[racket] Confusion with circular structs

Hi Nick!

Interesting observation.  I tried the same thing with immutable hash tables
in place of sets (just mapping each element to #t), and got the same
results.  So I don't think the set implementation is the culprit here.  I
suspect the built-in equality and/or hash functions for either immutable
hash tables, or transparent structs, or both, doesn't properly handle
circular values.

Carl Eastlund

On Fri, Jul 26, 2013 at 5:25 PM, Nicholas Labich <labichn at gmail.com> wrote:

> I'm running into what I find to be surprising behavior with circular
> structures (vertices of a simple graph).
>
> Graphically (see code below to reproduce):
> 0 --> 1 ;; no problem
> 2-+      ;; also no problem
>  ^-+
>
> But when I try to add an edge between 0 and 2, I hit an infinite loop at
> line 88 of racket/set.rkt (in 5.3.3) or line 447 of
> racket/private/set-types.rkt (in the latest 5.90.0.2), both of which are
> the hash-proc for set's gen:equal+hash.
>
> Is this the expected behavior (after all, OCaml makes no guarantee of
> termination for (=) on circular data)? I'm surprised that the self-edge on
> 2 doesn't result in the same. I've also tried to reproduce the same using
> `shared', but I'm assuming I still misunderstand its functionality because
> the result is not quite what I expect it to be.
>
> Any insight would be much appreciated.
>
> Nick
>
> Welcome to Racket v5.90.0.2.
> > (struct vertex (value preds succs) #:mutable #:transparent)
> > (define v0 (vertex 0 (set) (set)))
> > (define v1 (vertex 1 (set) (set)))
> > (set-vertex-succs! v0 (set v1))
> > (set-vertex-preds! v1 (set v0)) ;; 0 --> 1
> > (equal? v0 (set-first (vertex-preds v1)))
> #t
> > (define v2 (vertex 2 (set) (set)))
> > (set-vertex-preds! v2 (set v2)) ;; 2--+
> > (set-vertex-succs! v2 (set v2)) ;; ^--+
> > v2
> #0=(vertex 2 (set #0#) (set #0#))
> ;; try to add the already existing set
> > (set-vertex-succs! v0 (set-add (vertex-succs v0) v2)) ;; loops and loops
>   C-c C-c^Cuser break
>   context...:
>  .../racket/collects/racket/private/set-types.rkt:447:3: hash-proc
>  .../racket/collects/racket/private/set-types.rkt:447:3: hash-proc
>  .../racket/collects/racket/private/set-types.rkt:447:3: hash-proc
>  .../racket/collects/racket/private/set-types.rkt:447:3: hash-proc...
> ;; try to make a new set instead
> > (set-vertex-succs! v0 (set v1 v2))
>   C-c C-c^Cuser break
>   context...:
>   .../racket/collects/racket/private/set-types.rkt:447:3: hash-proc
>   .../racket/collects/racket/private/set-types.rkt:447:3: hash-proc
>   .../racket/collects/racket/private/set-types.rkt:447:3: hash-proc
>   .../racket/collects/racket/private/set-types.rkt:447:3: hash-proc...
>
> > (shared ([v0 (vertex 0 (set) succs0)]
>            [v1 (vertex 1 preds1 (set))]
>            [v2 (vertex 2 preds2 succs2)]
>            [succs0 (set v1 v2)]
>            [preds1 (set v0)]
>            [preds2 (set v0 v2)]
>            [succs2 (set v2)])
>     (values v0 v1 v2))
> (vertex 0 (set) (set #0=(vertex 2 (set #0#) (set #0#))))
> (vertex 1 (set (vertex 0 (set) (set #0=(vertex 2 (set #0#) (set #0#)))))
> (set))
> #0=(vertex 2 (set #0#) (set #0#))
> ;; I must be using `shared' incorrectly...
>
> ____________________
>   Racket Users list:
>   http://lists.racket-lang.org/users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.racket-lang.org/users/archive/attachments/20130726/8056640c/attachment.html>