[_ZeD_]

Just as note: I don't really would call "len(o)" a "legacy call", as in "something it's here just for backward compatibility, but it's ugly and please don't use it".

It's more "the good way to do it" :D

len() is a perfect example of the duck typing and the "protocol-based" philosophy of python: its implementation it's something like

    def len(object):
        return object.__len__()

as in "just return the result of invoking the object method called __len__".

this lets you just call len() on every "len-able" object. They can be list, dict, custom objects... instead of the need to, I don't know, implements an explicit interface, or define a "getLength()", "size()", "length()", "length" (just an attribute, not a function to call), you can just define a "magic" (as in "normally you don't need to call this directly") method called __len__.

you can create your "not really a container, but it has a length!" object as something like

    >>> class C:
    ...     def __len__(self):
    ...         return 42
    ...
    >>> len(C())
    42

As indirect effect, it enables you to go on the functional-style approach... a simple

   map(len, list_of_lists)

it's more readable than

   map(lambda l: l.__len__(), list_of_lists)

(probably [len(l) for l in list_of_lists] it's more readable, but bear with me...)

I'm not really sure about the "ruby C abstractions": do you refer to the use of "object-like" structs used in the C sources of the "ruby MRI" implementation of "ruby-the-language"?

regarding the GC... I don't know, but why is the kind of the GC used by an implementation of the language (or the complete lack of a GC) relevant to the "everything is an object"?

just to be clear, I don't dislike ruby :) it's just that I don't find the "everything is an object" a way to discriminate between ruby and python.

[estavaro]

You see, you wrote a bunch about how len is so cool in Python. But in the face of dynamic typing and polymorphism I didn't expect it to be any different really. Languages like Go that are statically typed make more of an issue about interfaces. Even in Dart they dropped explicit Interface usage in favor of implicit interfaces, considering that Dart is more explicit about matters than Ruby is.

Switching those function calls around reminds me of Delphi. When I first learned about Python, those calls reminded me of Delphi which I had used more. But once I learned Java, my frame of mind went from function(object) to object.function. With Ruby I found it quite intuitive. Now all languages work better if they approach things the way Java and Ruby do because of familiarity. That's why I like Dart for what it's worth.

Regarding C extensions of Ruby, they are like OO in C. Not sure how Python does it by default. But in Ruby all C extensions have a OO flavor. The first major book about Ruby writes about them: http://www.ruby-doc.org/docs/ProgrammingRuby/html/ext_ruby.h...

I think if you're counting references in the GC perhaps your C extensions are not very OO yet. In Ruby as in Java, the GC is an abstraction over reference counting. I think reference counting is often said to lead to more predictable performance at the cost of increasing the maintenance burden.

So again, not sure where Python is not "Objects all the way down." But Ruby has had the OO philosophy from early on. I think the Ruby OO approach started with the OO of the C extensions and went from there. Unlike Python that had a stronger procedural influence. The gap has closed since, but in Python a class is not written like this yet:

  class InRuby
  def aMethod
  end
  end

  class InDart {
  aMethod() {}
  }

[_ZeD_]

I don't get the last part...

in python a class is written as

    class InPython:
       def aMethod(self):
           pass

but... what I'm missing? I can't follow the reasoning...

regarding the C extensions... I didn't really needed to write one, but from http://docs.python.org/3/extending/extending.html#a-simple-e... I see

    static PyObject * 
    spam_system(PyObject *self, PyObject *args)
    ...

so I assume also on C level the python objects are mapped on some structs in OO fashion...

I honestly have no idea on how is (or if it has some meaning) to extend python in rpython on pypy, but in this case I assume there are some objects involved :D

[estavaro]

Exactly. In Python you still pass "self" to the method. In Ruby and Dart it's not needed.

In Python I've seen Python users discourage the use of classes. I haven't seen the same distrust of classes in Ruby for instance. It's as though Python has unresolved OO issues. Python could take pride in being "multi-paradigm", whereas languages like Ruby and Dart could take pride in being more OO.

Regarding the C extensions of Ruby, I think the bottomline is that from int, to booleans, to null, to classes, everything is considered OO even in the C code. So common functions can be applied to them even from C, say like Polymorphism would in other languages. From the primitives on up everything is like a high level Ruby. As Python is more "multi-paradigm" perhaps some of its C extension features aren't as OO as Ruby's are.

Funny though that those C extensions that make Python and Ruby so popular end up making other implementations of those languages that target other runtimes incompatible as they don't have access to those C extensions. So even in that regard they are similar.

Say we can't find anywhere in Python where Objects don't exist. Then welcome on board of true OO. Now show OO some love and stop discouraging the using of classes. Maybe drop the "multi-paradigm" approach. I know though that like JavaScript, sometimes you have to live with the shortcomings of the programming language in its support of OO. So while JavaScript could be said to have Objects everywhere, and it truly does, writing classes in JavaScript is not a settled issue. That's why I like Dart instead:

  class A {
    var aList = ['a', 'aa'];
    get length => aList.length; 
  }
  
  class B {
    var aList = ['b', 'bb', 'bbb'];
    get length => aList.length; 
  }
  
  class C extends A {
    var aList = ['c', 'cc', 'ccc', 'cccc'];
  }
  
  printIt(o) {
    print("${o.aList}: ${o.length}");
  }
  
  void main() {
    printIt(new A());
    printIt(new B());
    printIt(new C());
  }

Result of running it:

  [a, aa]: 2
  [b, bb, bbb]: 3
  [c, cc, ccc, cccc]: 4

[Ingaz]

>>Say we can't find anywhere in Python where Objects don't exist.

You're correct - we can't find a place in Python where Objects don't exist. Python and Ruby are exactly same in this.

>>Then welcome on board of true OO. Now show OO some love and stop discouraging the using of classes.

What do you mean by true OO?

It does not matter if you "love" objects or not. You can't write in python without OO. Every python module is an object, defs and vars - members of object. (Even classes and other modules are members of module object).

[philwelch]

Or, instead of having the convention that objects with a length implement __len__, you could have the convention where they just implement len and you could just call obj.len.

[_ZeD_]

In java the arrays objects expose a public int attribute called .length. The strings also expose a .length attribute, but it's a method. If, for some reason, you need to know how many elements are in enum, you need to call NameOfTheEnum.values().length. If you have a Collection (a Map, a List...), you need to call .size(). ...

Yeah. You are right about the fact that a possible "convention" would be to call .len, not .__len__ but you are forgetting 20+ years of backward compatibility.

I don't know when len() was introduced, and I'm not sure it's initial implementation was a simple "return object.__len__()" (or an equivalent...). What I know is that an explicit len function helped to hide the eventually different implementation details, and allowed the developers of the containers go crazy with attributes.

Moreover, with a len(obj) function and a obj.len() "public" method, you have two ways to do it.

[philwelch]

I'm not sure why Python has to be backwards compatible with Java, given that Python is a totally different language released four years before Java.