From: keiju@... (Keiju ISHITSUKA)
Date: 2001-05-16T14:48:14+09:00
Subject: [ruby-talk:15241] Re: Bug or feature? eval("x=5")

Hi,

In [ruby-talk :15237 ] the message: "[ruby-talk:15237] RE: Bug or
feature? eval("x=5") ", on May/16 13:38(JST) "Joseph McDonald" writes:

>>     eval("x=5")
>>     puts x             #  Never heard of x!

>weirdness.  It works in irb but not as a script:
>
>irb(main):001:0> eval("x=5")
>5
>irb(main):002:0> puts x
>5
>nil
>irb(main):003:0>

This is a difference of execution semantics of Ruby and irb.

See the following document in detail:

## ---
= Restrictions

Because irb evaluates the inputs immediately after the imput is
syntactically completed, irb gives slight different result than
directly use ruby. Known difference is pointed out here. 


== Declaration of the local variable

The following causes an error in ruby:

  eval "foo = 0"
  foo
  --
  -:2: undefined local variable or method `foo' for #<Object:0x40283118> (NameError)
  ---
  NameError

Though, the above will successfully done by irb. 

  >> eval "foo = 0"
 => 0
 >> foo
 => 0

Ruby evaluates a code after reading entire of code and determination
of the scope of local variables. On the other hand, irb do
immediately. More precisely, irb evaluate at first

  evel "foo = 0" 

then foo is defined on this timing. It is because of this
incompatibility.

If you'd like to detect those differences, begin...end can be used:

  >> begin
  ?>   eval "foo = 0"
  >>   foo
  >> end
  NameError: undefined local variable or method `foo' for #<Object:0x4013d0f0>
  (irb):3
  (irb_local_binding):1:in `eval'

-- keiju