agents and hiding your IP address, you should stop and check if you are adhering to the site’s Terms of Service and robots.txt. See the section Being a Good Scraper of the Web, and also the court case eBay v. Bidder’s Edge before you consider taking such steps to scrape a site.
5.2 Long Scrapes
If the job will take more than a few minutes to scrape, submit it as a batch job to Linstat. Of course, first test your script on a few pages to make sure it works fine!