I was getting stack overflow with the following code, then I tried it in SBCL and it worked. Wondering what causes the difference there.
Specifically: While I do plan to move to SBCL at some point, can this be made to work in CLISP?
(defvar *objs* nil) ; [1]
(defun parents (obj) (gethash :parents obj))
(defun obj (&rest parents) ; [2]
(let ((obj (make-hash-table)))
(push obj *objs*)
(setf (parents obj) parents)
obj))
(defun (setf parents) (val obj) ; [3]
(prog1 (setf (gethash :parents obj) val)
(make-precedence obj)))
(defun make-precedence (obj) ; [4]
(setf (gethash :preclist obj) (precedence obj))
(dolist (x *objs*)
(if (member obj (gethash :preclist x))
(setf (gethash :preclist x) (precedence x)))))
(defun precedence (obj) ; [5]
(delete-duplicates (traverse obj)))
(defun traverse (x) ; [6]
(cons x (mapcan #'traverse (gethash :parents x))))
;; [1] We'll store a list of objects we create in *obj*.
;; [2] Function to create an object, called like (setf scoundrel (obj)).
;; [3] Set an objects (multiple) parents & rebuild precedence list for all affected objs.
;; [4] Rebuild precedence list for obj, then for all affected objs.
;; [5] Returns a list of object and all its ancestors in precedence order as we define it.
;; (Can read it like (-> obj traverse delete-duplicates) if it helps)
;; [6] Cons an object to all its parents recursively; depth first search.
;; I pulled this out of labels in precedence above it for clarity & testability.
;; Source: PG's ANSI Common Lisp, Chapter 17, "Example: Objects".
Example - SBCL
(setf scoundrel (obj))
; #<HASH-TABLE :TEST EQL :COUNT 2 {1001A01893}>
(setf sc2 (obj scoundrel))
; #<HASH-TABLE :TEST EQL :COUNT 2 {1001A1F153}>
*objs*
; (#<HASH-TABLE :TEST EQL :COUNT 2 {1001A1F153}>
; #<HASH-TABLE :TEST EQL :COUNT 2 {1001A01893}>)
(parents scoundrel)
; NIL
; T
(parents sc2)
; (#<HASH-TABLE :TEST EQL :COUNT 2 {1001A01893}>)
; T
Example - GNU CLISP
(setf scoundrel (obj))
;; - Lisp stack overflow. RESET
*objs*
;; - Lisp stack overflow. RESET
It might be worth mentioning that I haven't studied the dual interpreted and compiled nature of lisp a lot yet. So far I've simply been using it as an interpreted language; by pasting the above functions into the clisp repl.
So I suspect that compiling all these functions could be one thing to consider. I note we can compile and compile-file, I can't see an operator which compiles all user defined functions though.
