Consider a second value indicating if an object value was accessed, regardless of being NIL?

[mmontone]

Hello.

I've been forced to patch the access library for the Djula template system (it uses this library for accessing), because I needed to know if an object was actually accessed, regardless of the result being NIL. In other words, in the current implementation, when access returns NIL, it is impossible to tell if a key with value NIL was accessed, or if the key could not be accessed. So, I introduced a second return value to the access function, indicating if the object value was accessed (like gethash does).

This is my wrapper, not a final implementation at all, but to show the idea, and I thought it could be interesting to add something like this to the library (look at the (values...) forms):

;; access library patches. we need to know if the value was accessed, regardless of
;; the result being nil. so, access should return if it the object was actually accessed
;; as a second value, regardless of result being nil. This is a problem with CL,
;; not having a real boolean datatype.

(defgeneric plist-val (id list &key test key)
  (:documentation "get a value out of a plist based on its key")
  (:method (id list &key (test #'access::equalper) (key #'identity))
    (iter (for (k v) on list by #'cddr)
      (let ((found (funcall test (funcall key k) id)))
        (if found
        (return-from plist-val (values v found)))))))

(defgeneric do-access  (o k &key type test key skip-call?)
  (:method ((o list) k &key type test key skip-call?)
    (declare (ignore skip-call?))
    (if (or (eql type :alist)
            (and (null type) (consp (first o))))
        ;;alist
    (let ((assoc (assoc k o :test test :key key)))
      (values (cdr assoc)
          (and assoc t)))
        ;;plist
    (plist-val k o :test test :key key)))

  (:method ((o array) k &key type test key skip-call?)
    (declare (ignore type test key skip-call?))
    (if (< k (length o))
    (values (aref o key) t)))

  (:method ((o hash-table) k &key type test key skip-call?)
    (declare (ignore type test key skip-call?))
    (multiple-value-bind (res found) (gethash k o)
      (if found
          (values res found)
          (awhen (ignore-errors (string k))
            (gethash it o)))))

  (:method (o  k &key type test key skip-call?)
    ;; not specializing on standard-object here
    ;; allows this same code path to work with conditions (in sbcl)
    (let ((actual-slot-name (access::has-slot? o k)))
      (cond
        ;; same package as requested, must be no accessor so handle slots
        ((eql actual-slot-name k)
         (when (slot-boundp o k)
           (values (slot-value o k) t)))

        ;; lets recheck for an accessor in the correct package
        (actual-slot-name
         (access o actual-slot-name :type type :test test :key key
                                    :skip-call? skip-call?))
        ))))

(defun access (o k &key type (test #'access::equalper) (key #'identity)
                   skip-call?)
  "Access plists, alists, arrays, hashtables and clos objects
   all through the same interface

   skip-call, skips trying to call "
  ;; make these easy to have the same defaults everywhere
  (unless test (setf test #'access::equalper))
  (unless key (setf key #'identity))
  (multiple-value-bind (res called)
      (unless skip-call?
        ;; lets suppress the warning if it is just being called through access
        (access::call-if-applicable o k :warn-if-not-a-fn? nil))
    (if called
        (values res t)
        (do-access o k :test test :key key :type type))))

[bobbysmith007]

Just returned from vacation and am catching up on things. If you get this into a patch and submit it as a pull request (hopefully with tests (pretty please)), I will be happy to include it.

[mmontone]

Ok. I'll see what I can do.

[bobbysmith007]

This seems to be a fixed issue! Thanks!