Mailing list for all users of the OCaml language and system.
 help / color / mirror / Atom feed
From: John Prevost <jmp@arsdigita.com>
To: Remi VANICAT <remi.vanicat@labri.u-bordeaux.fr>
Cc: caml-list@inria.fr
Subject: Re: bottom types and threaded exits
Date: 25 Sep 2000 14:38:25 -0400	[thread overview]
Message-ID: <87em28e2qm.fsf@localhost.localdomain.> (raw)
In-Reply-To: Remi VANICAT's message of "25 Sep 2000 12:08:50 +0200"

>>>>> "rv" == Remi VANICAT <remi.vanicat@labri.u-bordeaux.fr> writes:

    rv> Julian Assange <proff@iq.org> writes:
    >> Pervasives.exit is of type int -> 'a
    >> 
    >> Here we see ocaml using 'a to represent _|_. This hack is
    >> presumably so type unification still works in the face of
    >> potentially non-terminating computations, e.g:
    >> 
    >> let f a = try f a with Failure _ -> exit(1)
    >> 
    >> How can one force 'a? For instance, Thread.exit and
    >> 
    >> let f () = while true do () done
    >> 
    >> has a type of unit -> unit.
    >> 
    >> One can write something such as
    >> 
    >> let f () = while true do () done ; Pervasives.exit (1)
    >> 
    >> But this is clearly a hack.

>From what I can tell, what you're trying to do here is just make the
result type of a non-terminating computation go to 'a.  This is
actually quite easy--you just need to loop in a way that doesn't
constrain the type.  while ... do ... done always returns unit (which
is good, since usually you use it in a way that terminates.

The appropriate way to do it that will get type 'a is (in your case):

let rec f () = f ()

Note that if you choose the more direct translation

let rec f () = if true then f ()

you get constrained to type unit -> unit again, since a one-branched
if returns unit for the else branch.

So, an example of a vaguely useful function would be:

let rec call_forever f x = f x; call_forever f x

which has type:

val call_forever : ('a -> 'b) -> 'a -> 'c


So, it is not so much that 'a as bottom is a hack, as that the system
only produces such types when the result type is totally
unconstrained--and that doesn't happen with things like while loops,
only with recursion--where a function always returns the result of
calling itself (or another such function.)


John.




  reply	other threads:[~2000-09-26 19:16 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2000-09-23 14:25 Julian Assange
2000-09-24 19:14 ` Pierre Weis
2000-09-25 10:08 ` Remi VANICAT
2000-09-25 18:38   ` John Prevost [this message]
2000-09-26  9:16 ` Xavier Leroy
2000-09-30  9:12   ` Julian Assange
2000-09-30 14:25     ` Pierre Weis
2000-09-30 15:19       ` Markus Mottl
2000-09-30 14:28     ` Stefan Monnier
2000-10-02 13:11       ` Patrick M Doane
2000-10-02 16:28         ` Markus Mottl
2000-10-05 18:39           ` Pierre Weis
2000-09-25 19:56 Damien Doligez

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=87em28e2qm.fsf@localhost.localdomain. \
    --to=jmp@arsdigita.com \
    --cc=caml-list@inria.fr \
    --cc=remi.vanicat@labri.u-bordeaux.fr \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox