Why doesn't method_missing affect respond_to?

Ruby
1

Take this class:

class Foo
def method_missing(m, *args)
m.to_s
end
end

irb(main):006:0> Foo.new.foo
=> “foo”
irb(main):007:0> Foo.respond_to? :foo
=> false

This obviously just returns the name of any method that doesn’t exist
for the object. So if you can send the object any message and get a
result, why doesn’t it respond_to those messages?

Pat

2

On Monday 05 June 2006 5:39 am, Pat M. wrote:

This obviously just returns the name of any method that doesn’t exist
for the object. So if you can send the object any message and get a
result, why doesn’t it respond_to those messages?

Think about it from an implementation perspective. How is respond_to?
to know
what method_missing is going to do?

If you define a method_missing method, and you need your object’s
respond_to?
to reflect your method_missing magic, it is up to you to write a
respond_to
method that will do so.

Kirk H.

3

Pat M. wrote:

Take this class:

class Foo
def method_missing(m, *args)
m.to_s
end
end

irb(main):006:0> Foo.new.foo
=> “foo”
irb(main):007:0> Foo.respond_to? :foo
=> false

This obviously just returns the name of any method that doesn’t exist
for the object. So if you can send the object any message and get a
result, why doesn’t it respond_to those messages?

Pat

The respond_to method can’t know what method_missing is doing with
missing methods, so it’s the up to the class author to provide a version
of respond_to that agrees with method_missing.

4

Hi –

On Mon, 5 Jun 2006, Pat M. wrote:

irb(main):007:0> Foo.respond_to? :foo
=> false

This obviously just returns the name of any method that doesn’t exist
for the object. So if you can send the object any message and get a
result, why doesn’t it respond_to those messages?

Maybe because that would render #respond_to? essentially useless :slight_smile:

David

5

On Jun 5, 2006, at 8:23 AM, Tim H. wrote:

=> “foo”
missing methods, so it’s the up to the class author to provide a
version
of respond_to that agrees with method_missing.

More to the point, if you do this:

foo.rb

class Foo
def method_missing(m, *args)
eval “def self.#{m}; ‘#{m}’; end”
self.send(m)
end
end

END

irb(main):001:0> require ‘foo’
=> true
irb(main):002:0> f = Foo.new
=> #Foo:0x1cd438
irb(main):003:0> f.respond_to? :foo
=> false
irb(main):004:0> f.foo
=> “foo”
irb(main):005:0> f.respond_to? :foo
=> true
irb(main):006:0>

You’ve now defined a ‘foo’ method and further invocations of f.foo
will go to the newly minted method rather than method_missing each time.

-Rob

6

On 6/5/06, [email protected] [email protected] wrote:

end
Maybe because that would render #respond_to? essentially useless :slight_smile:

In my example where the object responds to everything, sure. It was
just an example though, and of course not every class is going to act
like this.

I guess the key is to write a predicate and stick it in both methods.

class Foo
def method_missing(m, *args)
valid_method?(m) ? m.to_s : super
end

def respond_to?(m)
valid_method?(m)
end

private
def valid_method?(m)
true
end
end

Of course it’s not particularly useful here, but you could change
valid_method? to
def valid_method?(m)
[ :foo, :bar ].include? m
end

Guess I just thought Ruby would automatically pick it up. However, I
don’t see how it’d be able to, short of creating a copy of the object
in memory and calling the method on the object to see if it returns a
result. If the copy gives a method missing error, then respond_to?
would return false. Obviously it would suck to create a copy and
actually call the method just to see if an object responds to it.

Pat

7

Pat M. wrote:

Guess I just thought Ruby would automatically pick it up. However, I
don’t see how it’d be able to, short of creating a copy of the object
in memory and calling the method on the object to see if it returns a
result. If the copy gives a method missing error, then respond_to?
would return false. Obviously it would suck to create a copy and
actually call the method just to see if an object responds to it.

Pat

Not to mention the problems that could occur when Ruby tries calling a
method just to see if it works:

class War
def start(type=:nuclear)
…
end
end

war = War.new
war.respond_to? :start