From: "Jesús Gabriel y Galán" <jgabrielygalan@...>
Date: 2007-10-09T16:32:42+09:00
Subject: Re: Question about Mechanize and Google API

On 10/9/07, mortee <mortee.lists@kavemalna.hu> wrote:
> Richard Conroy wrote:
> > From my experience Mechanize is a bit low level, especially in this area.
> > You are expected to write the code yourself to handle, redirects,
> > authentication,
> > SSL etc.
> >
> > You might want to try open-uri or Hpricot as well. But I think there is no
> > right-sized solution - you will still need to do a bit of code writing yourself.
>
> Just for the record, last time I checked, Mechanize *did* use Hpricot to
> parse pages. AFAIK, you can even provide Mechainze with your own page
> parser class before fetching the stuff.

Yes, Mechanize parses pages with Hpricot, and also allows you to use
an Hpricot-like API on them for search:

agent.get('http://rubyforge.org').search("//strong")

The problem I was having with it is the fact that I didn't know how to
specify arbitrary headers for my request, or make the Google
authentication stuff to work.

Working directly with net/https worked for me (code "inspired" by the
Gdata gem :-):

require 'net/http'
require 'net/https'
require 'uri'
require 'hpricot'

module Net
  class HTTPS < HTTP
    def initialize(address, port = nil)
      super(address, port)
      self.use_ssl = true
    end
  end
end


h = {"Email" => "xxxxx@gmail.com",
  "Passwd" => "xxxxx",
  "service" => "wise",
  "source" => "personal-test-1.0",
  "accountType" => "GOOGLE"
}

url = URI.parse('https://www.google.com/accounts/ClientLogin')
response = Net::HTTPS.post_form(url, h)
auth = ""
response.body.each do |line|
  if line =~ /Auth=/
    auth = line.split("=")[1].chomp
  end
end

headers = {"Authorization" => "GoogleLogin auth=#{auth}"}
http = Net::HTTP.new("spreadsheets.google.com", 80)
response, data = http.get("/feeds/spreadsheets/private/full", headers)
document = Hpricot(data)
document.search("//entry").each do |entry|
   p entry
end

This gives you a document with the list of private spreadsheets of the
user. Then you can use Hpricot or whatever you want to parse the atom
feed.

Jesus.