Re: Difference between class declarations

On 9/24/05, Mathieu Bouchard <[e-mail]> wrote:
> On Thu, 22 Sep 2005, TRANS wrote:
> > On 9/22/05, nobuyoshi nakada <[e-mail]> wrote:
> > A testcase needs to keep it's contents separate from other testcases,
> > just as it would if I were explicitly suclassing Test::Unit::TestCase.
> > (Note too the remark, how do I set the constant name at the toplevel?)
>
> I've also encountered a similar problem trying to use Class.new{}. It
> seems that the handling of constants in blocks changed just after
> Class.new{}, rendering the concept of Class.new{} almost useless. The
> point of RCR-41 was to make the "class A<B" notation superfluous, but who
> wants that? (except myself)

Peter and I have discusses how one might create blocks the close off
the scope, after he suggested a nice way to open them up in other
cases like this:

  x = "q"
  class A {
    define_method(x) {1}
  }
  A.new.q  #=> 1

But we couldn't think of a good way to do the opposite and close
blocks. The best idea (which I'm sure has been mentioned before) was
to have do..end be different from {...}.  But this break old programs
and also is problematic with the binding precedence difference between
the two. Also it would make the use do vs { very explict rather then
just a matter of personal choice. Personally I think it's a still fair
idea despite these issues, but I've come up with another idea:

  a = 1
  foo [ |x| a = 10 ]
  a #=> 1

Since { } is also used for Hash, I don't see why [ ] can't also be
used. Of course the above really cries out for a way to pass in
specific vars into the closed block. But it would allow:

  Class.new [
    # it's own scope
  ]

T.