Is there some way to execute a block within an arbitrary lexical scope?

Ruby
#1

Is there some way to execute a block within a certain lexical scope?

Let’s say the scope I want is inside B:
module A
class B
# … wanna get in here
end
end

And I wanted something like:

method_returning_B_in_lexical_scope_of_A.class_eval do
…
end

As most of you know ‘A::B.class_eval’ does not cut it, since A would not
be in the nesting.

The obvious reason I want this is to take advantage of the nesting for
lookups.

Thanks for any ideas!

– Ara V.

#2

On 01.05.2008 21:31, Ruby T. wrote:

method_returning_B_in_lexical_scope_of_A.class_eval do
…
end

As most of you know ‘A::B.class_eval’ does not cut it, since A would not
be in the nesting.

The obvious reason I want this is to take advantage of the nesting for
lookups.

Why do you need it? There is nested lookup already:

[email protected] ~
$ ruby -e ‘module A;X=1;class B;def x;X;end;end;end;p A::B.new.x’
1

Cheers

robert

#3

Let’s say I’m doing some metaprogramming from another class and I want
to do something in B’s full lexical class scope.

#4

On May 1, 2008, at 3:31 PM, Ruby T. wrote:

Let’s say I’m doing some metaprogramming from another class and I
want to do something in B’s full lexical class scope.

then you will need to use continuations.

a @ http://codeforpeople.com/

#5

Robert,

Well, let’s say that that something is a new class I want to define
within that scope. And I’d like the innards of that class to be able to
take advantage of the the natural, idiomatic lookup rules, where the
interpreter traipses up the ancestor chain.

To be more clear, let’s say I have

X = :bad
module A
X = :good
class B
#…let us say I want to execute something here where X == :good
end
end

And I’d like to execute something from outside that lexical scope (let’s
say from my metaprogramming classes) within that [A, A::B] lexical
scope.

I guess what I’m asking is: is there any way of faking lexical scope
(for the purposes of metaprogramming)?

#6

Ara,

Thanks for your suggestion. I’m a little hazy on how this would help. If
I defined a proc or code block or whatever elsewhere, it’s bound to the
lexical scope where it was defined. How does calling it from the
continuation help given that?

Anyway, maybe that’s not what you were thinking, and it’s likely I’m
missing the obvious here. I’m not the best at problem solving with
continuations.

– Ara V.

#7

2008/5/2 ara.t.howard [email protected]:

On May 1, 2008, at 3:31 PM, Ruby T. wrote:

Let’s say I’m doing some metaprogramming from another class and I want to
do something in B’s full lexical class scope.

What exactly is this “something”?

then you will need to use continuations.

It depends of course what should be achieved. A simple constant
lookup is easily done via #const_get. Other things can be achieved
via #instance_eval.

Cheers

robert

#8

On May 2, 2008, at 7:54 PM, Sean O’Halpin wrote:

Ara - could you shed some light on how continuations would help here?
I don’t see it myself.

ingore me, i’m insane. i was thinking of evil.rb.

a @ http://codeforpeople.com/

#9

On Thu, May 1, 2008 at 8:31 PM, Ruby T. [email protected]
wrote:

Thanks for any ideas!

– Ara V.

Hi,

there doesn’t appear to be any form of block invocation that changes
how constants are looked up.

Some experiments:

X = :bad
module A
X = :good
class B
#…let us say I want to execute something here where X == :good
end
end

as you state, this doesn’t work

A.module_eval { X } # => :bad

nor this

A.instance_eval { X } # => :bad

nor this

A::B.new.instance_eval { X } # => :bad

this does of course

module A
X # => :good
end

and this

module A
class B
X # => :good
end
end

…but this surprised me

class A::B
X # => :bad
end

this doesn’t work either

context_A = A.module_eval { binding }
eval(“X”, context_A) # => :bad

you have to use a string

context_A = A.module_eval “binding”
eval(“X”, context_A) # => :good

but again

context_B = A::B.module_eval “binding”
eval(“X”, context_B) # => :bad

so you can’t parameterize this which means you have to do something

like:

def eval_in_namespace(namespace, str)
constants = []
namespace.split(/::/).reject{|x|
x.empty?}.inject(Module.const_get(self.class.to_s)) { |prev, this|
(constants << prev.const_get(this)).last
}
prefix = constants.map{ |x| “#{x.class.to_s.downcase} #{x}”}.join(’;’)
suffix = constants.map{ ‘end’}.join(’;’)
eval “#{prefix}; #{str}; #{suffix}”
end
eval_in_namespace(“A::B”, “X”) # => :good

not very pretty :S

of course, you could just use:

A::X # => :good

but not

A::b::X # => :bad # !> toplevel constant X referenced by A::b::X

Ara - could you shed some light on how continuations would help here?
I don’t see it myself.

Regards,
Sean

#10

Please do not top post.

On 02.05.2008 19:26, Ruby T. wrote:

class B
#…let us say I want to execute something here where X == :good
end
end

And I’d like to execute something from outside that lexical scope (let’s
say from my metaprogramming classes) within that [A, A::B] lexical scope.

I guess what I’m asking is: is there any way of faking lexical scope
(for the purposes of metaprogramming)?

For constant lookup there is:

[email protected] /cygdrive/c/Temp
$ ruby lex.rb
2

[email protected] /cygdrive/c/Temp
$ ruby lex.rb
2
2

[email protected] /cygdrive/c/Temp
$ cat lex.rb

class Module
def lex_lookup©
name.split(’::’).inject([TOPLEVEL_BINDING,[]]) do |(b,o),n|
b = (b.const_get n rescue eval(n,b))
[b, o << b]
end.last.reverse.each do |cl|
return cl.const_get© if cl.const_defined? c
end
raise NameError, c
end
end

X=1
class Foo
X=2
class Bar
puts lex_lookup(“X”)
end
puts lex_lookup(“X”)
end

[email protected] /cygdrive/c/Temp
$

You can easily extend that to include top level constants as well as
other evaluations.

Cheers

robert