Hacker News Reader

Comment by torben-friis | Hacker News Reader

Maybe it is clearer if I explain it in syntactic terms? In my mental model objects are nouns (described entities) and methods are verbs - actions over the noun.

process.start() is the action of starting done by the the noun that is the process.

It's not exactly a matter of naming, as some methods are not explicitly verbs, but there is almost always an implicit action there: word.to_string() clearly has the convert/transform implication, even if ommitted for brevity.

I see no path where 5 is a noun and times the verb, nor any verb I can put there that makes it make sense. If you try to stick a verb (iterate?) it becomes clear that 5 is not the noun, the thing performing the iteration, but a complement - X iterates (5 times). Perhaps the block itself having a times object with 5 as an input would make it more standard to me (?).

But I do understand that if something is extremely practical a purist/conceptual argument doesn't go very far.

View on HN

It’s not just about practicality. Ruby is using message passing, not method calling. This is fundamentally different and a bit foreign to the larger community. Then ruby layers syntactic sugar on top that hides this.

Behind the scenes everything is a message passed using __send__ and you can do this directly as well, but you generally don’t.

So when you write

5.times { puts "Hello" }

It’s sort of expected by the average programmer that you are telling 5 to call the times method and expect it to exist and do what it’s told.

In reality you have indirectly sent a message that looks like

5.__send__(:times) { puts "Hello" }

What we are really doing is sending a message to 5 (the receiver) and giving it the opportunity to decide how to respond. This is where method_missing comes in to allow responding in a custom fashion regardless if a method was explicitly defined.

So you’re not telling 5 to call the method times, rather you are asking, “Hey 5, do you know how to handle the message times?”

These are fundamentally different things. This is actually super important and honestly hard to really grok _especially_ in ruby because of the syntactic sugar. I came from a C/C++ background originally, then Java and then moved to Ruby. After a few years I thought I understood this difference, but honestly it wasn’t until I spent a couple years using Objective-C where message passing is happening much more explicitly that I was able to truly understand the difference in a way that it became intuitive.

I’m a rubyist, but how is message passing fundamentally different from method calling? I get that method_missing adds a twist, but your comment doesn’t explain what the fundamental difference is.

Especially in the context of Fixnum#times. How does message passing vs method calling matter there? #times is just a method on Fixnum.

Because it leaves it up to the object being called what to actually do with the message. The object you're talking to might be forwarding its messages to another object in another ruby instance on another machine (if the current machine is getting fully loaded, etc), and the caller would be none the wiser. And crucially, the caller wouldn't have to be modified to enable this. The logic for this would be entirely within the object being called.

So the difference isn't just with method_missing.

With "method calling" as you put it, the program blows up if that object doesn't have that method, WHEN YOU CALL IT.

Basically, this smalltalk oo paradigm is about shifting where you put more of your logic. At & around the "call" site, or within the object whom you're calling & entrusting to do something useful with it.

All hearkening back to Alan Kay's original ideas about biology influencing how we organise code, and having a program be 1000's of "little black boxes" where work gets done by all these boxes talking to each other.

Which is why smalltalk (& ruby implements the Smalltalk object model to its entirety) actually has an awful lot in common with Erlang & other BEAM runtimes, even though those are functional languages. Because once you get beyond the techy buzzwords, the main ethos behind them is actually quite similar.

I guess what I’m getting at, is that I don’t understand how the difference actually informs anything concretely, as in the example of Fixnum#times, where this discussion started. Why is it super important to understand this fundamental difference?

Fixnum#times isn’t a great example, I only used it since the parent used it to illustrate their confusion and quite frankly a concrete useful example is to complex for this format.

ActiveRecord has changed a lot over the years, but as an example in the original ActiveRecord you used dynamic finders. None of the finder methods existed initially, but if you passed a message to an active record object for a non existent method rather than fail it would determine if that should be a method and then it would build and persist a method to the process for future calls.

It allows for some really interesting and powerful applications in horizontally scaling as well.

> I’m a rubyist, but how is message passing fundamentally different from method calling?

The difference is fairly subtle most of the time in practice, which is why dynamic OO languages like Ruby that use message passing can mostly look like OO languages that use method calling, though it is significant behind the scenes and opens up a lot of possibilities; one of the more obvious practical differences is that it is why Ruby can capture and handle unknown "method calls" via "method_missing", which works the way it does because the version of __send__ defined in BasicObject looks for an existing method matching its first argument in the object's inheritance chain, and if it finds that calls it, and if it doesn't find one does the same thing with the "method_missing" method.

There's a lot that can be built on top of this, either leveraging method_missing or more directly.

Though I think the more relevant issue upthread isn't message passing vs method calling but narrow vs broad conception of nouns/agents.

In a pervasively object oriented system/language, anything that can be represented in the system is an object (=noun =potential agent of action).

Ruby is almost a pervasive OO system (falling short because blocks are not objects, though they can be reified into objects with some efficiency overhead.) As such, unreified blocks are the only thing in Ruby that cannot be the agent of action.

For an iteration of a block a fixed number of times, the potential agents are the block and the number of times. But blocks aren't objects, so they can't be the agent, leaving only the number.

Hence, Integer#times

I have been doing Ruby for so long that it feels very natural to apply a method in this way on the instance.

false.not

applies the not method on the false instance in the same way that

car.start

in every OO language calls the start method on car as the receiver.

So filter(list) feels just wrong when you are clearly filtering the list itself.

Although I prefer Elixir currently I agree that ruby at least goes all the way in on OO and not having to remember which feature is implemented as a language syntax and what is just a method invocation is a strength not a weakness. It is different in other languages for historical performance reasons really.

list.filter is ok! Filtering is an action that applies to a list

false.not is borderline but if read as false.negate it makes sense (negating is an action that applies to a Boolean value). That wording screws the chaining though.

5.times is where the pattern breaks: times is not an action that applies to a number (nor an action at all). It’s the block the one that should repeat/iterate - but Ruby breaks the rule there and blocks are not an object (!). If they were you could block.repeat(5) which IMO is cleaner.

There is a bit of personal preference in what "applies to a number", but I see what you mean.

As a slight correction, a block is indeed an object! They are received by methods as an instance of the Proc class:

  def inspects_block(&block)
    puts block
    puts block.class
  end
  inspects_block { "foo" }
  # => #<Proc:0x0000000000000000>
  # => Proc

You can even add a 'repeat' method to these in the way that you specified, although you will need to add '->' to declare the block (as a lambda, which is also just an instance of Proc) before you call #repeat on it:

  class Proc
    def repeat(n)
      n.times { self.call }
    end
  end
  ->{ puts("foo") }.repeat(3)
  # => foo
  # => foo
  # => foo

Blocks are actually instances of the Proc class. There are helper methods to handle blocks passed to methods in a lightweight manner but you can also accept the block as a method argument, e.g.,

  class Integer
    def times(&blk)
      i = 0
      while i < self
        blk.call(i)
        i += 1
      end
    end
  end

> Blocks are actually instances of the Proc class.

No, they aren't.

Blocks can be reified into instances of the Proc class, but they are not objects and reifying them into objects has overhead.

Using a & argument asks for a block passed to the method to be reified into a proc accessible in the body of the method, which is useful if you are going to do something with it that you can't do with a bare block.

Fun fact - there's no Boolean class in Ruby. True is an instance of TrueClass and false is an instance of FalseClass

Hmm, but that's not really a "breakage" in ruby, it's more an aesthetic argument over which objects should have which logic. It's like naming things. Smalltalk (& ruby) has 'select'. Everyone else uses 'filter'.

To some, 5.times seems very readable & logical. It's like arguing over the "right" colour scheme to use while coding (BTW, the correct answer is solarised light, but with black foreground text!!)

I think I feel you. However, I think you have conceptually loaded "method" with more meaning. I think a less loaded way to think of the object/method is that the object is the first argument to the method that is called on it. So

    5.times { puts 'hi' }

is equivalent to

    times(5) { puts 'hi' }

which you could expand to

    my_function = -> { puts 'hi' } 
    repeat_times(5, &my_function)

And here is another reason for the disconnect: in a purely functional language repeating a function five times is useless, you're doing something for side effects only. Looping in itself is kind of a wrong (i.e. incomplete) abstraction for functional dev, because you're usually thinking in higher level concepts, such as `reduce`-based transformations. Maybe that's another part of the reason why `5.times { … }` feels off.

After my foray into functional programming, I actually ended up appreciating Ruby more, because it lets you have it both ways: program your computer directly, and harness functional concepts. Since computer hardware is not functional I don't want the extra ceremony and abstraction over it for the sake of purity.

All that said, going back and forth between Ruby and Elixir really conceptually crystallized for me that the method call receiver is basically just the first argument to the method, accessible with the keyword `self` (which in Python is made explicit for example).

Reply on news.ycombinator.com