Mechanize users - Jul 2007 - WWW::Mechanize::Link.inspect needs some TLC

The problem: users trying to debug Mechanize apps with Komodo
are finding the debugger times out once it''s loaded a web
page.  They don''t run into this in the ruby-debug debugger,
or running in normal mode.

The reason: Komodo''s debugger is graphical, which means that
whenever it hits a breakpoint it automatically shows the
contents of each local variable.  It has a limit on how
much data it will retrieve, but it currently doesn''t guard
an object from loading too much data (which it should, but
that''s a separate bug).

I can duplicate the cause of this crash in ruby-debug as
well.  Here''s a sample session, with this code:

require ''rubygems''
require ''mechanize''
require ''logger''

agent = WWW::Mechanize.new { |a| a.log = Logger.new("mech.log") }
agent.user_agent_alias = ''Mac Safari''
page = agent.get("http://www.google.com/")
search_form = page.forms.name("f").first
search_form.fields.name("q").value = "bratislava tournament"
search_results = agent.submit(search_form)
puts search_results.body

Given this ruby-debug session:

bugs $ rdebug mechanize01.rb
./mechanize01.rb:1 require ''rubygems''
(rdb:1) b 11
Set breakpoint 1 at mechanize01.rb:11
(rdb:1) c
Breakpoint 1 at mechanize01.rb:11
./mechanize01.rb:11 puts search_results.body
(rdb:1) p page.links[0].inspect.size
1521039
# That''s way too big, since the page is a simple google results page:
(rdb:1) p page.body.size
3441

Using mechanize/inspect.rb:

(rdb:1) p page.links[0].pretty_inspect
"#<WWW::Mechanize::Link\n \"iGoogle\"\n
\"/url?sa=p&pref=ig&pval=3&q=http://www.google.ca/ig%3Fhl%3Den&usg=AFQjCNH9TTed08sJL_DKraFsuSMDFvW1gw\">\n"
(rdb:1) p page.links[0].pretty_inspect.size
138
(rdb:1) p page.inspect.size
1480219
(rdb:1) p page.pretty_inspect.size
2172

With this change:
--- mechanize/inspect.rb~       2007-07-12 10:55:20.375000000 -0700
+++ mechanize/inspect.rb        2007-07-12 11:42:58.203125000 -0700
@@ -40,6 +40,7 @@
            }
          }
        end
+      alias :inspect  :pretty_inspect
      end

      class Link
@@ -49,6 +50,7 @@
            q.breakable; q.pp href
          }
        end
+      alias :inspect  :pretty_inspect
      end

      class Form
lib $

I get these much better results:

(rdb:1) p page.links[0].inspect.size
138
(rdb:1) p page.body.size
3441
(rdb:1) p page.inspect.size
2172

Is this patch reasonable or have I missed something?

Thanks,
Eric

On 7/12/07, Eric Promislow <ericp at activestate.com>
wrote:
>
> The problem: users trying to debug Mechanize apps with Komodo
> are finding the debugger times out once it''s loaded a web
> page.  They don''t run into this in the ruby-debug debugger,
> or running in normal mode.
>
> The reason: Komodo''s debugger is graphical, which means that
> whenever it hits a breakpoint it automatically shows the
> contents of each local variable.  It has a limit on how
> much data it will retrieve, but it currently doesn''t guard
> an object from loading too much data (which it should, but
> that''s a separate bug).
>
> I can duplicate the cause of this crash in ruby-debug as
> well.  Here''s a sample session, with this code:
>
> require ''rubygems''
> require ''mechanize''
> require ''logger''
>
> agent = WWW::Mechanize.new { |a| a.log = Logger.new("mech.log") }
> agent.user_agent_alias = ''Mac Safari''
> page = agent.get("http://www.google.com/")
> search_form = page.forms.name("f").first
> search_form.fields.name("q").value = "bratislava
tournament"
> search_results = agent.submit(search_form)
> puts search_results.body
>
> Given this ruby-debug session:
>
> bugs $ rdebug mechanize01.rb
> ./mechanize01.rb:1 require ''rubygems''
> (rdb:1) b 11
> Set breakpoint 1 at mechanize01.rb:11
> (rdb:1) c
> Breakpoint 1 at mechanize01.rb:11
> ./mechanize01.rb:11 puts search_results.body
> (rdb:1) p page.links[0].inspect.size
> 1521039
> # That''s way too big, since the page is a simple google results
page:
> (rdb:1) p page.body.size
> 3441
>
> Using mechanize/inspect.rb:
>
> (rdb:1) p page.links[0].pretty_inspect
> "#<WWW::Mechanize::Link\n \"iGoogle\"\n
\"/url?sa=p&pref=ig&pval=3&q>
http://www.google.ca/ig%3Fhl%3Den&usg=AFQjCNH9TTed08sJL_DKraFsuSMDFvW1gw\
> ">\n"
> (rdb:1) p page.links[0].pretty_inspect.size
> 138
> (rdb:1) p page.inspect.size
> 1480219
> (rdb:1) p page.pretty_inspect.size
> 2172
>
> With this change:
> --- mechanize/inspect.rb~       2007-07-12 10:55:20.375000000 -0700
> +++ mechanize/inspect.rb        2007-07-12 11:42:58.203125000 -0700
> @@ -40,6 +40,7 @@
>             }
>           }
>         end
> +      alias :inspect  :pretty_inspect
>       end
>
>       class Link
> @@ -49,6 +50,7 @@
>             q.breakable; q.pp href
>           }
>         end
> +      alias :inspect  :pretty_inspect
>       end
>
>       class Form
> lib $
>
> I get these much better results:
>
> (rdb:1) p page.links[0].inspect.size
> 138
> (rdb:1) p page.body.size
> 3441
> (rdb:1) p page.inspect.size
> 2172
>
> Is this patch reasonable or have I missed something?

This patch seems reasonable.  I''ll apply it and make sure the tests
pass.
:-)


-- 
Aaron Patterson
http://tenderlovemaking.com/
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://rubyforge.org/pipermail/mechanize-users/attachments/20070713/92713504/attachment.html