Ruby

Is respond_to?(no_method_error.name, false) not enough for this purpose?

I think he wants to distinguish self.foo and foo().
Only public methods can be candidates in the former case, but also private methods can be in the latter case.

Nakada-san, that's exactly what I want. A code example would be something like this:

begin self.do_something rescue NoMethodError => no_method_error no_method_error.private_method_callable? # => false  end begin do_something() rescue NoMethodError => no_method_error no_method_error.private_method_callable? # => true  end 

In the first begin...end part, you call self. first, so you can't call private methods while you can in the second one.

So this would essentially be a way to tell if the call was a "fcall", without needing to parse the exception message?
In other terms, telling if this NoMethodError is a "private method called" error.
Maybe exc.private_method_error? or exc.private_method_called?

I think I should've been more specific and also should've mentioned that in the example above, the method you are trying to call doesn't actually exist.

begin self.method_that_does_not_exist rescue NoMethodError => e e.message # => undefined local variable or method `method_that_does_not_exist' ...  e.private_method_callable? # => false  end begin method_that_does_not_exist() rescue NoMethodError => e e.message # => undefined local variable or method `method_that_does_not_exist' ...  e.private_method_callable? # => true  end 

So, I guess what I'm proposing here is adding a way to tell if a "fcall" could be made.

The first error message is actually
"undefined method `method_that_does_not_exist' for main:Object".
But that does not help since anyway the error message below would be the same since it has "()" (and the same problem if it would have arguments)

"private_method_callable?" sounds like a specific private method would be callable.

Maybe e.private_call?
"Returns true if the call which generated the error was allowed to call private methods (because it had no receiver, was using self.assign= or used #send)"

Does it satisfy your use case that raising another exception when private method call with self. ?

begin self.a_private_method rescue NoMethodError => e e.is_a? PrivateMethodCallError #=> true end begin self.method_not_exist rescue NoMethodError => e e.is_a? PrivateMethodCallError #=> false end begin method_not_exist rescue NoMethodError => e e.is_a? PrivateMethodCallError #=> false end 

Of course, PrivateMethodCallError inherits from NoMethodError.

Benoit, you are absolutely right about the error message. I was a bad person and didn't really check after copying & pasting...
Regarding the method name, #private_call? sounds better to me than private_method_callable?.

Nakamura-san, sorry I wasn't clear enough in the first place. It would be great if you could read the further explanation I posted above.

https://github.com/ruby/ruby/compare/trunk...nobu:feature/12043-NoMethodError%23private_call-p

Good suggestion IMHO, +1

The did_you_mean gem is great. If distinguishing between "()" and no "" will
make things even greater then I am all for it. \o/

This feature is only for "did_you_mean" gem, so I think that any name is OK, including implementation details.

private_call?
explicit_receiver?
fcall?
were_you_a_function?
demon_from_the_nose?
etc.

Nobuyoshi Nakada wrote:

This feature is only for "did_you_mean" gem, so I think that any name is OK, including implementation details.

private_call?
explicit_receiver?
fcall?
were_you_a_function?
demon_from_the_nose?
etc.

It's still public API.

Let's choose private_call?, since I and the OP agree on this, unless somebody has an objection.