From: Trans <transfire@...>
Date: 2007-07-28T06:51:02+09:00
Subject: Re: Facets error while using Time.parse



On Jul 27, 1:55 pm, Ezra Zygmuntowicz <ezmob...@gmail.com> wrote:
> On Jul 27, 2007, at 12:34 PM, Trans wrote:
>
>
>
>
>
> > On Jul 27, 10:40 am, "mer...@gmail.com" <mer...@gmail.com> wrote:
> >> I have install latest facets using gem. I am having this issue with
> >> Time.parse with facets. Any ideas?
>
> >> [spatel@taamserver ~]$ ruby --version
> >> ruby 1.8.6 (2007-06-07 patchlevel 36) [i686-linux]
>
> >> Everything is fine here:
>
> >> [spatel@taamserver ~]$ cat x1.rb
> >> #require 'facets'
> >> p Time.parse("3:40:40")
> >> p Time.parse('4:41:41')
> >> p Time.parse('5:42:42')
> >> [spatel@taamserver ~]$ ruby x1.rb
> >> Fri Jul 27 03:40:40 -0500 2007
> >> Fri Jul 27 04:41:41 -0500 2007
> >> Fri Jul 27 05:42:42 -0500 2007
>
> >> Same program with first line uncommented
>
> >> [spatel@taamserver ~]$ cat x1.rb
> >> require 'facets'
> >> p Time.parse("3:40:40")
> >> p Time.parse('4:41:41')
> >> p Time.parse('5:42:42')
> >> [spatel@taamserver ~]$ ruby x1.rb
> >> Fri Jul 27 03:40:40 -0500 2007
> >> Fri Jul 27 13:24:55 -0500 2007
> >> Fri Jul 27 13:24:55 -0500 2007
>
> > Want to see the culprit? Would you believe I tracked it to this:
>
> >   class Symbol
> >     # Same functionality as before, just a touch more efficient.
>
> >     def to_s
> >       @to_s || (@to_s = id2name)
> >     end
> >   end
>
> > I'm not sure how exactly that could break anything. But somehow it
> > does. Anyone?
>
> > A quick fix, remark out the line in lib/facets.rb (in facets' gem
> > dir):
>
> >   #require 'facets/core/symbol/to_s'
>
> > T.
>
> Trans-
>
>         I ran into this same thing, I had defined the same Symbol#to_s in
> merb. ActifveRecord was dropping created_at and updated_at attributes
> after the first time you would access them.
>
>         Turns out it was some code deep in the Date stdlib that does some
> weird concatenating of symbols that was not compatible with
> Symbol#to_s as you have it defined. I had to remove it from my
> library because it caused very hard to track down bugs.

Thanks Ezra. It's gone now. I always felt a little off about that
method, but I could never see any reason not to have it. I benchmarked
and it did make symbol to_s slightly faster. Oh well I guess.

I can't make sense of it either, trying:

 class Symbol
   def to_s
     @_to_s ||= id2name
     puts "%10s %10s %10s %10s" % [inspect, id2name.inspect,
@_to_s.inspect, (id2name == @_to_s) ]
     @_to_s
     #id2name
   end
 end

Gives me this nutty result:

   :_comp=   "_comp="   "_comp="       true
    :hour=    "hour="    "hour="       true
     :min=     "min="     "min="       true
     :sec=     "sec="     "sec="       true
    :_comp    "_comp"    "_comp"       true
     :zone     "zone"     "zone"       true
    :_comp    "_comp"    "_comp"       true
      :sec      "sec"      "sec"       true
     :hour     "hour"     "hour"       true
      :min      "min"      "min"       true
  Fri Jul 27 03:40:40 -0700 2007
   :_comp=   "_comp="    "_comp"      false
    :hour=    "hour="     "hour"      false
     :min=     "min="      "min"      false
     :sec=     "sec="      "sec"      false
    :_comp    "_comp"    "_comp"       true
     :zone     "zone"     "zone"       true
  Fri Jul 27 14:45:18 -0700 2007
   :_comp=   "_comp="    "_comp"      false
    :hour=    "hour="     "hour"      false
     :min=     "min="      "min"      false
     :sec=     "sec="      "sec"      false
    :_comp    "_comp"    "_comp"       true
     :zone     "zone"     "zone"       true
  Fri Jul 27 14:45:18 -0700 2007

But if I unremark the last line, they all come up true.

My guess is that their's deeper problem in Ruby's time/data libs that
needs to be fixed.

T.