From: Trans <transfire@...>
Date: 2007-06-13T09:05:54+09:00
Subject: Re: Extending Symbol with #to_str effects Struct.new



On Jun 12, 6:51 pm, "Chris Carter" <cdcar...@gmail.com> wrote:
> On 6/12/07, Trans <transf...@gmail.com> wrote:
>
>
>
> > Not arguing either way a s to whether this makes sense or not (it's
> > had it's lengthy discussions before).
>
> >   > class Symbol
> >   >    def to_str
> >   >      String(self)
> >   >    end
> >   >  end
>
> > But it does cause this strangeness:
>
> >   > Struct.new(:x)
> >   NameError: identifier x needs to be constant
> >         from (irb):6:in `new'
> >         from (irb):6
> >         from :0
>
> > Is it really THAT unreasonable an extension? I wonder what else it
> > would effect, and why. I don't really even understand the error
> > message, but one gets the same result with:
>
> >   > Struct.new('x')
> >   NameError: identifier x needs to be constant
> >         from (irb):1:in `new'
> >         from (irb):1
> >         from :0
>
> > That in itself seems silly.
>
> > T.
>
> This is because if Struct gets a String or a "string like object
> (responds_to? :to_str)" it uses an alternate behavior where the first
> argument, if a proper Constant name in a string, is defined to be the
> Struct class generated under the Struct:: namespace.

Okay. Thanks. I get the error now. Though, I don't really get how this
alternate behavior is useful. Why would one want to use the Struct
namespace? And if you did, what's wrong with

  Struct::MyStruct = Struct.new( ... )

But the main thing, differentiating functionality based on String vs.
Symbol is just generally a bad idea. And very bad in core libs, IMHO.

T.