I'm reading section 3.5 of the SICP book, and I'm trying to implement
streams using Elisp. The book implements streams using the Scheme language as follows:
The core structure of a stream is a pair whose car is the current value of the sequence and its cdr is the promise of evaluating the next item, which is achieved with the following construction:
(define (cons-stream a b) (cons a (delay b)))
Where (delay b) is equivalent to (lambda () b).
When the cdr of this pair is evaluated, it forces the evaluation of the delayed element of the stream so we can get the next element and the implementation is as follows:
(define (stream-cdr stream) (force (cdr stream)))
Where (force delayed-object) is just the execution of (delayed-object).
With this construction we can now easily build recursive procedures with streams transparently as if they were regular lists like in map:
(define (stream-map proc stream)
(if (stream-null? stream)
the-empty-stream
(cons-stream (proc (stream-car s))
(stream-map p (stream-cdr s)))))
I'm trying to write a similar stream implementation using Elisp, but I haven't found a way to implement delay so that it allows me to write recursive procedures in a similar fashion as with Scheme, currently my workaround is to put the delay as a lambda expression inside the recursive procedure like this:
(defun stream-map (proc stream)
(lexical-let ((p proc)
(s stream))
(if (stream-null? stream)
the-empty-stream
(cons-stream (funcall p (stream-car s))
(lambda ()
(stream-map p (stream-cdr s)))))))
Which obviously doesn't look as nice as the Scheme implementation.
These are my Elisp version of the core functions for creating streams:
(defun cons-stream (a b) (cons a b))
(defun stream-cdr (stream) (sicp-force (cdr stream)))
(defun sicp-force (exp) (funcall exp))
How can I do to write the delay and the other stream functions so I don't have to put the lambda inside the recursive procedure?
Thanks to #Gerstmann's suggestion I was able to write the construction I was looking for. The new implementation now looks like:
(setq the-empty-stream nil)
(defun stream-null? (s) (eq s the-empty-stream))
(defun my/sicp-delay (exp) `(lambda () ,exp))
(defun my/sicp-force (exp) (funcall exp))
(defmacro my/cons-stream (a b) (list 'cons a (my/sicp-delay b)))
(defun my/stream-car (stream) (car stream))
(defun my/stream-cdr (stream) (my/sicp-force (cdr stream)))
Now I can write procedures that look just as clean as the Scheme implementation:
(defun my/stream-enumerate-interval (low high)
(lexical-let ((l low)
(h high))
(if (> l h)
the-empty-stream
(my/cons-stream low
(my/stream-enumerate-interval (1+ l) h)))))
; this returns 12
(my/stream-car (my/stream-cdr (my/stream-cdr (my/stream-enumerate-interval 10 20))))
Related
I'm trying to call a function (lambda) stored within an alist. Below is a small snippet that demonstrates what I'm trying to do:
(defvar *db* '((:add (lambda (a b)
(+ a b)))
(:sub (lambda (a b)
(- a b)))))
(defun perform-operation-on-numbers (operation a b)
"Performs specified operation on the supplied numbers."
(let ((func (second (find operation
*db*
:key #'car))))
;; TODO: Call `func` on `a` and `b`
(print func)))
(perform-operation-on-numbers :add 1 2)
No matter what I do, not even funcall is able to let me call the lambda stored against :add. How should I reference the retrieved lambda as a lambda?
Your use of quote lead to your inability to use funcall.
Look:
(setf *mydb* '((:add #'+)
(:sub #'-)))
;; ((:ADD #'+) (:SUB #'-))
I can't use funcall. But:
(setf *mydb* (list (cons :add #'+)
(cons :sub #'-)))
;; ((:ADD . #<FUNCTION +>) (:SUB . #<FUNCTION ->))
;;
;; ^^^^ "FUNCTION" ? That's better! <----------
;;
I can (funcall (cdr (first *MYDB*)) 2)
Then the succinct notation is with back-quote and comma.
As pointed out by other answers, you are manipulating code as data, where the forms below (lambda ...) are unevaluated. But even with your data:
(defvar *db* '((:add (lambda (a b)
(+ a b)))
(:sub (lambda (a b)
(- a b)))))
You can use funcall or apply, if you first use COERCE:
If the result-type is function, and object is a lambda expression, then the result is a closure of object in the null lexical environment.
For example, let's access the form associated with :add:
CL-USER> (second (assoc :add *db*))
(LAMBDA (A B) (+ A B))
The value is an unevaluated form.
You can coerce it to a function:
CL-USER> (coerce (second (assoc :add *db*)) 'function)
#<FUNCTION (LAMBDA (A B)) {536B988B}>
Maybe you want to walk the terms to check that the lambda are only using a restricted set of operations, in which case it makes sense to keep them as data. But at some point you'll want to turn these code snippets to actual functions, and you can do that with coerce:
CL-USER> (defvar *db-fns*
(loop
for (n c) in *db*
collect (list n (coerce c 'function))))
*DB-FNS*
Here you compute the functions once, and can reuse them later instead of calling coerce each time.
CL-USER> *db-fns*
((:ADD #<FUNCTION (LAMBDA (A B)) {536B9B5B}>)
(:SUB #<FUNCTION (LAMBDA (A B)) {536B9C0B}>))
(it is equivalent to calling eval on the lambda form)
That's not a function: it's a list beginning (lambda ...). If you want a function have a function, for instance by
(defvar *db* `((:add ,(lambda (a b)
(+ a b)))
(:sub ,(lambda (a b)
(- a b)))))
or, better, don't wrap the thing in some useless baggage:
(defvar *db* `((:add ,#'+
(:sub ,#'-))
When I compile the following code, SBCL complains that g!-unit-value and g!-unit are undefined. I'm not sure how to debug this. As far as I can tell, flatten is failing.
When flatten reaches the unquoted part of defunits, it seems like the entire part is being treated as an atom. Does that sound correct?
The following uses code from the book Let over Lambda:
Paul Graham Utilities
(defun symb (&rest args)
(values (intern (apply #'mkstr args))))
(defun mkstr (&rest args)
(with-output-to-string (s)
(dolist (a args) (princ a s))))
(defun group (source n)
(if (zerop n) (error "zero length"))
(labels ((rec (source acc)
(let ((rest (nthcdr n source)))
(if (consp rest)
(rec rest (cons (subseq source 0 n) acc))
(nreverse (cons source acc))))))
(if source (rec source nil) nil)))
(defun flatten (x)
(labels ((rec (x acc)
(cond ((null x) acc)
((atom x) (cons x acc))
(t (rec (car x) (rec (cdr x) acc))))))
(rec x nil)))
Let Over Lambda Utilities - Chapter 3
(defmacro defmacro/g! (name args &rest body)
(let ((g!-symbols (remove-duplicates
(remove-if-not #'g!-symbol-p
(flatten body)))))
`(defmacro ,name ,args
(let ,(mapcar
(lambda (g!-symbol)
`(,g!-symbol (gensym ,(subseq
(symbol-name g!-symbol)
2))))
g!-symbols)
,#body))))
(defun g!-symbol-p (symbol-to-test)
(and (symbolp symbol-to-test)
(> (length (symbol-name symbol-to-test)) 2)
(string= (symbol-name symbol-to-test)
"G!"
:start1 0
:end1 2)))
(defmacro defmacro! (name args &rest body)
(let* ((o!-symbols (remove-if-not #'o!-symbol-p args))
(g!-symbols (mapcar #'o!-symbol-to-g!-symbol o!-symbols)))
`(defmacro/g! ,name ,args
`(let ,(mapcar #'list (list ,#g!-symbols) (list ,#o!-symbols))
,(progn ,#body)))))
(defun o!-symbol-p (symbol-to-test)
(and (symbolp symbol-to-test)
(> (length (symbol-name symbol-to-test)) 2)
(string= (symbol-name symbol-to-test)
"O!"
:start1 0
:end1 2)))
(defun o!-symbol-to-g!-symbol (o!-symbol)
(symb "G!" (subseq (symbol-name o!-symbol) 2)))
Let Over Lambda - Chapter 5
(defun defunits-chaining (u units prev)
(if (member u prev)
(error "~{ ~a~^ depends on~}"
(cons u prev)))
(let ((spec (find u units :key #'car)))
(if (null spec)
(error "Unknown unit ~a" u)
(let ((chain (second spec)))
(if (listp chain)
(* (car chain)
(defunits-chaining
(second chain)
units
(cons u prev)))
chain)))))
(defmacro! defunits (quantity base-unit &rest units)
`(defmacro ,(symb 'unit-of- quantity)
(,g!-unit-value ,g!-unit)
`(* ,,g!-unit-value
,(case ,g!-unit
((,base-unit) 1)
,#(mapcar (lambda (x)
`((,(car x))
,(defunits-chaining
(car x)
(cons
`(,base-unit 1)
(group units 2))
nil)))
(group units 2))))))
This is kind of tricky:
Problem: you assume that backquote/comma expressions are plain lists.
You need to ask yourself this question:
What is the representation of a backquote/comma expression?
Is it a list?
Actually the full representation is unspecified. See here: CLHS: Section 2.4.6.1 Notes about Backquote
We are using SBCL. See this:
* (setf *print-pretty* nil)
NIL
* '`(a ,b)
(SB-INT:QUASIQUOTE (A #S(SB-IMPL::COMMA :EXPR B :KIND 0)))
So a comma expression is represented by a structure of type SB-IMPL::COMMA. The SBCL developers thought that this representation helps when such backquote lists need to be printed by the pretty printer.
Since your flatten treats structures as atoms, it won't look inside...
But this is the specific representation of SBCL. Clozure CL does something else and LispWorks again does something else.
Clozure CL:
? '`(a ,b)
(LIST* 'A (LIST B))
LispWorks:
CL-USER 87 > '`(a ,b)
(SYSTEM::BQ-LIST (QUOTE A) B)
Debugging
Since you found out that somehow flatten was involved, the next debugging steps are:
First: trace the function flatten and see with which data it is called and what it returns.
Since we are not sure what the data actually is, one can INSPECT it.
A debugging example using SBCL:
* (defun flatten (x)
(inspect x)
(labels ((rec (x acc)
(cond ((null x) acc)
((atom x) (cons x acc))
(t (rec (car x) (rec (cdr x) acc))))))
(rec x nil)))
STYLE-WARNING: redefining COMMON-LISP-USER::FLATTEN in DEFUN
FLATTEN
Above calls INSPECT on the argument data. In Common Lisp, the Inspector usually is something where one can interactively inspect data structures.
As an example we are calling flatten with a backquote expression:
* (flatten '`(a ,b))
The object is a proper list of length 2.
0. 0: SB-INT:QUASIQUOTE
1. 1: (A ,B)
We are in the interactive Inspector. The commands now available:
> help
help for INSPECT:
Q, E - Quit the inspector.
<integer> - Inspect the numbered slot.
R - Redisplay current inspected object.
U - Move upward/backward to previous inspected object.
?, H, Help - Show this help.
<other> - Evaluate the input as an expression.
Within the inspector, the special variable SB-EXT:*INSPECTED* is bound
to the current inspected object, so that it can be referred to in
evaluated expressions.
So the command 1 walks into the data structure, here a list.
> 1
The object is a proper list of length 2.
0. 0: A
1. 1: ,B
Walk in further:
> 1
The object is a STRUCTURE-OBJECT of type SB-IMPL::COMMA.
0. EXPR: B
1. KIND: 0
Here the Inspector tells us that the object is a structure of a certain type. That's what we wanted to know.
We now leave the Inspector using the command q and the flatten function continues and returns a value:
> q
(SB-INT:QUASIQUOTE A ,B)
For anyone else who is trying to get defmacro! to work on SBCL, a temporary solution to this problem is to grope inside the unquote structure during the flatten procedure recursively flatten its contents:
(defun flatten (x)
(labels ((flatten-recursively (x flattening-list)
(cond ((null x) flattening-list)
((eq (type-of x) 'SB-IMPL::COMMA) (flatten-recursively (sb-impl::comma-expr x) flattening-list))
((atom x) (cons x flattening-list))
(t (flatten-recursively (car x) (flatten-recursively (cdr x) flattening-list))))))
(flatten-recursively x nil)))
But this is horribly platform dependant. If I find a better way, I'll post it.
In case anyone's still interested in this one, here are my three cents. My objection to the above modification of flatten is that it might be more naturally useful as it were originally, while the problem with representations of unquote is rather endemic to defmacro/g!. I came up with a not-too-pretty modification of defmacro/g! using features to decide what to do. Namely, when dealing with non-SBCL implementations (#-sbcl) we proceed as before, while in the case of SBCL (#+sbcl) we dig into the sb-impl::comma structure, use its expr attribute when necessary and use equalp in remove-duplicates, as we are now dealing with structures, not symbols. Here's the code:
(defmacro defmacro/g! (name args &rest body)
(let ((syms (remove-duplicates
(remove-if-not #-sbcl #'g!-symbol-p
#+sbcl #'(lambda (s)
(and (sb-impl::comma-p s)
(g!-symbol-p (sb-impl::comma-expr s))))
(flatten body))
:test #-sbcl #'eql #+sbcl #'equalp)))
`(defmacro ,name ,args
(let ,(mapcar
(lambda (s)
`(#-sbcl ,s #+sbcl ,(sb-impl::comma-expr s)
(gensym ,(subseq
#-sbcl
(symbol-name s)
#+sbcl
(symbol-name (sb-impl::comma-expr s))
2))))
syms)
,#body))))
It works with SBCL. I have yet to test it thoroughly on other implementations.
Assume that the macro would take the boolean types a and b . If a is nil, then the macro should return nil (without ever evaluating b), otherwise it returns b. How do you do this?
This really depends on what you can use.
E.g., is or available? if? cond?
Here is one example:
(defmacro and (a b)
`(if ,a ,b nil)
EDIT. In response to a comment, or is more complicated because we have to avoid double evaluation:
(defmacro or (a b)
(let ((v (gensym "OR")))
`(let ((,v ,a))
(if ,v ,v ,b))))
sds's answer is nice and concise, but it has two limitations:
It only works with two arguments, whereas the built in and and or take any number of arguments. It's not too hard to update the solution to take any number of arguments, but it would be a bit more complicated.
More importantly, it's based very directly in terms of delayed operations that are already present in the language. I.e., it takes advantage of the fact that if doesn't evaluate the then or else parts until it has first evaluated the condition.
It might be a good exercise, then, to note that when a macro needs to delay evaluation of some forms, it's often the simplest strategy (in terms of implementation, but not necessarily the most efficient) to use a macro that expands to a function call that takes a function. For instance, a naive implementation of with-open-file might be:
(defun %call-with-open-file (pathname function)
(funcall function (open pathname)))
(defmacro my-with-open-file ((var pathname) &body body)
`(%call-with-open-file
,pathname
(lambda (,var)
,#body)))
Using a technique like this, you can easily get a binary and (and or):
(defun %and (a b)
(if (funcall a)
(funcall b)
nil))
(defmacro my-and (a b)
`(%and (lambda () ,a)
(lambda () ,b)))
CL-USER> (my-and t (print "hello"))
"hello" ; printed output
"hello" ; return value
CL-USER> (my-and nil (print "hello"))
NIL
or is similar:
(defun %or (a b)
(let ((aa (funcall a)))
(if aa
aa
(funcall b))))
(defmacro my-or (a b)
`(%or (lambda () ,a)
(lambda () ,b)))
To handle the n-ary case (since and and or actually take any number of arguments), you could write a function that takes a list of lambda functions and calls each of them until you get to one that would short circuit (or else reaches the end). Common Lisp actually already has functions like that: every and some. With this approach, you could implement and in terms of every by wrapping all the arguments in lambda functions:
(defmacro my-and (&rest args)
`(every #'funcall
(list ,#(mapcar #'(lambda (form)
`(lambda () ,form))
args))))
For instance, with this implementation,
(my-and (listp '()) (evenp 3) (null 'x))
expands to:
(EVERY #'FUNCALL
(LIST (LAMBDA () (LISTP 'NIL))
(LAMBDA () (EVENP 3))
(LAMBDA () (NULL 'X))))
Since all the forms are now wrapped in lambda functions, they won't get called until every gets that far.
The only difference is that and is specially defined to return the value of the last argument if all the preceding ones are true (e.g., (and t t 3) returns 3, not t, whereas the specific return value of every is not specified (except that it would be a true value).
With this approach, implementing or (using some) is no more complicated than implementing and:
(defmacro my-or (&rest args)
`(some #'funcall ,#(mapcar #'(lambda (form)
`(lambda () ,form))
args)))
I have a lisp homework I am having a hard time with it.
I have to write a function that perform a union operation. The function takes 2 inputs, either in the form of either atom or list and unions every element, preserving the order and stripping off all levels of parenthesis.
The output for the function:
(my-union 'a 'b) ;; (a b)
(my-union 'a '(b)) ;; (a b)
(my-union '(a b) '(b c)) ;; (a b c)
(my-union '(((a))) '(b(c((d e))a))) ;; (a b c d e)
I am fairly new to lisp.
Here is what I have written so far and it works only for the third example:
(defun new-union (a b)
(if (not b)
a
(if (member (car b) a)
(new-union a (cdr b))
(new-union (append a (list (car b))) (cdr b)))))
Any help would be appreciated!
Since this is your first homework, and you are new to Lisp, here is a very simple top-down approach, not worrying about performance, and making good use of the tools CL offers:
In Common Lisp, there is already a function which removes duplicates: remove-duplicates. Using it with the :from-end keyword-argument will "preserve order". Now, imagine you had a function flatten, which flattens arbitrarily nested lists. Then the solution to your question would be:
(defun new-union (list1 list2)
(remove-duplicates (flatten (list list1 list2)) :from-end t))
This is how I would approach the problem when no further restrictions are given, and there is no real reason to worry much about performance. Use as much of the present toolbox as possible and do not reinvent wheels unless necessary.
If you approach the problem like this, it boils down to writing the flatten function, which I will leave as an exercise for you. It is not too hard, one easy option is to write a recursive function, approaching the problem like this:
If the first element of the list to be flattened is itself a list, append the flattened first element to the flattened rest. If the first element is not a list, just prepend it to the flattened rest of the list. If the input is not a list at all, just return it.
That should be a nice exercise for you, and can be done in just a few lines of code.
(If you want to be very correct, use a helper function to do the work and check in the wrapping function whether the argument really is a list. Otherwise, flatten will work on atoms, too, which may or may not be a problem for you.)
Now, assuming you have written flatten:
> (defun new-union (list1 list2)
(remove-duplicates (flatten (list list1 list2)) :from-end t))
NEW-UNION
> (new-union 'a 'b)
(A B)
> (new-union 'a '(b))
(A B)
> (new-union '(a b) '(b c))
(A B C)
> (new-union '(((a))) '(b (c ((d e)) a)))
(A B C D E)
One way to approach this is to separate your concerns. One is flattening; another is duplicates-removing; yet another is result-building.
Starting with empty list as your result, proceed to add into it the elements of the first list, skipping such elements that are already in the result.
Then do the same with the second list's elements, adding them to the same result list.
(defun my-union (a b &aux (res (list 1)) (p res))
(nadd-elts p a)
(nadd-elts p b)
(cdr res))
nadd-elts would add to the end of list, destructively updating its last cell (pointed to by p) using e.g. rplacd. An example is here.
To add elements, nadd-elts would emulate the flattening procedure, and add each leaf element into p after checking res for duplicates.
Working in functional style, without destructive update, the general approach stays the same: start with empty result list, add first list into it - without duplicates - then second.
(defun my-union (a b &aux res)
(setq res (add-into res a))
(setq res (add-into res b))
res)
Now we're left with implementing the add-into function.
(defun add-into (res a &aux r1 r2)
(cond
((atom a) .... )
(T (setq r1 (add-into res (car a)))
(setq r2 (............ (cdr a)))
r2)))
The above can be re-written without the auxiliary variables and without set primitives. Try to find out how... OK here's what I meant by that:
(defun my-union (a b) (add-into NIL (cons a b)))
(defun add-into (res a)
(cond
((atom a) .... )
(T (add-into (add-into res (car a))
(cdr a)))))
Unless you are not allowed to use hash table (for some reason I've encountered this as a requirement before), you could come up with an ordering function that will help you build the resulting set in the way, that you don't have to repeat the search over and over again.
Also, since nested lists are allowed your problem scales down to only removing duplicates in a tree (as you can simply append as many lists as you want before you start processing them.
Now, I'll try to show few examples of how you could do it:
;; Large difference between best and worst case.
;; Lists containing all the same items will be processed
;; in square time
(defun union-naive (list &rest lists)
(when lists (setf list (append list lists)))
(let (result)
(labels ((%union-naive (tree)
(if (consp tree)
(progn
(%union-naive (car tree))
(when (cdr tree) (%union-naive (cdr tree))))
(unless (member tree result)
(setq result (cons tree result))))))
(%union-naive list) result)))
;; Perhaps the best solution, it is practically linear time
(defun union-hash (list &rest lists)
(when lists (setf list (append list lists)))
(let ((hash (make-hash-table)) result)
(labels ((%union-hash (tree)
(if (consp tree)
(progn
(%union-hash (car tree))
(when (cdr tree) (%union-hash (cdr tree))))
(setf (gethash tree hash) t))))
(%union-hash list))
(maphash
#'(lambda (a b)
(declare (ignore b))
(push a result)) hash)
result))
;; This will do the job in more time, then the
;; solution with the hash-map, but it requires
;; significantly less memory. Memory is, in fact
;; a more precious resource some times, but you
;; have to decide what algo to use based on the
;; data size
(defun union-flatten (list &rest lists)
(when lists (setf list (append list lists)))
(labels ((%flatten (tree)
(if (consp tree)
(if (cdr tree)
(nconc (%flatten (car tree))
(%flatten (cdr tree)))
(%flatten (car tree)))
(list tree))))
;; the code below is trying to do something
;; that you could've done using
;; (remove-duplicates (%flatten list))
;; however sorting and then removing duplicates
;; may prove to be more efficient
(reduce
#'(lambda (a b)
(cond
((atom a) (list a))
((eql (car a) b) b)
(t (cons b a))))
(sort (%flatten list)
#'(lambda (a b)
(string< (symbol-name a)
(symbol-name b)))))))
(union-naive '(((a))) '(b(c((d e))a)))
(union-hash '(((a))) '(b(c((d e))a)))
(union-flatten '(((a))) '(b(c((d e))a)))
Notice that the function I've used to order elements is not universal, but you would probably be able to come up with an alternative function for any sort of data. Any fast hashing function in general would do, I've used this one for simplicity.
I'm a Lisp beginner. I'm trying to memoize a recursive function for calculating the number of terms in a Collatz sequence (for problem 14 in Project Euler). My code as of yet is:
(defun collatz-steps (n)
(if (= 1 n) 0
(if (evenp n)
(1+ (collatz-steps (/ n 2)))
(1+ (collatz-steps (1+ (* 3 n)))))))
(defun p14 ()
(defvar m-collatz-steps (memoize #'collatz-steps))
(let
((maxsteps (funcall m-collatz-steps 2))
(n 2)
(steps))
(loop for i from 1 to 1000000
do
(setq steps (funcall m-collatz-steps i))
(cond
((> steps maxsteps)
(setq maxsteps steps)
(setq n i))
(t ())))
n))
(defun memoize (fn)
(let ((cache (make-hash-table :test #'equal)))
#'(lambda (&rest args)
(multiple-value-bind
(result exists)
(gethash args cache)
(if exists
result
(setf (gethash args cache)
(apply fn args)))))))
The memoize function is the same as the one given in the On Lisp book.
This code doesn't actually give any speedup compared to the non-memoized version. I believe it's due to the recursive calls calling the non-memoized version of the function, which sort of defeats the purpose. In that case, what is the correct way to do the memoization here? Is there any way to have all calls to the original function call the memoized version itself, removing the need for the special m-collatz-steps symbol?
EDIT: Corrected the code to have
(defvar m-collatz-steps (memoize #'collatz-steps))
which is what I had in my code.
Before the edit I had erroneously put:
(defvar collatz-steps (memoize #'collatz-steps))
Seeing that error gave me another idea, and I tried using this last defvar itself and changing the recursive calls to
(1+ (funcall collatz-steps (/ n 2)))
(1+ (funcall collatz-steps (1+ (* 3 n))))
This does seem to perform the memoization (speedup from about 60 seconds to 1.5 seconds), but requires changing the original function. Is there a cleaner solution which doesn't involve changing the original function?
I assume you're using Common-Lisp, which has separate namespaces for variable and function names. In order to memoize the function named by a symbol, you need to change its function binding, through the accessor `fdefinition':
(setf (fdefinition 'collatz-steps) (memoize #'collatz-steps))
(defun p14 ()
(let ((mx 0) (my 0))
(loop for x from 1 to 1000000
for y = (collatz-steps x)
when (< my y) do (setf my y mx x))
mx))
Here is a memoize function that rebinds the symbol function:
(defun memoize-function (function-name)
(setf (symbol-function function-name)
(let ((cache (make-hash-table :test #'equal)))
#'(lambda (&rest args)
(multiple-value-bind
(result exists)
(gethash args cache)
(if exists
result
(setf (gethash args cache)
(apply fn args)))))))
You would then do something like this:
(defun collatz-steps (n)
(if (= 1 n) 0
(if (evenp n)
(1+ (collatz-steps (/ n 2)))
(1+ (collatz-steps (1+ (* 3 n)))))))
(memoize-function 'collatz-steps)
I'll leave it up to you to make an unmemoize-function.
something like this:
(setf collatz-steps (memoize lambda (n)
(if (= 1 n) 0
(if (evenp n)
(1+ (collatz-steps (/ n 2)))
(1+ (collatz-steps (1+ (* 3 n))))))))
IOW: your original (non-memoized) function is anonymous, and you only give a name to the result of memoizing it.
Note a few things:
(defun foo (bar)
... (foo 3) ...)
Above is a function that has a call to itself.
In Common Lisp the file compiler can assume that FOO does not change. It will NOT call an updated FOO later. If you change the function binding of FOO, then the call of the original function will still go to the old function.
So memoizing a self recursive function will NOT work in the general case. Especially not if you are using a good compiler.
You can work around it to go always through the symbol for example: (funcall 'foo 3)
(DEFVAR ...) is a top-level form. Don't use it inside functions. If you have declared a variable, set it with SETQ or SETF later.
For your problem, I'd just use a hash table to store the intermediate results.
Changing the "original" function is necessary, because, as you say, there's no other way for the recursive call(s) to be updated to call the memoized version.
Fortunately, the way lisp works is to find the function by name each time it needs to be called. This means that it is sufficient to replace the function binding with the memoized version of the function, so that recursive calls will automatically look up and reenter through the memoization.
huaiyuan's code shows the key step:
(setf (fdefinition 'collatz-steps) (memoize #'collatz-steps))
This trick also works in Perl. In a language like C, however, a memoized version of a function must be coded separately.
Some lisp implementations provide a system called "advice", which provides a standardized structure for replacing functions with enhanced versions of themselves. In addition to functional upgrades like memoization, this can be extremely useful in debugging by inserting debug prints (or completely stopping and giving a continuable prompt) without modifying the original code.
This function is exactly the one Peter Norvig gives as an example of a function that seems like a good candidate for memoization, but which is not.
See figure 3 (the function 'Hailstone') of his original paper on memoization ("Using Automatic Memoization as a Software Engineering Tool in Real-World AI Systems").
So I'm guessing, even if you get the mechanics of memoization working, it won't really speed it up in this case.
A while ago I wrote a little memoization routine for Scheme that used a chain of closures to keep track of the memoized state:
(define (memoize op)
(letrec ((get (lambda (key) (list #f)))
(set (lambda (key item)
(let ((old-get get))
(set! get (lambda (new-key)
(if (equal? key new-key) (cons #t item)
(old-get new-key))))))))
(lambda args
(let ((ans (get args)))
(if (car ans) (cdr ans)
(let ((new-ans (apply op args)))
(set args new-ans)
new-ans))))))
This needs to be used like so:
(define fib (memoize (lambda (x)
(if (< x 2) x
(+ (fib (- x 1)) (fib (- x 2)))))))
I'm sure that this can be ported to your favorite lexically scoped Lisp flavor with ease.
I'd probably do something like:
(let ((memo (make-hash-table :test #'equal)))
(defun collatz-steps (n)
(or (gethash n memo)
(setf (gethash n memo)
(cond ((= n 1) 0)
((oddp n) (1+ (collatz-steps (+ 1 n n n))))
(t (1+ (collatz-steps (/ n 2)))))))))
It's not Nice and Functional, but, then, it's not much hassle and it does work. Downside is that you don't get a handy unmemoized version to test with and clearing the cache is bordering on "very difficult".