Change variable defined in let function - lisp

Given the following let function:
(let ((foo (list a b c d)))
foo)
How can i modify the list foo?
(let ((foo (list a b c d)))
;some code
foo)
so the returned foo looks like eks: '(some new list) or (a b modified d)
i tried (set) but foo will still return as the original list.

You can use setf to modify a specific element in a list.
(let ((foo (list 'a 'b 'c 'd)))
(setf (third foo) 'modified)
foo)
This will return (a b modified d).
Or if you want to replace the whole variable, you can assign directly to it with setq:
(let ((foo (list 'a 'b 'c 'd)))
(setq foo (list 'some 'new 'list))
foo)
You can't use set with lexical variables, only special variables.

Related

to form a new symbol using list items

say I have two lists in lisp
(setq a '(p q))
(setq b '(1 2))
(car a) is p
(car b) is 1
now I want to define a symbol '(test p 1) but if I use below
(setq c '(test (car a) (car b)))
I get '(test (car a) (car b))
it is understandable, but I just want to know how can I substitute those (car a) to p and (car b) to 1 and form a new symbol of '(test p 1)
Thanks
First off, setq should not be used on unbound variables. You can use setq on established variables. Also for global variables you should use *earmuffs*.
(defparameter *a* '(p q))
(defparameter *b* '(1 2))
(car *a*) ; ==> p
(car *b*) ; ==> 1
The quote will use the quotes structure as data. That means everything expr where you write 'expr will never be evaluated beyond taking the data verbatim. New lists are created with cons. eg.
;; creates/updates binding *x* to point at the newly created list (test p 1)
(defparameter *c* (cons 'test
(cons (car *a*)
(cons (car *b*)
'()))))
cons is the primitive, but CL has several other ways to create lists. eg. the same with the function list:
;; creates/updates binding *x* to point at the newly created list (test p 1)
(defparameter *c* (list 'test (car *a*) (car *b*)))
The second the structure becomes more complex using quasiquote/unquote/unquote-splice is a lot easier.
;; creates/updates binding *x* to point at the newly created list (test p 1)
(defparameter *c* `(test ,(car *a*) ,(car *b*)))
;; more complex example
(defmacro my-let ((&rest bindings) &body body)
`((lambda ,(mapcar #'car bindings)
,#body)
,(mapcar #'cadr bindings)))
(macroexpand-1 '(my-let ((a 10) (b 20)) (print "hello") (+ (* a a) (* b b))))
; ==> ((lambda (a b)
; (print "hello")
; (+ (* a a) (* b b)))
; (10 20))
Note that this is just sugar for the identical structure made with cons, list, and append. It might be optimized for minimal memory use so will share structure. eg. `(,x b c) in a procedure will do (cons x '(b c)) which means if you create two versions their cdr will be eq and you should refrain from mutating these parts.
If you want to make a list the function you want is list:
(list 'test (car a) (car b))`
Will be the list (test p 1).
Note that the purpose of quote (abbreviated ', so '(x) is identical to (quote (x))) is simply to tell the evaluator that what follows is literal data, not code. So, in (list 'test ...), which is the same as (list (quote test) ...) then quote tells the evaluator that test is being used as a literal datum, rather than as the name of a binding, and similarly '(p q) means 'this is a literal list with elements p and q', while (p q) means 'this is a form for evaluation, whose meaning depends on what p is')
To complete the answer from tfb, you can write
`(test ,(car a) ,(car b)
This is strictly the same of
(list 'test (car a) (car b)

Any good way to declare unused variables in destructuring-bind?

I can't figure, is there any way to put something like _ in erlang, for "unused value" in destructuring-bind?
For example there we have something like that:
(destructuring-bind ((_SNIPPET
(_TITLE . title)
(_DESCRIPTION . description)
_RESOURCE-ID (_VIDEO-ID . video-id)))) entry
(declare (ignore
_SNIPPET _TITLE _DESCRIPTION _RESOURCE-ID _VIDEO-ID))
(list video-id title description)))
It'll be great not to put specific variable for every unused value, and write something like that:
(destructuring-bind ((_
(_ . title)
(_ . description)
(_ (_ . video-id)))) entry
(list video-id title description)))
Is there any way to get such behavior with standart destructuring-bind or any other standart macros? Or I have to use some ML-like pattern matching library, and if so - which one?
It's not possible with DESTRUCTURING-BIND (you can't use a variable more than once, some compiler will complain). You can enumerate the variables, _1, _2, ... But then you have to ignore each of them.
LOOP can do it:
CL-USER 23 > (loop for ((a b nil c) nil d) in '(((1 2 3 4) 5 6)
((1 2 3 4) 5 6))
collect (list a b c d))
((1 2 4 6) (1 2 4 6))
NIL is used as the wildcard variable.
You can reuse the LOOP macro:
(defmacro match-bind (pattern object &body body)
`(loop with ,pattern = ,object
while nil
finally (return (progn ,#body))))
CL-USER 37 > (match-bind ((a b nil c) nil d)
'((1 2 3 4) 5 6)
(list a b c d))
(1 2 4 6)
You can use some LET-MATCH from some library. For example: https://github.com/schani/clickr/blob/master/let-match.lisp
There are probably more fancy versions.
There's nothing built into the language for this. Rainer Joswig's answer points out that loop can do some destructuring, but it doesn't do nearly as much. In an earlier version of this answer, I suggested traversing the destructuring lambda list and collecting a list of all the symbols that begin with _ and adding a declaration to the form to ignore those variables. A safer version replaces each one with a fresh variable (so that there are no repeated variables), and ignores them all. Thus something like
(destructuring-bind (_a (_b c)) object
c)
would expand into
(destructuring-bind (#:g1 (#:g2 c)) object
(declare (ignore #:g1 #:g2))
c)
This approach will work OK if you're only using the "data-directed" described in 3.4.4.1.1 Data-directed Destructuring by Lambda Lists. However, if you're using "lambda-list-directed" approach described in 3.4.4.1.2 Lambda-list-directed Destructuring by Lambda Lists, where you can use lambda-list keywords like &optional, &key, etc., then things are much more complicated, because you shouldn't replace variables in some parts of those. For instance, if you have
&optional (_x '_default-x)
then it might be OK to replace _x with something, but not _default-x, because the latter isn't a pattern. But, in Lisp, code is data, so we can still write a macro that maps over the destructuring-lambda-list and replaces only in locations that are patterns. Here's somewhat hairy code that does just that. This takes a function and a destructuring lambda list, and calls the function for each pattern variable in the lambda list, along with the type of the argument (whole, required, optional, etc.).
(defun map-dll (fn list)
(let ((result '())
(orig list)
(keywords '(&allow-other-keys &aux &body
&key &optional &rest &whole)))
(labels ((save (x)
(push x result))
(handle (type parameter)
(etypecase parameter
(list (map-dll fn parameter))
(symbol (funcall fn type parameter)))))
(macrolet ((parse-keyword ((&rest symbols) &body body)
`(progn
(when (and (not (atom list))
(member (first list) ',symbols))
(save (pop list))
,#body)))
(doparameters ((var) &body body)
`(do () ((or (atom list) (member (first list) keywords)))
(save (let ((,var (pop list)))
,#body)))))
(parse-keyword (&whole)
(save (handle :whole (pop list))))
(doparameters (required)
(handle :required required))
(parse-keyword (&optional)
(doparameters (opt)
(if (symbolp opt)
(handle :optional opt)
(list* (handle :optional (first opt)) (rest opt)))))
(when (and (atom list) (not (null list))) ; turn (... . REST)
(setq list (list '&rest list))) ; into (... &rest REST)
(parse-keyword (&rest &body)
(save (handle :rest (pop list))))
(parse-keyword (&key)
(doparameters (key)
(if (symbolp key)
(handle :key key)
(destructuring-bind (keyspec . more) key
(if (symbolp keyspec)
(list* (handle :key keyspec) more)
(destructuring-bind (keyword var) keyspec
(list* (list keyword (handle :key var)) more)))))))
(parse-keyword (&allow-other-keys))
(parse-keyword (&aux)
(doparameters (aux) aux))
(unless (null list)
(error "Bad destructuring lambda list: ~A." orig))
(nreverse result)))))
Using this, it's pretty easy to write a destructuring-bind* that replaces each pattern variable beginning with _ with a fresh variable that will be ignored in the body.
(defmacro destructuring-bind* (lambda-list object &body body)
(let* ((ignores '())
(lambda-list (map-dll (lambda (type var)
(declare (ignore type))
(if (and (> (length (symbol-name var)) 0)
(char= #\_ (char (symbol-name var) 0)))
(let ((var (gensym)))
(push var ignores)
var)
var))
lambda-list)))
`(destructuring-bind ,lambda-list ,object
(declare (ignore ,#(nreverse ignores)))
,#body)))
Now we should look at the expansions it produces:
(macroexpand-1
'(destructuring-bind* (&whole (a _ . b)
c _ d
&optional e (f '_f)
&key g _h
&aux (_i '_j))
object
(list a b c d e f g)))
;=>
(DESTRUCTURING-BIND
(&WHOLE (A #:G1041 &REST B) C #:G1042 D
&OPTIONAL E (F '_F)
&KEY G #:G1043
&AUX (_I '_J))
OBJECT
(DECLARE (IGNORE #:G1041 #:G1042 #:G1043))
(LIST A B C D E F G))
We haven't replaced anywhere we shouldn't (init forms, aux variables, etc.), but we've taken care of the places that we should. We can see this work in your example too:
(macroexpand-1
'(destructuring-bind* ((_ (_ . title)
(_ . description)
_
(_ . video-id)))
entry
(list video-id title description)))
;=>
(DESTRUCTURING-BIND ((#:G1044 (#:G1045 &REST TITLE)
(#:G1046 &REST DESCRIPTION)
#:G1047
(#:G1048 &REST VIDEO-ID)))
ENTRY
(DECLARE (IGNORE #:G1044 #:G1045 #:G1046 #:G1047 #:G1048))
(LIST VIDEO-ID TITLE DESCRIPTION))

Apply and keyword arguments in lisp

I can use keyword parameters in LISP
(member 'a '(a b c) :test #'eq)
However, when I tried to use apply to invoke member method
(apply #'member 'a '(a b c) :test #'eq)
I have error message as follows:
MEMBER: keyword arguments in (:TEST) should occur pairwise
[Condition of type SYSTEM::SIMPLE-PROGRAM-ERROR]
The solution was
(apply #'member 'a '(a b c) '(:test eq))
Whereas without the keyword arguments
(apply #'member 'a '((a b c)))
What's the logic behind this? Why '(:test #'eq) raises an error?
ADDED
This is the reason why I asked this question.
I have code from ANSI Common Lispbook page 103.
(defun our-adjoin (obj lst &rest args)
(if (apply #'member obj lst args)
lst
(cons obj lst)))
When I tried (our-adjoin 'a '(a b c)) it returns the result (A B C), but the our-adjoin can't be translated as (apply #'member 'a '(a b c)), because it will raise an error (as is asked in Apply and keyword arguments in lisp).
What I can think about is that the value from &rest args is given to make something like (apply #member 'a '(a b c) '()) not to raise an error.
apply expects its final argument to be a list. That is,
(apply #'foo (list 1 2 3)) == (foo 1 2 3)
(apply #'member 'a '(a b c) :test #'eq) == ??? ; is an error - #'eq isn't a list
I don't know what apply is making of #'eq (a function) where a list is expected, but that's the problem.
You might be looking for funcall instead of apply:
(funcall #'foo 1 2 3) == (foo 1 2 3)
(funcall #'member 'a '(a b c) :test #'eq) == (member 'a '(a b c) :test #'eq)
Edit: (apply #'member 'a '(a b c))
This is the same as
(member 'a 'a 'b 'c)
which of course is nonsense. Think of apply as "expanding" its last argument.
Edit 2: The our-adjoin code
(our-adjoin 'a '(a b c) :test #'eq)
;; is equivalent to
(if (apply #'member 'a '(a b c) (list :test #'eq))
lst
(cons obj lst))
;; is equivalent to
(if (member 'a '(a b c) :test #'eq) ...)
(our-adjoin 'a '(a b c))
;; is equivalent to
(if (apply #'member 'a '(a b c) (list)) ...) ; (list) == nil
;; is equivalent to
(if (member 'a '(a b c)) ...)
So your hypothesis (that the equivalent was (apply #'member 'a '(a b c) '())) is correct. (FYI, there is no difference between nil, 'nil, (), '(), and (list).)
APPLY is provided so that we can call functions with a computed argument list.
APPLY has the following syntax:
apply function &rest args+ => result*
the first parameter is a function
then several arguments, at least one, where the last argument needs to be a list

How to quote all items after comma-at them from a list

I want to expand (foo a b c d e ...) to ===> (bar 'a 'b 'c 'd 'e ...)
So far I only get this solution:
(defmacro foo (a1 &rest a2)
`(bar ',a1 '(,#a2)))
But it results in:
(foo a b c d) ===> (bar 'a '(b c d))
which is not what I want.
Does anyone have any idea?
'whatever is shorthand for (quote whatever). If you have a list of symbols like A, B, C, D, etc, and you want a list that contains the structure (bar (quote a) (quote b) (quote c) ...), you can do something like this:
`(bar ,#(mapcar (lambda (symbol) (list 'quote symbol)) symbols))

What is the difference between '(a b c) and (list 'a 'b 'c)?

I am reading "On lisp" and encountered this code (I simplified it a bit).
CL-USER> (defun foo ()
'(a b c))
FOO
CL-USER> (foo)
(A B C)
CL-USER> (nconc * '(D E))
(A B C D E)
CL-USER> (foo)
(A B C D E)
CL-USER> (defun foo ()
(list 'a 'b 'c))
STYLE-WARNING: redefining FOO in DEFUN
FOO
CL-USER> (foo)
(A B C)
CL-USER> (nconc * '(D E))
(A B C D E)
CL-USER> (foo)
(A B C)
What exactly does * mean? Is it the previous function call? Is it suitable to use in real world code?
Why does (nconc * '(D E)) change the return value of the first foo function?
I always thought (list 'a 'b 'c) and '(a b c) are the same? What is the difference?
The call to LIST creates a new list each time it is evaluated. The list literal might be placed in a read-only memory segment after compilation. A destructive update on the lists with NCONC is then problematic, possibly with undefined consequences (segmentation error, changing the literal for future references, or nothing at all).
Variables *, ** and *** are specified by the language standard and they are quite useful when testing things. They are a feature of the REPL, and so are not, and not supposed to, be useful in a "real world code".