I'm thinking of trying Beautiful Soup [1], a Python package for HTML scraping. Are there any other HTML scraping packages I should be looking at? Python is not a requirement, I'm actually interested in hearing about other languages as well.
The story so far:
The Ruby world's equivalent to Beautiful Soup is why_the_lucky_stiff's Hpricot [1].
[1] https://github.com/hpricot/hpricotIn the .NET world, I recommend the HTML Agility Pack. Not near as simple as some of the above options (like HTMLSQL), but it's very flexible. It lets you maniuplate poorly formed HTML as if it were well formed XML, so you can use XPATH or just itereate over nodes.
http://www.codeplex.com/htmlagilitypack
BeautifulSoup is a great way to go for HTML scraping. My previous job had me doing a lot of scraping and I wish I knew about BeautifulSoup when I started. It's like the DOM with a lot more useful options and is a lot more pythonic. If you want to try Ruby they ported BeautifulSoup calling it RubyfulSoup but it hasn't been updated in a while.
Other useful tools are HTMLParser or sgmllib.SGMLParser which are part of the standard Python library. These work by calling methods every time you enter/exit a tag and encounter html text. They're like Expat if you're familiar with that. These libraries are especially useful if you are going to parse very large files and creating a DOM tree would be long and expensive.
Regular expressions aren't very necessary. BeautifulSoup handles regular expressions so if you need their power you can utilize it there. I say go with BeautifulSoup unless you need speed and a smaller memory footprint. If you find a better HTML parser on Python, let me know.
I found HTMLSQL [1] to be a ridiculously simple way to screenscrape. It takes literally minutes to get results with it.
The queries are super-intuitive - like:
SELECT title from img WHERE $class == 'userpic'
There are now some other alternatives that take the same approach.
[1] http://www.jonasjohn.de/lab/htmlsql.htmThe Python lxml [1] library acts as a Pythonic binding for the libxml2 and libxslt libraries. I like particularly its XPath support and pretty-printing of the in-memory XML structure. It also supports parsing broken HTML. And I don't think you can find other Python libraries/bindings that parse XML faster than lxml.
[1] http://codespeak.net/lxml/For Perl, there's WWW::Mechanize.
Python has several options for HTML scraping in addition to Beatiful Soup. Here are some others:
WWW:Mechanize
. Gives you a browser like object to ineract with web pageslibwww
. Supports various options to traverse and select elements (e.g.
XPath
[3] and CSS selection)'Simple HTML DOM Parser' is a good option for PHP, if your familiar with jQuery or JavaScript selectors then you will find yourself at home.
Find it here [1]
There is also a blog post about it here. [2]
[1] http://simplehtmldom.sourceforge.net/Why has no one mentioned JSOUP yet for Java? http://jsoup.org/
The templatemaker [1] utility from Adrian Holovaty (of Django [2] fame) uses a very interesting approach: You feed it variations of the same page and it "learns" where the "holes" for variable data are. It's not HTML specific, so it would be good for scraping any other plaintext content as well. I've used it also for PDFs and HTML converted to plaintext (with pdftotext and lynx, respectively).
[1] http://code.google.com/p/templatemaker/I know and love Screen-Scraper [1].
Screen-Scraper is a tool for extracting data from websites. Screen-Scraper automates:
* Clicking links on websites
* Entering data into forms and submitting
* Iterating through search result pages
* Downloading files (PDF, MS Word, images, etc.)
Common uses:
* Download all products, records from a website
* Build a shopping comparison site
* Perform market research
* Integrate or migrate data
Technical:
* Graphical interface--easy automation
* Cross platform (Linux, Mac, Windows, etc.)
* Integrates with most programming languages (Java, PHP, .NET, ASP, Ruby, etc.)
* Runs on workstations or servers
Three editions of screen-scraper:
* Enterprise: The most feature-rich edition of screen-scraper. All capabilities are enabled.
* Professional: Designed to be capable of handling most common scraping projects.
* Basic: Works great for simple projects, but not nearly as many features as its two older brothers.
[1] http://www.screen-scraper.comI would first find out if the site(s) in question provide an API server or RSS Feeds for access the data you require.
Scraping Stack Overflow is especially easy with Shoes [1] and Hpricot [2].
require 'hpricot'
Shoes.app :title => "Ask Stack Overflow", :width => 370 do
SO_URL = "http://stackoverflow.com"
stack do
stack do
caption "What is your question?"
flow do
@lookup = edit_line "stackoverflow", :width => "-115px"
button "Ask", :width => "90px" do
download SO_URL + "/search?s=" + @lookup.text do |s|
doc = Hpricot(s.response.body)
@rez.clear()
(doc/:a).each do |l|
href = l["href"]
if href.to_s =~ /\/questions\/[0-9]+/ then
@rez.append do
para(link(l.inner_text) { visit(SO_URL + href) })
end
end
end
@rez.show()
end
end
end
end
stack :margin => 25 do
background white, :radius => 20
@rez = stack do
end
end
@rez.hide()
end
end
[1] http://code.whytheluckystiff.net/shoes/Another option for Perl would be Web::Scraper [1] which is based on Ruby's Scrapi [2]. In a nutshell, with nice and concise syntax, you can get a robust scraper directly into data structures.
[1] http://search.cpan.org/~miyagawa/Web-Scraper/lib/Web/Scraper.pmI've had some success with HtmlUnit [1], in Java. It's a simple framework for writing unit tests on web UI's, but equally useful for HTML scraping.
[1] http://htmlunit.sourceforge.netYahoo! Query Language or YQL can be used alongwith jQuery, AJAX, JSONP to screen scrape web pages [1]
[1] http://projects.ischool.washington.edu/tabrooks/343INFOAutumn09/JSONP/jsonpJqueryYQL.htmThere is this solution too: netty HttpClient [1]
[1] http://docs.jboss.org/netty/3.2/xref/org/jboss/netty/example/http/snoop/HttpClient.htmlI use Hpricot on Ruby. As an example this is a snippet of code that I use to retrieve all book titles from the six pages of my HireThings account (as they don't seem to provide a single page with this information):
pagerange = 1..6
proxy = Net::HTTP::Proxy(proxy, port, user, pwd)
proxy.start('www.hirethings.co.nz') do |http|
pagerange.each do |page|
resp, data = http.get "/perth_dotnet?page=#{page}"
if resp.class == Net::HTTPOK
(Hpricot(data)/"h3 a").each { |a| puts a.innerText }
end
end
end
It's pretty much complete. All that comes before this are library imports and the settings for my proxy.
I've used Beautiful Soup a lot with Python. It is much better than regular expression checking, because it works like using the DOM [1], even if the HTML is poorly formatted. You can quickly find HTML tags and text with simpler syntax than regular expressions. Once you find an element, you can iterate over it and its children, which is more useful for understanding the contents in code than it is with regular expressions. I wish Beautiful Soup existed years ago when I had to do a lot of screenscraping -- it would have saved me a lot of time and headache since HTML structure was so poor before people started validating it.
[1] http://en.wikipedia.org/wiki/Document_Object_ModelAlthough it was designed for .NET [1] web-testing, I've been using the WatiN [2] framework for this purpose. Since it is DOM-based, it is pretty easy to capture HTML, text, or images. Recentely, I used it to dump a list of links from a MediaWiki [3] All Pages namespace query into an Excel spreadsheet. The following VB.NET [4] code fragement is pretty crude, but it works.
Sub GetLinks(ByVal PagesIE As IE, ByVal MyWorkSheet As Excel.Worksheet)
Dim PagesLink As Link
For Each PagesLink In PagesIE.TableBodies(2).Links
With MyWorkSheet
.Cells(XLRowCounterInt, 1) = PagesLink.Text
.Cells(XLRowCounterInt, 2) = PagesLink.Url
End With
XLRowCounterInt = XLRowCounterInt + 1
Next
End Sub
[1] http://en.wikipedia.org/wiki/.NET_FrameworkI have used LWP [1] and HTML::TreeBuilder [2] with Perl and have found them very useful.
LWP (short for libwww-perl) lets you connect to websites and scrape the HTML, you can get the module here [3] and the O'Reilly book seems to be online here [4].
TreeBuilder allows you to construct a tree from the HTML, and documentation and source are available in HTML::TreeBuilder - Parser that builds a HTML syntax tree [5].
There might be too much heavy-lifting still to do with something like this approach though. I have not looked at the Mechanize module [6] suggested by another answer, so I may well do that.
[1] http://en.wikipedia.org/wiki/Library_for_WWW_in_PerlYou would be a fool not to use Perl.. Here come the flames..
Bone up on the following modules and ginsu any scrape around.
use LWP
use HTML::TableExtract
use HTML::TreeBuilder
use HTML::Form
use Data::Dumper
Implementations of the HTML5 parsing algorithm [1]: html5lib [2] (Python, Ruby), Validator.nu HTML Parser [3] (Java, JavaScript; C++ in development), Hubbub [4] (C), Twintsam [5] (C#; upcoming).
[1] http://www.whatwg.org/specs/web-apps/current-work/multipage/parsing.htmlWell, if you want it done from the client side using only a browser you have jcrawl.com [1]. After having designed your scrapping service from the web application (http://www.jcrawl.com/app.html), you only need to add the generated script to an HTML page to start using/presenting your data.
All the scrapping logic happens on the the browser via JavaScript. I hope you find it useful. Click this link for a live example that extracts the latest news from Yahoo tennis [2].
[1] http://www.jcrawl.comYou probably have as much already, but I think this is what you are trying to do:
from __future__ import with_statement
import re, os
profile = ""
os.system('wget --no-cookies --header "Cookie: soba=(SeCreTCODe)" http://stackoverflow.com/users/30/myProfile.html')
with open("myProfile.html") as f:
for line in f:
profile = profile + line
f.close()
p = re.compile('summarycount">(\d+)</div>') #Rep is found here
print p
m = p.search(profile)
print m
print m.group(1)
os.system("espeak \"Rep is at " + m.group(1) + " points\""
os.remove("myProfile.html")
I've had mixed results in .NET using SgmlReader which was originally started by Chris Lovett [1] and appears to have been updated by MindTouch [2].
[1] http://robgarrett.com/cs/blogs/software/archive/2005/08/09/1499.aspxI've also had great success using Aptana's Jaxer + jQuery to parse pages. It's not as fast or 'script-like' in nature, but jQuery selectors + real JavaScript/DOM is a lifesaver on more complicated (or malformed) pages.
I like Google Spreadsheets' ImportXML(URL, XPath) function.
It will repeat cells down the column if your XPath expression returns more than one value.
You can have up to 50 importxml()
functions on one spreadsheet.
RapidMiner's Web Plugin is also pretty easy to use. It can do posts, accepts cookies, and can set the user-agent [1].
[1] http://en.wikipedia.org/wiki/User_agentRegular expressions work pretty well for HTML scraping as well ;-) Though after looking at Beautiful Soup, I can see why this would be a valuable tool.
Scrubyt [1] uses Ruby and Hpricot to do nice and easy web scraping. I wrote a scraper for my university's library service using this in about 30 minutes.
[1] https://github.com/scrubber/scrubytFor more complex scraping applications, I would recommend the IRobotSoft web scraper. It is a dedicated free software for screen scraping. It has a strong query language for HTML pages, and it provides a very simple web recording interface that will free you from many programming effort.
The recent talk by Dav Glass Welcome to the Jungle! (YUIConf 2011 Opening Keynote) [1] shows how you can use YUI [2] 3 on Node.js [3] to do clientside-like programming (with DOM selectors instead of string processing) on the server. It is very impressive.
[1] http://developer.yahoo.com/yui/theater/video.php?v=glass-nodeI've been using Feedity - http://feedity.com for some of the scraping work (and conversion into RSS feeds) at my library. It works well for most webpages.
I do a lot of advanced web scraping so wanted to have total control over my stack and understand the limitations. This webscraping library [1] is the result.
[1] http://code.google.com/p/webscraping/I made a very nice library Internet Tools [1] for web scraping.
The idea is to match a template against the web page, which will extract all data from the page and also validate if the page structure is unchanged.
So you can just take the HTML of the web page you want to process, remove all dynamical or irrelevant content and annotate the interesting parts.
E.g. the HTML for a new question on the stackoverflow.com index page is:
<div id="question-summary-11326954" class="question-summary narrow">
<!-- skipped, this is getting too long -->
<div class="summary">
<h3><a title="Some times my tree list have vertical scroll ,then I scrolled very fast and the tree list shivered .Have any solution for this.
" class="question-hyperlink" href="/questions/11326954/about-scroll-bar-issue-in-tree">About Scroll bar issue in Tree</a></h3>
<!-- skipped -->
</div>
</div>
So you just remove this certain id, title and summary, to create a template that will read all new questions in title, summary, link-arrays:
<t:loop>
<div class="question-summary narrow">
<div class="summary">
<h3>
<a class="question-hyperlink">
{title:=text(), summary:=@title, link:=@href}
</a>
</h3>
</div>
</div>
</t:loop>
And of course it also supports the basic techniques, CSS 3 selectors, XPath 2 and XQuery 1 expressions.
The only problem is that I was so stupid to make it a Free Pascal [2] library. But there is also language independent web demo [3].
[1] http://www.benibela.de/sources_en.html#internettoolsFor those that would prefer a graphical workflow tool, RapidMiner (FOSS) has a nice web crawling and scraping facility.
Here's a series of videos:
http://vancouverdata.blogspot.com/2011/04/rapidminer-web-crawling-rapid-miner-web.html
When it comes to extracting data from an HTML document on the server-side, Node.js [1] is a fantastic option. I have used it successfully with two modules called request [2] and cheerio [3].
You can see an example how it works here [4].
[1] http://en.wikipedia.org/wiki/Node.jsIt's basically jQuery for C#. It depends on HTML Agility Pack [2] for parsing the HTML.
[1] http://BeautifulSoup
XPathExpression
can be used (which exists since Java 1.5) - Mark Jeronimus