Continuations by example: Exceptions, time-traveling search, threads, and more | Hacker News Reader

Continuations by example: Exceptions, time-traveling search, threads, and more

matt.might.net

151 pointsalschwalm15 comments

I always start Matt Might articles thinking "Okay, THIS is the one I'm going to understand all the way to the very end without unlearning what I just read, else I die," and alas, I have again died.

Me to. Level failed, but then i found this cheat code from apache and at lest i now understand basics.

http://commons.apache.org/sandbox/commons-javaflow/

AFAIK Javaflow's been inactive for over a decade. Newer implementations for the same concept include https://github.com/offbynull/coroutines and https://github.com/vsilaev/tascalate-javaflow/.

Full disclosure: I'm the one who built / owns the first project.

Love it. Lmk if you never need a hand. I was on the core analysis team at Fortify.

Thanks for sharing this background and cool project.

On the topic of inactive continuation implementations for Java: I implemented native support for both delimited (shift/reset) and non-delimited (call/cc) continuations in the Avian JVM. See https://github.com/ReadyTalk/avian/blob/master/test/extra/Co... for some mind-bending examples.

I'm having an ADHD night and I had the same thought, except I haven't died yet. Felt like giving up three times already, but I did have a click moment after playing around with a scheme repl for a couple of minutes. I understand the first example of call/cc now!

  (display
    (call/cc (lambda (cc)
              (display "I got here.\n")
              (cc "This string was passed to the continuation.\n")
              (display "But not here.\n"))))

Edit: Tried playing around with mit-scheme, but it doesn't work with the ctrl-v'ing the examples. Chezscheme does work.

It seems that MIT-Scheme needs the cc function to be a function of 1 argument (the return value for the continuation)[0]

Changing it to the following should work:

  (let ((start #f))
    (if (not start)
      (call/cc (lambda (cc)
                 (set! start cc))))
  
    (display "Going to invoke (start)\n")
    
    (start #f))

[0] https://www.gnu.org/software/mit-scheme/documentation/stable...

For what it's worth, almost any time Matt writes about Scheme he's likely using Racket so it'd be worth trying that for his posts!

I was actually relieved when I found this article [1] which argues against call/cc as a core language feature. Alas, I think the argument does not dismiss _delimited continuations_, so now my burden has shifted from mastering call/cc to mastering shift/reset :-/ [2].

1: http://okmij.org/ftp/continuations/against-callcc.html

2: https://en.wikipedia.org/wiki/Delimited_continuation

I think I've found a good use for continuations in a parser combinator library I'm writing, though I haven't gotten around to working on the corresponding feature yet.

Has anyone used continuations in a serious/non-academic way?

I used to implement an equivalent of exit or return statement, like in this example.

  #lang racket

  ; implement sqrt in a very inefficient way
  (let/ec exit
    (for ([i 10])
      (when (= (* i i) 49)
        (exit i)))
    "not found :(")
  ;==> 7

There are more idiomatic and easy ways to write this, but when you have to many nested and not nested for's, sometimes it is useful to have a magical exit-everything-now.

You could dispute "serious", but I've used it in a not!lisp for various control flow purposes.

  func a0 a1 = callcc \return:
    # ...
    callcc (again =)
      # ...
      x ?? return y
    again again

Continuations can be implicitly used. If you have a language that is compiled using transformation to CPS. Then continuations are everywhere under the hood.

Wow, imagine writing in such an unreadable language

Reply on news.ycombinator.com