From ngrandy at gmail.com  Sat May  3 04:27:31 2008
From: ngrandy at gmail.com (Nick Grandy)
Date: Sat, 3 May 2008 01:27:31 -0700
Subject: [Mechanize-users] marshalling mechanized
Message-ID: <9a6cef0c0805030127p38008ac5mbe218f4d186520f8@mail.gmail.com>

hi all,

in the course of my hacking i wanted to marshal mechanized objects
(whether the agent, page, or other).  this may be a rather particular
use-case, but i figured i would report.  marshalling does not work on
certain objects, in particular IO objects, and since mechanized uses
Net::HTTP it makes use of IO objects.  so it turns out that it you
just do

a = WWW::Mechanize.new
page = a.get "http://nytimes.com"
Marhshal.dump(a)

you get a problem.  in particular, "TypeError: can't dump TCPSocket".
basically, don't try to marshal a socket.

however, if we make sure to finish the http connection after each use,
then the agent can be marshalled safely.  i accomplished this by
inserting:
http_obj.finish

immediately prior to this line:
      return page if res_klass <= Net::HTTPSuccess
in mechanize.rb, which is approximately line 587 in v 0.7.0

this is a crude hack that doesn't, i think, respect keep-alive
connections, but it does at least work for the purpose of ensuring the
agent can be marshalled.  i invite thoughts on more elegant solutions!
nick