[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Scheme-reports] Proposed new SRFI for immutable lists

On Wed, 3 Sep 2014 15:17:22 +0900, Alex Shinn <alexshinn@x> said:

> On Wed, Sep 3, 2014 at 2:00 PM, Kevin Wortman <kwortman@x> wrote:
>> ...
>> Are cycles actually possible in strict immutable structures? When you
>> create a head node, there's no way for it to have a circular link to a tail
>> node that doesn't exist yet.

> Yes, if you provide implementation support for it,
> see http://docs.racket-lang.org/reference/shared.html

> Also for lazy data structures see
> http://www.haskell.org/haskellwiki/Tying_the_Knot,
> for which I have a Scheme version lying around somewhere.

  By the way, some languages offer yet
  another approach, in principle similar to
  the latter, but more procedural.  It is
  to make the object construction protocol
  a little more elaborate so the user (not
  just the implementor) can see object
  _allocation_ as a separate step from
  object _initialization_.  Probably
  needless to add, in this way a reference
  to the object becomes available (for
  plugging into other objects) before the
  construction of the object is finished
  (with references to those other objects
  plugged into it).  This is just with
  regards to the possibility of cycles in
  strictly immutable structures; whether
  this can be a candidate for adopting by
  Scheme is another question.

  There is also the more general approach
  of "arbitrarily delayed once-only
  component initialization", where any
  component of an immutable object can be
  initialized at any time after object
  creation (no matter how much later), but
  only once.  If I recall correctly, Oz
  allows (something like) that, but I
  forget how exactly it is called there.


Vassil Nikolov | Васил Николов | <vnikolov@x>

"Be careful how you fix what you don't understand."  (Brooks 2010, 185)

Scheme-reports mailing list