equivalent of "this" in goops?

equivalent of "this" in goops?

[Greg Troxel]

I'm using guile 1.6.x,and doing network mapping.

In a class representing a node, I want to have a slot with an
init-thunk, and I would like to call a procedure with the class object
as an argument.  The slot holds a class object which is basically a
hash table of links in a network, and it needs the node object so it
can hash on 'other end node id'.

So, I'd like to pass 'this' in c++ terms, and I don't see how to do
that from reading goops.info.  I realize there are issues with the
class not being fully initialized, and in my case that's safe since
I'm just storing the object in a variable and not using it until a
link is added, which doesn't happen during object initialization.
Thus I realize that what I want to do is vaguely unschemely.

My fallback is to initialize the linkset on first use, or after object
createion, or something like that.

-- 
        Greg Troxel <gdt@ir.bbn.com>


_______________________________________________
Guile-user mailing list
Guile-user@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-user

Re: equivalent of "this" in goops?

[Andreas Rottmann]

Greg Troxel <gdt@ir.bbn.com> writes:

> I'm using guile 1.6.x,and doing network mapping.
>
> In a class representing a node, I want to have a slot with an
> init-thunk, and I would like to call a procedure with the class object
> as an argument.  The slot holds a class object which is basically a
> hash table of links in a network, and it needs the node object so it
> can hash on 'other end node id'.
>
> So, I'd like to pass 'this' in c++ terms, and I don't see how to do
> that from reading goops.info.  I realize there are issues with the
> class not being fully initialized, and in my case that's safe since
> I'm just storing the object in a variable and not using it until a
> link is added, which doesn't happen during object initialization.
> Thus I realize that what I want to do is vaguely unschemely.

FWICT (not being a GOOPS expert), this is not possible without
extending GOOPS: the init-thunk of a slot is specified at class
creation time, and the earliest time "this" is avaiable is in the
"initialize" generic. However, since the init-thunk is called at
initialization time, "this" would be available at the time it is run,
but there is no way to pass it, since the init-thunk, is, well, a
thunk ;). I could imagine a init-function (name to be argued), which
gets passed all arguments from "initialize" could be a useful
extension.

> My fallback is to initialize the linkset on first use, or after object
> createion, or something like that.
>
Given a GOOPS with this hypothetical extension is far from official
release ;), a workaround is probably the only feasibly short-term
solution (given my assumtion that this is not possible with current
GOOPS).

Cheers, Rotty
-- 
Andreas Rottmann         | Rotty@ICQ      | 118634484@ICQ | a.rottmann@gmx.at
http://yi.org/rotty      | GnuPG Key: http://yi.org/rotty/gpg.asc
Fingerprint              | DFB4 4EB4 78A4 5EEE 6219  F228 F92F CFC5 01FD 5B62

The best way to accelerate a Windows machine is at 9.81 m/s^2



_______________________________________________
Guile-user mailing list
Guile-user@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-user

Re: equivalent of "this" in goops?

[Neil Jerram]

Andreas Rottmann wrote:
> Greg Troxel <gdt@ir.bbn.com> writes:
> 
>>In a class representing a node, I want to have a slot with an
>>init-thunk, [...] So, I'd like to pass 'this' in c++ terms, and I don't see how to do
>>that from reading goops.info.
> 
> FWICT (not being a GOOPS expert), this is not possible without
> extending GOOPS: the init-thunk of a slot is specified at class
> creation time, and the earliest time "this" is avaiable is in the
> "initialize" generic. However, since the init-thunk is called at
> initialization time, "this" would be available at the time it is run,
> but there is no way to pass it, since the init-thunk, is, well, a
> thunk ;). I could imagine a init-function (name to be argued), which
> gets passed all arguments from "initialize" could be a useful
> extension.

Absolutely right.  Your best bet at this point is to customize the 
initialize method for the class:

(define-method (initialize (this <node>) initargs)
   (next-method)
   (slot-set this 'name-of-slot-with-init-thunk
             (init-function this)))

(It's probably also possible to use the MOP to support #:init-proc as a 
new slot option, but that would require more thought.)

Regards,
	Neil


_______________________________________________
Guile-user mailing list
Guile-user@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-user

Re: equivalent of "this" in goops?

[Greg Troxel]

Thanks very much for your suggestion.  I now have the following (but
haven't tested yet):

;; set of links
(define-class <linkset> (<sinew-object>)
  ;;  hash key is on other endpoint
  (link-hash #:accessor link-hash
             #:init-thunk (lambda () (make-hash-table 61)))
  (my-node #:accessor my-node #:init-keyword my-node))

(define-class <node> (<sinew-object>)
  [other stuff]

  ;; links is a set of links, findable by the other endpoint
  (links-ip #:accessor links-ip)
  (links-udaan #:accessor links-udaan))

;; override initialize in order to pass this to linkset constructor
(define-method (initialize (this <node>) initargs)
   (next-method)
   (slot-set this 'links-ip (make <linkset> #:my-node this))
   (slot-set this 'links-udaan (make <linkset> #:my-node this))

-- 
        Greg Troxel <gdt@ir.bbn.com>


_______________________________________________
Guile-user mailing list
Guile-user@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-user