Anyone scraping dynamic AJAX sites?


#1

Hello.

Is there anyone who has successfully found a way to scrape a dynamically
generated AJAX web site? If I view the source, it gives me the
variables. If I use Firebug to view the DOM, it gives me the actual
values. Any ideas?

Thanks.


#2

The problem is you need a DOM-aware Javascript interpreter in your
code to execute the Javascript, manipulate the DOM in the HTML, and
then allow you to extract the data you need.

There are projects like Rhino, which is a Javascript engine you can
embed in other apps, but you still won’t have the DOM of the page nor
will you be able to manipulate it then extract the values, at least as
far as I understand.

You could use something like Ruby driving some sort of WebKit
interface on Mac OS or Linux, but I have no idea where to start. That,
to me, seems like the best answer. Maybe even a Ruby-based Cocoa app
would be the trick.


#3

On Sat, Nov 29, 2008 at 7:25 PM, Becca G. removed_email_address@domain.invalid wrote:

Hello.

Is there anyone who has successfully found a way to scrape a dynamically
generated AJAX web site? If I view the source, it gives me the
variables. If I use Firebug to view the DOM, it gives me the actual
values. Any ideas?

http://code.google.com/p/firewatir/


#4

scRUBYt! - http://scrubyt.org

e.g. scraping your linkedin contacts:

require ‘rubygems’
require ‘scrubyt’

property_data = Scrubyt::Extractor.define :agent => :firefox do

fetch ‘https://www.linkedin.com/secure/login
fill_textfield ‘session_key’, ‘
fill_textfield ‘session_password’, '

submit

click_link_and_wait ‘Connections’, 5

vcard “//li[@class=‘vcard’]” do
first_name “//span[@class=‘given-name’]”
second_name “//span[@class=‘family-name’]”
email “//a[@class=‘email’]”
end

end

puts property_data.to_xml

Cheers,
Peter


http://www.rubyrailways.com
http://scrubyt.org


#5

Just for completeness sake: scRUBYt! (since 0.4.05) is using FireWatir
as the agent (or mechanize - you can choose whether you want scrape
AJAX or not) so you can do full blown AJAX scraping - but with a
scraping DSL which usually speeds up the scraper creation, especially
in the case of complicated scrapers.

Cheers,
Peter


http://www.rubyrailways.com
http://scrubyt.org


#6

On Sat, Nov 29, 2008 at 6:25 PM, Becca G. removed_email_address@domain.invalid wrote:

Hello.

Is there anyone who has successfully found a way to scrape a dynamically
generated AJAX web site? If I view the source, it gives me the
variables. If I use Firebug to view the DOM, it gives me the actual
values. Any ideas?

Thanks.

Posted via http://www.ruby-forum.com/.

As gf pointed out, the problem is you need a full DOM and working
javascript for this, sometimes even working css, to really do it
properly, you need a full blown, fully supported, web browser.

Short story, use the WATIR library to interact with your browser’s DOM
to do this.

http://wtr.rubyforge.org/

I used to do this all the time for work, in a testing capacity. I
tried a number of diferent solutions, and found WATIR far superior to
anything else out there, including the very pricey pay packages. If
you cut through all the marketing BS, half the pay-packages are
functional the same as WATIR, and the other half are more primitive.

–Kyle


#7

On Mon, Dec 1, 2008 at 3:48 AM, Peter S. removed_email_address@domain.invalid
wrote:

Just for completeness sake: scRUBYt! (since 0.4.05) is using FireWatir as
the agent (or mechanize - you can choose whether you want scrape AJAX or
not) so you can do full blown AJAX scraping - but with a scraping DSL which
usually speeds up the scraper creation, especially in the case of
complicated scrapers.

Cheers,
Peter

Peter
Neat. I’ll have to give that a try next time I need to revisit
scraping.


#8

On Mon, Dec 1, 2008 at 10:23 AM, Florian G. removed_email_address@domain.invalid
wrote:

personal session) and usability on a deployment server.

Regards,
Florian G.

[1]: Preferably a virtal one on a console-only machine.
[2]: Sadly, afaik, firefox has no hidden-mode.

http://coderrr.wordpress.com/2007/10/15/patch-to-firewatir-and-jssh-to-support-testing-with-multiple-concurrent-firefox-browsers/


#9

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Actually, firewatir and scRUBYt! are nice.

But is there a possibility to start firefox with a second profile (so
that it circumvents the “one instance”-rule) and rendering to a hidden
display? [1][2]

Otherwise, this really hurts testablity (as the browser might retain
your personal session) and usability on a deployment server.

Regards,
Florian G.

[1]: Preferably a virtal one on a console-only machine.
[2]: Sadly, afaik, firefox has no hidden-mode.

On Dec 1, 2008, at 10:48 AM, Peter S. wrote:

http://scrubyt.org

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.8 (Darwin)

iEYEARECAAYFAkk0ApYACgkQyLKU2FMxSOJHmgCgjZnIkD/c4yoq//bcF31fOpD7
80sAnRNrqb3QBtaOpVJCE0Z8LgNb1TIz
=k3XY
-----END PGP SIGNATURE-----


#10

[1]: Preferably a virtal one on a console-only machine.
[2]: Sadly, afaik, firefox has no hidden-mode.

You could try using a virtual frame buffer if you are using Linux or
similar.

Xfvb :99 -ac &
export DISPLAY=:99

Will


#11

If the site is truely AJAX, i.e. the data is loaded from an HTTP call
from
JavaScript, you could monitor the HTTP requests made by the browser. On
Firefox, I use the LiveHTTPHeaders extension. Just load go
view–>sidebar–>HTTP Headers, load the page with whatever data, and
look
thru the requests for anything interesting.

I used this method to get Facebook contact info and it worked fairly
well.
As a bonus, any data found with this method is usually in a very
machine-understandable format like JSON or RSS. There are Ruby
libraries
for both.

Dan


#12

On Sat, Dec 20, 2008 at 7:27 AM, Will S. removed_email_address@domain.invalid wrote:

[1]: Preferably a virtal one on a console-only machine.
[2]: Sadly, afaik, firefox has no hidden-mode.

I’ve never used it, but Celerity appears to have Javascript support:

http://celerity.rubyforge.org/

You could try using a virtual frame buffer if you are using Linux or
similar.

Xfvb :99 -ac &
export DISPLAY=:99

Or, start a vncserver with xstartup set to launch the scraper script.