[chowells]

Flexible in a way, sure. But non-locality is generally a bad property, not good. Adding it is a workaround for all the enumeration methods using blocks in a way that makes people think they're weird looping syntax instead of a fundamentally different idea.

People want to do early returns from looping over a collection, so take the easy solution of adding more messy language semantics instead of finding a semantically simple solution instead. (For that matter, have fun working out the semantics of break and next when used in a block that isn't an argument to an enumeration method. How do you as a method author opt in to distinguishing between the two after yielding to a block?)

This is generally the case with Ruby everywhere. Why does that thing have an edge case with weird semantics? To work around the edge case with weird semantics somewhere else. It's all fine if you want to just try things until something works. But if you want to really understand what's going on and write code that's a first-class participant in the language features, it gets really frustrating to try to deal with it all.

[vidarh]

It's really not that complex.

For my (buggy, unfinished, languishing without updates) prototype Ruby compiler, lambda and proc (and blocks) are all implemented nearly the same way, with the exception that for proc/blocks, return/break/next will act as if having unwound the stack to the scope where the proc was defined first (or throw an error if escaped from there).

The distinction is very obvious when you think of it in that way - a proc acts as if called in the context/scope it was defined, while a lambda acts as if called in the scope it is called in.

> How do you as a method author opt in to distinguishing between the two after yielding to a block?

You don't. A block acts as a proc, not a lambda. If you want lambda semantics, then take a lambda as an argument - don't take a block/proc and be surprised it doesn't act like something it isn't.

[chowells]

Given that you completely ignored what I said I wanted to do and gave an answer for some other question, I'm pretty sure it's more complicated than you think.

I want to write a method that takes a block and distinguishes between next and break, exactly like the methods of enumeration do. It's obviously possible because a super common interface does it.

Last time I looked, that interface does it by being written in native code that interfaces with the interpreter. That is, it's not part of the language semantics. It's a special case with weird rules unlike what anything else gets to do.

Or at least it was. Maybe the language has actually made it accessible since then, but I'm not optimistic. That's not the ruby way.

[vidarh]

Your sentence structure did not at all make it clear that "the two" referred to next and break rather than proc and lambda.

Assuming you don't care whether a "next" was actually called, but only whether you've exited the block and whether or not you exited it via a break, you can do this check in a number of ways, but it I will agree it's a bit dirty that sometimes you do need to rely on standard library functionality rather than language constructs if you want to do these things.

Here's one way of doing it, since "break" within a Fiber triggers a LocalJumpError:

    def next_or_break(&block)
      Fiber.new(&block).resume
      :next_or_end_of_block_reached
    rescue LocalJumpError
      :break
    end

    p(next_or_break do next end)
    p(next_or_break do break end)

[dragonwriter]

> but it I will agree it's a bit dirty that sometimes you do need to rely on standard library functionality rather than language constructs if you want to do these things

You don't need to introduce a Fiber or rely on the standard library, you can just use the core language, leveraging the fact that code in ensure sections gets called even when you are returning "past" the calling method.

(You may need a dirty catch-all rescue so you can set a flag before reraising that lets you distinguish "bypassing direct return by exception" from "bypassing direct return by break or proc-semantics return", but that's still the core language, not standard library.)

I put example code in another comment: https://news.ycombinator.com/item?id=44065226

[vidarh]

Ah, of course. I'm embarassed I didn't think of that now... That's absolutely better than abusing Fiber. Great suggestion.

[kayodelycaon]

I think you’re seeing the base language.

Effectively blocks are self-contained chunks of code. You can change things around them, but you can’t change how keywords work inside them. Because you’re crossing a method boundary when you call a block you’re not able to access next and break. (Or capture return.)

Ruby is defining scope here and C methods are not limited by the language they define.

[vidarh]

Exactly. You can leverage some of the standard library methods to figure out if a block did a next or exit normally vs. did a break, by e.g. abusing Fiber's (there are probably other ways too), and that might feel a bit dirty, but those are also part of the language.

I think it'd be nice if Ruby language constructs offered a few of the building blocks used by the standard library so that you could implement more of the standard library without dipping down into some low level mechanism - for my prototype Ruby compiler a priority was to implement as much as possible in Ruby, and there are certainly more things in the core classes that can't be done in pure Ruby than I'd like.

But it's not that much, and while there are plenty of warts, the inconsistencies are often smaller than people think.

[drnewman]

It seems you're pretty upset about your experience with Ruby. I'm sorry that's been the case for you.

However, in Ruby blocks aren't just about flexibility, more importantly they're about generality. They're not there to resolve an edge case at all (Ruby also has keywords for loops). They're a generalization of control flow that is just slightly less general than continuations. In practical use they provide an alterative to Lisp macros for many use cases.

These were some of the problems Matz was trying to sort out with his design of Ruby--creating a language that was fun and easy to use for day-to-day programming, with as much of the meta-programming power of Lisp and Smalltalk as possible. Blocks are one of his true innovatations that came from trying to balance that tension.

[igouy]

> Blocks are one of his true innovatations …

How do they differ from Smallalk blocks? (I don't know.)

https://drcuis.github.io/TheCuisBook/Block-syntax.html

[drnewman]

Blocks in Smalltalk (to my understanding) are closures. Blocks in Ruby are closures that also bring the call stack they were created in with them.

One way to think of about it is this: anonymous functions as originally implemented in early Lisps are code as an object, closures are code with its lexical environment as an object. You can think of a Ruby block as code with its lexical environment and its call stack as an object.

So they don't just handle return differently than closures, they have access to the call stack of the site where they're created like a continuation. This is why they handle return differently, but this is just one of the things that falls out from that. It also comes with other control flow features like "redo", "retry", "next", "rescue", "finally", and others. These are all call stack control (control flow) conveniences, just like return is. All of them can be thought of as being abstractions built on top of continuations (just ask a Scheme hacker).

Originally Ruby was basically a Lisp without macros, but with continuations, a Smalltalk like object system and a lot of syntactic affordances inspired by Perl, and other languages. Blocks are one of the conveniences built on top of the Lispy semantics.

Note that I'm explaining how blocks work as an abstraction (vidarh below explains how they work as a concretion, as implemented in MRI).

[igouy]

> … other control flow features like "redo", "retry", "next", "rescue", "finally", and others.

At-a-glance afaict Smalltalk provides those features too, so I would guess Smalltalk blocks may have access to the call stack too?

https://drcuis.github.io/TheCuisBook/The-Debugger.html

[drnewman]

Yes Smalltalk has continuations. So it can do all of those things as well. But I don't think they're explicitly tied to blocks like they are in Ruby. This really isn't a problem for Smalltalk since it's not as syntax oriented as Ruby.

The invovation is to have those features tied to convenient syntax.

[igouy]

So is the innovation to make something that was available in Lisps / Smalltalks, available within the different constraints of Ruby.

(I should check how Smalltalk blocks behave.)

[kazinator]

I don't believe this is correct. Blocks are closures.

Continuations were introduced in Ruby 1.8, via a callcc method in a Kernel module or something like that.

Since 1.9 they are in some sort of deprecated status.

I don't think that even the yield() stuff is implemented using continuations.

[drnewman]

I agree blocks are closures. But they’re call stack aware closures. Which gives them _some_ of the power of continuations.

I’m also sure you’re right that they’re not implemented using continuations. However, my understanding is that Ruby was originally conceived as a language with continuations. I’ll see if I can find a reference for that. But from what I recall reading in a blog post from someone who was at a programming language conference in 1997 when Matz introduced the language that’s how he described it.

[kazinator]

> they’re call stack aware closures

What does that even mean?

If you create a block deeply within a cascade of nested function calls, nineteen activation levels deep, and return that block out of all those nestings, is it still aware of the nineteen levels that have terminated, and to what purpose/benefit?

What example Ruby code would break without continuing access to the dynamic scope that has terminated, rather than just the lexical scope?

[chowells]

No, return exiting the enclosing function scope from a block is the special case that's there to resolve an edge case.

Blocks should be nothing special. They're anonymous functions that capture the environment mutably. The only new part is all the special bits added to handle the weird edge cases that they're trying to pretend don't exist.

[HellzStormer]

I disagree that it's just there to handle edge cases. It's a useful generalization.

I think the "Building an intuition" section of my blog post[1] makes a good case for that.

  When dealing with loops, you have 3 nested constructs interacting: a wrapping function, a loop statement and the loop's body; and you have 3 keywords to choose where the flow of the code goes.

    return returns from the wrapping function
    break leaves the loop statement
    next / continue leaves loop's body

  When dealing with blocks or anonymous functions, it's instead 3 nested "functions" that are interacting: a wrapping function, a called function and an anonymous functions (or block).

    Ruby's blocks, let you use the same 3 keywords to choose where the flow of the code goes.

    return returns from the wrapping function (ex: my_func)
    break returns from the called function (ex: each, map)
    next returns from the block

  Quite consistent. But since we are talking about functions instead of statements (loop), return values are also involved. Allowing both break and next to provide a return value fits well in that model and is quite useful. The 3 keywords are basically return, but they have different targets.

[1] https://maxlap.dev/blog/2022/02/10/what-makes-ruby-blocks-gr...

[dragonwriter]

> No, return exiting the enclosing function scope from a block is the special case that's there to resolve an edge case.

What is the edge case? It seems to be there so that Ruby enumeration methods can provide the behavior expected of looping statements (which is kind of necessary if you want the looping statements to just be semantic sugar for enumeration methods so that they can work correctly with any enumerable.)

> Blocks should be nothing special.

"Special" compared to what?

[dragonwriter]

> have fun working out the semantics of break and next when used in a block that isn't an argument to an enumeration method.

The semantics are:

* next returns from the block, optionally taking a return value that will be returned as the result of the function (this is identical to the behavior of "return" from a Proc with lambda semantics)

* break returns from the function that yielded to the block, optionally taking a return value that will be returned as the result of the function (EDIT: deleted comparison to returns from a Proc with proc rather than lambda semantics here, because it wasn't quite accurate.)

This is, incidentally, also exactly the semantics when they are called from a block passed to an enumeration method, there is no special case there.

> How do you as a method author opt in to distinguishing between the two after yielding to a block?

If control returns to your method, it was a return from the block, either via return (from a block with lambda semantics), next, or running through the end of the block. If, OTOH, break is called (or return from a block with proc semantics), control won't return to the calling method normally, but code in any "ensure" sections applicable will be run, which can be used to identify that this has occurred, and even to override the return value provided by the break.

The simplest possible function illustrating this:

  def how_exited
    yield
    direct = true
    return "next, completion, or lambda return"
  rescue
    direct = true
    return "exception"
  ensure
    return "break" if not direct
  end

[pmahoney]

Can you explain further? I'm not sure I follow.

> How do you as a method author opt in to distinguishing between [break and next] after yielding to a block?

I don't use Ruby much lately, but if I yield to a block which calls break, control will pass to the code following the block (and not back to my method). If the block calls next or simply finishes, control passes back to me though I cannot know if next was called or not (but do I care? I can go ahead and yield the next element of a collection either way)

[chowells]

Ah, you're right. They do have observably different behavior. But I can't even close an open file handle after a break. That's still pretty miserable.

[dragonwriter]

> But I can't even close an open file handle after a break.

Yes, you can. That's what ensure (the Ruby equivalent of "finally") is for. Or, better using File.open with a block where you do the work, which uses ensure under the hood.

[HellzStormer]

Yes you can, you use a finally clause around your call to yield, as you would already do to handle possible exceptions happening in the block. So no extra consideration is needed.