Web scraping from youtube with nokogiri - ruby

Question

I want to scrape all the names of the users who commented below a youtube video.
I'm using ruby and nokogiri.
require 'rubygems'
require 'nokogiri'
require 'open-uri'
url = "https://www.youtube.com/watch?v=tntOCGkgt98"
doc = Nokogiri::HTML(open(url))
doc.css(".comment-thread-renderer > .comment-renderer").each do |comment|
name = comment.css("#comment-section-renderer-items .g-hovercard").text
puts name
end
But it's not working, I'm not getting any output, no error either.

Answer 1

I won't be able to give you a solution, but at least I can give you a couple of hints that may help you to move forward.
The code you have is not working because the comments section is loaded via an ajax call after the page is loaded. If you do a hard reload in your browser, you will see that there is a spinner icon and a Loading... text in the sections comment, waiting for the content to be loaded. When Nokogiri gets the page via the http request, it gets the html content that you see before the comments are loaded. As a matter of fact the place where the contents will be later added looks like:
<div id="watch-discussion" class="branded-page-box yt-card">
<div id="comment-section-renderer"
class="comment-section-renderer vve-check"
data-visibility-tracking="CCsQuy8iEwjr3P3u1uzNAhXIepAKHRV9D8Ao-B0=">
<div class="action-panel-loading">
<p class="yt-spinner ">
<span class="yt-spinner-img yt-sprite" title="Loading icon">
</span>
<span class="yt-spinner-message">Loading...</span>
</p>
</div>
</div>
</div>
That is the reason why you won't find the divs you are looking for, because they aren't part of the html you have.
Looking at the network console in the browser, it seems that the ajax request to get the comments data is being sent to https://www.youtube.com/watch_fragments_ajax?v=tntOCGkgt98&tr=time&distiller=1&ctoken=EhYSC3RudE9DR2tndDk4wAEAyAEA4AEBGAY%253D&frags=comments&spf=load. As you can see the v parameter is the video id, however there are a couple of caveats:
There is a ctoken param, which you can get by scraping the original page contents. It is inside a <script> tag, in the form of
'COMMENTS_TOKEN': "<token>".
However, you still need to send a session_token as a form data in the body of the AJAX request (which is a POST). That I don't know where is coming from :(.
I think that you will be pushing the limits of Nokogiri here, as AFAIK it is not intended to follow ajax requests or handling Javascript. Maybe the ruby Selenium driver is better suited for this.
HTH

Answer 2

I think you need name.css("#comment-section..."
The each statement will iterate over the elements, using the variable name.
You may want to use node instead of name:
doc.css(".comment-thread-renderer > .comment-renderer").each do |node|
name = node.css("#comment-section-renderer-items .g-hovercard").text
puts name
end

Answer 3

I wrote this rails app using nokogiri to see all the tags that a page has before any javascript is run in the browser. The source code is here, so you can adjust it if you need to add more info about the node in the view.
That can easily tell you if the particular tag element that you are looking for is something you can retrieve without having to do some JS eval.
Most web crawlers don't support client-side rendering, which gives you an idea that it's not a trivial task to execute JS when scraping content.

Answer 4

YouTube is a dynamically rendered JavaScript website, though it could be parsed with Nokogiri without using Selenium or another package. Try open the Network tab in dev tools, scroll to the comment section, and see what request being send.
You need to make a post request in order to fetch comments data. You can preview the output in the "Preview" tab.
Preview output:
Which is equivalent to this comment:
Note: Since this comment brings very little value, this answer will be updated with the attached code once there will be an available solution.