Issue #4085 has been updated by headius (Charles Nutter).


shugo (Shugo Maeda) wrote:
> headius (Charles Nutter) wrote:
> > {:upcase => {
> >   :default => <builtin upcase>,
> >   X => <upcase patched>}
> > }
> > 
> > Method lookup would then proceed as normal in all situations. The result of lookup would be a table mapping refinements to methods with a default entry if the method is defined directly on String.
> 
> It's an interesting idea.
> How does method lookup work if multiple modules are used by using?
> 
> module Z
>   using X
>   using Y
>   "foo".upcase
> end
> 
> Are both X and Y used as a key of the table in the reverse order they are used by using?

I would think they stack like module includes, so at lookup time we'd see refined methods on String, look in calling scope in reverse order, and use the first refinement we encounter as the key.


> And, how does super work?

Well, I'm still questioning how super should work in general. Refinements are not actually modifying class hierarchy, so the current behavior of super calling the old method seems like magic to me.

You have it implemented currently as though used refinements are "virtually" in the hierarchy beneath the class, similar to prepend. So stacked refinements fire "super" in reverse order:

irb(main):001:0> module X; refine(String) { def upcase; puts 'first'; super; end }; end
=> #<Module:0x007ffbf30a4898>
irb(main):002:0> module Y; refine(String) { def upcase; puts 'second'; super; end }; end
=> #<Module:0x007ffbf3082ae0>
irb(main):003:0> using X
=> main
irb(main):004:0> using Y
=> main
irb(main):005:0> 'foo'.upcase
second
first
=> "FOO"

But there are some oddities when refinements are mixed into multiple elements of the hierarchy:

irb(main):001:0> module A; refine(Numeric) { def blah; puts 'A'; super; end }; end
=> #<Module:0x007ffeeb930d90>
irb(main):002:0> module B; refine(Integer) { def blah; puts 'B'; super; end }; end
=> #<Module:0x007ffeeb90f578>
irb(main):003:0> module C; refine(Fixnum) { def blah; puts 'C'; super; end }; end
=> #<Module:0x007ffeeb8f1870>
irb(main):004:0> using A; using B; using C
=> main
irb(main):005:0> 1.blah
C
NoMethodError: super: no superclass method `blah' for 1:Fixnum
        from (irb):3:in `blah'
        from (irb):6
        from /usr/local/bin/irb-2.0.0:12:in `<main>'

Obviously refined super is not simulating the full hierarchy here. It would be difficult to do so, but I feel like you either need to support super in refinements consistently or not at all.

If I add modules to the same locations in the hierarchy, the supers fire "properly". So basically, refined "super" is only working for one level up from the refinement itself.

irb(main):007:0> module A2; def blah; puts 'A2'; super; end; end
=> nil
irb(main):008:0> module B2; def blah; puts 'B2'; super; end; end
=> nil
irb(main):009:0> module C2; def blah; puts 'C2'; super; end; end
=> nil
irb(main):010:0> class Fixnum; prepend C2; end
=> Fixnum
irb(main):011:0> class Integer; prepend B2; end
=> Integer
irb(main):012:0> class Numeric; prepend A2; end
=> Numeric
irb(main):013:0> 1.blah
C
C2
C
B2
B
A2
A
NoMethodError: super: no superclass method `blah' for 1:Fixnum
        from (irb):1:in `blah'
        from (irb):7:in `blah'
        from (irb):2:in `blah'
        from (irb):8:in `blah'
        from (irb):3:in `blah'
        from (irb):9:in `blah'
        from (irb):3:in `blah'
        from (irb):13
        from /usr/local/bin/irb-2.0.0:12:in `<main>'

The double call of C's blah here is unexpected as well.

Another example showing that refinements don't honor refined hierarchies for "super":

irb(main):026:0> class Foo
irb(main):027:1> def blah; puts 'in Foo'; end
irb(main):028:1> end
=> nil
irb(main):029:0> class Bar < Foo
irb(main):030:1> def blah; puts 'in Bar'; super; end
irb(main):031:1> end
=> nil
irb(main):032:0> module Baz
irb(main):033:1> refine Foo do
irb(main):034:2* def blah; puts 'in Baz'; super; end
irb(main):035:2> end
irb(main):036:1> end
=> #<Module:0x007ffeeb05f978>
irb(main):037:0> using Baz
=> main
irb(main):038:0> Bar.new.blah
in Bar
in Foo
=> nil

Again, inconsistent behavior, but I'm not sure which specification is correct.

FWIW, there's something similar to refinements in the form of extension methods in C# and defender methods in Java 8. It would be worth researching how those features handle super. In Java 8, the defender methods live only on interfaces and are somewhat "virtually" in the hierarchy, so there's a lot of oddities surrounding the process of selecting the proper super method.

> > I admit I am a bit reluctant to suggest this, because I still have concerns about the feature itself. But it would be possible for call sites to only need a reference to their calling scope (determined at parse time) to implement dynamic refinements without severe impact to normal code. Dynamic refinements, as in module_eval, would work by simply invalidating the call sites they contain.
> 
> FYI, in my new implementation (http://shugo.net/tmp/refinement_fix_1119.diff), refined methods are not stored in inline cache, so there's no need to invalidate inline cache for module_eval.
> Instead, refined method invocations are slower than the implementation in the trunk HEAD.

I considered this possibility, but are you willing to accept that large parts of Rails code will have slower overall performance because they want to use refinements? I will revise my earlier refinements requirements: refined calls should exhibit exactly the same performance characteristics as regular calls. I believe if refinements go in, many many libraries will want to start using them. We should not force Ruby perf to take a major step backward just by introducing a new and potentially popular feature that has implementation problems.

> > In any case, I would really like more time for this dialog to continue. If we push refinements into Ruby in their current form, we're not giving adequate time to flesh out the edge cases. If we push a partial implementation now, we may be making a future implementation harder and we would not be protecting ourselves from mistakes. I want to work with you to find a definition and implementation of refinements that meets requirements without punishing future Rubyists.
> 
> I'm starting to think it's not good to rush to introduce Refinements as an official feature.
> What do you think of introducing Refinements as an experimental feature as Endoh-san suggested?
> I don't know what does Endoh-san mean by "an experimental feature", but it may require an explicit compile option or a runtime option (e.g., require "refinements" like continuation) to enable it.  If Refinements are enabled, warning should be shown not to make the current Refinements de-facto standard.

I do not have any objection to refinements being included as an experimental feature.

If it's a compile-time feature, I'm not sure I see the value in having it in 2.0.0 at all; people could download source and build that.

If it's a flag or require, I assume you'd have to enable it to turn on parse/compile-time flagging of refined methods/calls, correct? I think that's easy enough in JRuby too.
----------------------------------------
Feature #4085: Refinements and nested methods
https://bugs.ruby-lang.org/issues/4085#change-33968

Author: shugo (Shugo Maeda)
Status: Assigned
Priority: Normal
Assignee: matz (Yukihiro Matsumoto)
Category: core
Target version: 2.0.0


=begin
 As I said at RubyConf 2010, I'd like to propose a new features called
 "Refinements."
 
 Refinements are similar to Classboxes.  However, Refinements doesn't
 support local rebinding as mentioned later.  In this sense,
 Refinements might be more similar to selector namespaces, but I'm not
 sure because I have never seen any implementation of selector
 namespaces.
 
 In Refinements, a Ruby module is used as a namespace (or classbox) for
 class extensions.  Such class extensions are called refinements.  For
 example, the following module refines Fixnum.
 
   module MathN
     refine Fixnum do
       def /(other) quo(other) end
     end
   end
 
 Module#refine(klass) takes one argument, which is a class to be
 extended.  Module#refine also takes a block, where additional or
 overriding methods of klass can be defined.  In this example, MathN
 refines Fixnum so that 1 / 2 returns a rational number (1/2) instead
 of an integer 0.
 
 This refinement can be enabled by the method using.
 
   class Foo
     using MathN
 
     def foo
       p 1 / 2
     end
   end
 
   f = Foo.new
   f.foo #=> (1/2)
   p 1 / 2
 
 In this example, the refinement in MathN is enabled in the definition
 of Foo.  The effective scope of the refinement is the innermost class,
 module, or method where using is called; however the refinement is not
 enabled before the call of using.  If there is no such class, module,
 or method, then the effective scope is the file where using is called.
 Note that refinements are pseudo-lexically scoped.  For example,
 foo.baz prints not "FooExt#bar" but "Foo#bar" in the following code:
 
   class Foo
     def bar
       puts "Foo#bar"
     end
 
     def baz
       bar
     end
   end
 
   module FooExt
     refine Foo do
       def bar
         puts "FooExt#bar"
       end
     end
   end
 
   module Quux
     using FooExt
 
     foo = Foo.new
     foo.bar  # => FooExt#bar
     foo.baz  # => Foo#bar
   end
 
 Refinements are also enabled in reopened definitions of classes using
 refinements and definitions of their subclasses, so they are
 *pseudo*-lexically scoped.
 
   class Foo
     using MathN
   end
 
   class Foo
     # MathN is enabled in a reopened definition.
     p 1 / 2  #=> (1/2)
   end
 
   class Bar < Foo
     # MathN is enabled in a subclass definition.
     p 1 / 2  #=> (1/2)
   end
 
 If a module or class is using refinements, they are enabled in
 module_eval, class_eval, and instance_eval if the receiver is the
 class or module, or an instance of the class.
 
   module A
     using MathN
   end
   class B
     using MathN
   end
   MathN.module_eval do
     p 1 / 2  #=> (1/2)
   end
   A.module_eval do
     p 1 / 2  #=> (1/2)
   end
   B.class_eval do
     p 1 / 2  #=> (1/2)
   end
   B.new.instance_eval do
     p 1 / 2  #=> (1/2)
   end
 
 Besides refinements, I'd like to propose new behavior of nested methods.
 Currently, the scope of a nested method is not closed in the outer method.
 
   def foo
     def bar
       puts "bar"
     end
     bar
   end
   foo  #=> bar
   bar  #=> bar
 
 In Ruby, there are no functions, but only methods.  So there are no
 right places where nested methods are defined.  However, if
 refinements are introduced, a refinement enabled only in the outer
 method would be the right place.  For example, the above code is
 almost equivalent to the following code:
 
   def foo
     klass = self.class
     m = Module.new {
       refine klass do
         def bar
           puts "bar"
         end
       end
     }
     using m
     bar
   end
   foo  #=> bar
   bar  #=> NoMethodError
 
 The attached patch is based on SVN trunk r29837.
=end



-- 
http://bugs.ruby-lang.org/