From: Eric Mahurin <eric_mahurin@...>
Date: 2005-10-27T21:52:56+09:00
Subject: Re: overriding #inspect



--- Sean O'Halpin <sean.ohalpin@gmail.com> wrote:

> On 10/26/05, Eric Mahurin <eric_mahurin@yahoo.com> wrote:
> > > --- Sean O'Halpin <sean.ohalpin@gmail.com> wrote:
> > >
> > > > On 10/26/05, Eric Mahurin <eric_mahurin@yahoo.com>
> wrote:
> > > > > Anybody know what the best way to override inspect
> where
> > > > you
> > > > > may have a recursive data structure?
> > > >
> > > > Check out Tanaka Akira's 'pp' lib (if only to convince
> > > > yourself that
> > > > there's no silver bullet!)
> >
> > Actually, as soon as I allowed myself to make an instance
> > variable, it didn't seem so bad.  Here's what I came up
> with:
> >
> > def inspect
> >    to_s[0..-2].concat(" #{
> >        if defined?(@inspecting)
> >            "..."
> >        else
> >            begin
> >                @inspecting = true
> >                ... generate inspect contents here ...
> >            ensure
> >                remove_instance_variable(:@inspecting)
> >            end
> >        end
> >    }>")
> > end
> 
> But as soon as you start using flags to guard against cycles
> you have
> to at least acknowledge that it won't be thread safe, unless
> you take
> an approach like 'pp' (which stashes object_id in
> Thread.current[InspectKey]).

Yep.  Here's one that uses a Hash to effectively make thread
local instance variables:

def inspect # :nodoc:
    to_s[0..-2].concat(" #{
        if (@inspecting||=Hash.new)[Thread.current]
            "..."
        else
            begin
                @inspecting[Thread.current] = true
                ... generate inspect contents here ...
            ensure
                @inspecting.delete(Thread.current)
            end
        end
    }>")
end

This seems better that storing this info in thread attributes
(global variables to a thread).  Better encapsulation.

I'm not sure how much of these operations on @inspecting are
atomic, so we may need to use Thread.critical to fix this.  Can
we make any assumptions about what is atomic?



	
		
__________________________________ 
Yahoo! Mail - PC Magazine Editors' Choice 2005 
http://mail.yahoo.com