With these emac lisp definitions given to me I need to get the correct results for (defun operand (n ast)). Currently, the first child works like it's supposed to but for the second child (operand (- n 1) (cadr ast)) gives the second child as (INT_LITERAL pos) and not the rest of the child ((INT_LITERAL pos) (77)). Not sure where to go from here. As you can see I've done some guess and testing to fix my solution but nothing has worked yet. From my understanding when my results are nil that means that frame has no parent frame but I'm not sure why it does not print out the whole operand.
(defun store (offset value alist)
"Insert the value for this offset, replacing the previous value (if any)."
(cond
((null alist) (list (cons offset value))) ; ((offset . value))
((eq offset (caar alist)) (cons (cons offset value) (cdr alist)))
(t (cons (car alist)
(store offset value (cdr alist))))
)
)
(defun lookup (offset alist)
"Return the value associated with this offset, or raise an error."
(cond
((null alist) (user-error "UNINITIALISED %s" offset) (exit))
((eq (caar alist) offset) (cdar alist))
(t (lookup offset (cdr alist)))
)
)
;;(setq a (store 1 19 (store 0 17 ())))
;; a
;; (setq a (store 2 20 a))
;; (setq a (store 1 29 a))
;; (lookup 3 ())
;; (lookup 3 a)
;;(lookup 1 a)
;;; Accessors for the various fields in an AST node
(defun position (ast)
"The position stored in an AST node"
(cadar ast)
)
(defun kind (ast)
(caar ast)
)
(defun operand (n ast)
;; Your code goes here.
(if (eq n 0)
(caadr ast) ;;first child
(operand (- n 1)(cadr ast)) ;;second child
)
)
;;(operand (- n 1)(cadr (cadr ast))) gives 77 (#o115, #x4d, ?M)
;;(operand (- n 1)(cadr ast)) gives (INT_LITERAL pos)
;;(operand (- n 1) (cadr (cddr ast))) gives nil
;;(operand (- n 1) (cdr (cadr ast))) gives nil
;; (operand (- n 1)(caddr ast)) gives nil
;;(operand (- n 1)(car ast)) gives wrong type argument listp, pos
;;(operand (- n 1)(cdr ast)) gives nil
;;cadadr, cadr, cadddr, cdadr, caddr, car, cdr
;; (setq ast '((PLUS pos) (( (VARIABLE pos) (b 1) ) ((INT_LITERAL pos) (77) ) ) ))
;; (kind ast) = PLUS
;; (position ast) = pos
;; (operand 0 at) = ((VARIABLE pos)(b 1))
;; (kind (operand 0 ast))= VARIABLE
;; (operand 1 ast)= supposed to equal ((INT_LITERAL pos) (77))
;; (kind (operand 1 ast)) = supposed to equal INT_LITERAL
Your question is not easy to follow -- I'm confident that you could pare all that code down to something far more minimal for these purposes.
At present you're calling operand recursively, but the ast data does not have the nested structure required by that recursion, and so things quickly break down.
I think you just want this?
(defun operand (n ast)
(nth n (cadr ast)))
Related
I'm writing a macro to generate codes used by another macro in Common Lisp. But I'm new at this and have difficulty in constructing a macro that takes in a list (bar1 bar2 ... barn) and produces the following codes by a loop.
`(foo
,#bar1
,#bar2
...
,#barn)
I wonder whether this can be achieved not involving implement-dependent words such as SB-IMPL::UNQUOTE-SPLICE in sbcl.
Maybe I didn't give a clear description about my problem. In fact I want to write a macro gen-case such that
(gen-case
(simple-array simple-vector)
('(dotimes ($1 $5)
(when (and (= (aref $4 $2 $1) 1) (zerop (aref $3 $1)))
$0))
'(dolist ($1 (aref $4 $2))
(when (zerop (aref $3 $1))
$0)))
objname body)
produces something like
`(case (car (type-of ,objname))
(simple-array
,#(progn
(setf temp
'(dotimes ($1 $5)
(when (and (= (aref $4 $2 $1) 1) (zerop (aref $3 $1)))
$0)))
(code-gen body)))
(simple-vector
,#(progn
(setf temp
'(dolist ($1 (aref $4 $2))
(when (zerop (aref $3 $1))
$0)))
(code-gen body))))
In general cases, the lists taken in by gen-case may contain more than two items.
I have tried
``(case (car (type-of ,,objname))
,',#(#|Some codes that produce target codes|#))
but the target codes are inserted to the quote block and thus throw an exception in the macro who calls the macro gen-case. Moreover, I have no way to insert ,# to the target codes as a straightforward insertion will cause a "comma not inside a backquote" exception.
The codes generated are part of another macro
(defmacro DSI-Layer ((obj-name tag-name) &body body)
"Data Structure Independent Layer."
(let ((temp))
(defun code-gen (c)
(if (atom c) c
(if (eq (car c) tag-name)
(let ((args (cadr c)) (codes (code-gen (cddr c))) (flag nil))
(defun gen-code (c)
(if (atom c) c
(if (eq (car c) *arg*)
(let ((n (cadr c)))
(if (zerop n) (progn (setf flag t) codes)
(nth (1- n) args)))
(let ((h (gen-code (car c))))
(if flag
(progn
(setf flag nil)
(append h (gen-code (cdr c))))
(cons h (gen-code (cdr c))))))))
(gen-code temp))
(cons (code-gen (car c)) (code-gen (cdr c))))))
`(case (car (type-of ,obj-name))
(simple-array
,#(progn
(setf temp
'(dotimes ($1 $5)
(when (and (= (aref $4 $2 $1) 1) (zerop (aref $3 $1)))
$0)))
(code-gen body)))
(simple-vector
,#(progn
(setf temp
'(dolist ($1 (aref $4 $2))
(when (zerop (aref $3 $1))
$0)))
(code-gen body))))))
and I've set up a read-macro
(defvar *arg* (make-symbol "ARG"))
(set-macro-character #\$
#'(lambda (stream char)
(declare (ignore char))
(list *arg* (read stream t nil t))))
The intention of DSI-Layer is to add a piece of code to determine the type of input parameters. For example, the codes
(defun BFS (G v)
(let* ((n (car (array-dimensions G)))
(visited (make-array n :initial-element 0))
(queue (list v))
(vl nil))
(incf (aref visited v))
(DSI-Layer (G next-vertex)
(do nil ((null queue) nil)
(setf v (pop queue)) (push v vl)
(next-vertex (i v visited G n)
(setf queue (nconc queue (list i)))
(incf (aref visited i)))))
vl))
will be converted to
(defun BFS (G v)
(let* ((n (car (array-dimensions G)))
(visited (make-array n :initial-element 0))
(queue (list v))
(vl nil))
(incf (aref visited v))
(case (car (type-of G))
(simple-array
(do nil ((null queue) nil)
(setf v (pop queue))
(push v vl)
(dotimes (i n)
(when (and (= (aref G v i) 1) (zerop (aref visited i)))
(setf queue (nconc queue (list i)))
(incf (aref visited i))))))
(simple-vector
(do nil ((null queue) nil)
(setf v (pop queue))
(push v vl)
(dolist (i (aref G v))
(when (zerop (aref visited i))
(setf queue (nconc queue (list i)))
(incf (aref visited i)))))))))
Now I just wonder that whether the DSI-Layer can be generated from another macro gen-case by passing the type names and corresponding code templates to it or not.
By the way, I don't think the specific meaning of generated codes matters in my problem. They are just treated as data.
Don't be tempted to use internal details of backquote. If you have the lists you want to append in distinct variables, simply append them:
`(foo
,#(append b1 b2 ... bn))
If you have a list of them in some single variable (for instance if they've come from an &rest or &body argument) then do something like
`(foo
,#(loop for b in bs
appending b))
I see your problem - you need it not for a function call
but for a macro-call with case.
One cannot use dynamically macros - in a safe way.
One has to use eval but it is not safe for scoping.
#tfb as well as me answered in this question for type-case
lengthily.
previous answer (wrong for this case)
No need for a macro.
`(foo
,#bar1
,#bar2
...
,#barn)
with evaluation of its result
by pure functions would be:
(apply foo (loop for bar in '(bar1 bar2 ... barn)
nconc bar))
nconc or nconcing instead of collect fuses lists together and is very useful in loop. - Ah I see my previous answerer used append btw appending - nconc nconcing however is the "destructive" form of "append". Since the local variable bar is destructed here which we don't need outside of the loop form, using the "destructive" form is safe here - and comes with a performance advantage (less elements are copied than when using append). That is why I wired my brain always to use nconc instead of append inside a loop.
Of course, if you want to get the code construct, one could do
`(foo ,#(loop for bar in list-of-lists
nconc bar))
Try it out:
`(foo ,#(loop for bar in '((1 2 3) (a b c) (:a :b :c)) nconc bar))
;; => (FOO 1 2 3 A B C :A :B :C)
The answers of all of you inspired me, and I came up with a solution to my problem. The macro
(defmacro Layer-Generator (obj-name tag-name callback body)
(let ((temp (gensym)) (code-gen (gensym)))
`(let ((,temp))
(defun ,code-gen (c)
(if (atom c) c
(if (eq (car c) ,tag-name)
(let ((args (cadr c)) (codes (,code-gen (cddr c))) (flag nil))
(defun gen-code (c)
(if (atom c) c
(if (eq (car c) *arg*)
(let ((n (cadr c)))
(if (zerop n) (progn (setf flag t) codes)
(nth (1- n) args)))
(let ((h (gen-code (car c))))
(if flag
(progn
(setf flag nil)
(append h (gen-code (cdr c))))
(cons h (gen-code (cdr c))))))))
(gen-code ,temp))
(cons (,code-gen (car c)) (,code-gen (cdr c))))))
(list 'case `(car (type-of ,,obj-name))
,#(let ((codes nil))
(dolist (item callback)
(push
`(cons ',(car item)
(progn
(setf ,temp ,(cadr item))
(,code-gen ,body)))
codes))
(nreverse codes))))))
produces codes which are not the same as DSI-Layer but produce codes coincident with what the latter produces. Because the codes
`(case (car (type-of ,obj-name))
(tag1
,#(#|codes1|#))
(tag2
,#(#|codes2|#))
...)
are equivalent to
(list 'case `(car (type-of ,obj-name))
(cons 'tag1 (#|codes1|#))
(cons 'tag2 (#|codes2|#))
...)
And now we can use a loop to generate it just as what the Layer-Generator does.
I tried to solve the twoSum Problem with primitive tools of car and cdr
Given an array of integers, return indices of the two numbers such
that they add up to a specific target.
You may assume that each input would have exactly one solution, and
you may not use the same element twice.
Example:
Given nums = [2, 7, 11, 15], target = 9,
Because nums[0] + nums[1] = 2 + 7 = 9, return [0, 1].
The idea is to take a x from nums, then check if x's complement (target -x) is member of set nums-x
The key logic is
if ((memberp complement (remove-first x nums))
then (list x complement))
Begin with a helper function try nums
(defun two-sum (nums target)
(try nums))
The main function:
(defun try (nums)
(let ((x (car nums))
(complement (- target x)))
(cond
((null x) '())
((memberp complement (remove-first x nums))
(list x complement))
(t (try (cdr nums)))
)))
Then I realize that nums in ((memberp complement (remove-first x nums)) should be stay unchanged and independent from the local scope of let.
How could get such a nums?
memberp and `remove-first'
(defun remove-first (item sequence)
(filter (lambda (x) (not (= x item)))
sequence))
(defun filter (predicate sequence)
(cond ((null sequence) nil)
((funcall predicate (car sequence))
(cons (car sequence)
(filter predicate
(cdr sequence))))
(t (filter predicate
(cdr sequence)))))
(defun memberp(item x)
(cond ((null x) 'false)
((equal item (car x)) x)
(t (memq item (cdr x)))))
Here is a simple recursive function to compute the indexes:
(defun two-sum (list target &optional (pos 0))
(if (null (cdr list))
nil
(let ((p (my-position (- target (car list)) list)))
(if p
(list pos (+ pos p))
(two-sum (cdr list) target (1+ pos))))))
(defun my-position (element list &optional (pos 0))
(cond ((null list) nil)
((eql element (car list)) pos)
(t (my-position element (cdr list) (1+ pos)))))
The function is initially called with the list and the target. The parameter pos, which initially is not passed to the function, is assigned automatically to 0, and in the subsequent calls it will be incremented by one, so that it tracks the index of the current element of the list.
The first condition checks if the list has less than two elements: if it is empty (or its cdr is empty) the result is nil since no solution is possibile (note that in Common Lisp (cdr nil) is nil).
Otherwise we compute the position of the “complement” of the number in the rest of the list (note that position is a primitive function, so I called my-position its rewriting). If the element is present, we return both pos and (+ pos p) (since the position found is relative to the current position), otherwise (my-position returns nil when no element is found) we recur on the rest of the list.
Note that with this method there is no need to consider every time all the elements of the list.
I am trying to check if a list has a mountain aspect or not in lisp.
e.g:1,5,9,6,4,3
l is my list and aux is 0-the ascending part of l or 1-the descending part of the list.
muntemain just call munte starting with aux=0,the ascending part
my error is :
Badly formed lambda: (AND (< (CAR L) (CAR (CDR L))) (EQ AUX 0))
and I can't see the problem.Can someone help please?
(defun munte (l aux)
(cond
((and (atom l) (null aux)) NIL)
((and (null l) (null aux)) NIL)
((and (atom l) (eq aux 1)) T)
((and (null l) (eq aux 1) T)
((and (< (car l) (car(cdr l))) (eq aux 0)) (munte(cdr l) 0))
((and (or (> (car l) (cadr l)) (= (car l) (cadr l))) (eq aux 0))(munte(cdr l) 1))
( and (> (car l) (cadr l)) (eq aux 1)) (munte(cdr l) 1))
(T NIL)
)
)
(defun muntemain (l)
(cond
((> (car l) (cadr l)) NIL)
((< (length l) 2) NIL)
(T (munte l 0))
)
)
Formatting
As noted by Barmar, you really need to use an editor to help you with the parenthesis. There are many tutorials for installing Emacs+Slime. Take some time to install proper tools.
Don't use EQ for numbers and characters
An implementation is permitted to make "copies" of characters and
numbers at any time. The effect is that Common Lisp makes no guarantee
that eq is true even when both its arguments are "the same thing" if
that thing is a character or number.
Factorize tests
((and (atom l) (null aux)) NIL)
((and (null l) (null aux)) NIL)
((and (atom l) (eq aux 1)) T)
((and (null l) (eq aux 1) T)
From the definition of atom, NIL is an atom, so you don't need (null L). The different cases for aux can be grouped too. The clause below is sufficient to account for all the above ones:
((atom L) (eql aux 1))
But I don't understand why aux is not a boolean in the first place if you always bind it to 0 or 1. Just use t and nil and return aux in the above clause.
Use meaningful functions
(< (car l) (car(cdr l)))
Of course, (car(cdr ..)) is known as (cadr ..), but also as second. The above test is equivalent to:
(< (first L) (second L))
And what if your list has no second element? You will compare a number against nil and signal an error (not what you want). You need more tests. In muntemain, you seem to have a special case for when length is below 2, but the test is done only if the previous returns nil, which won't happen if an error is signaled.
An iterative alternative
Here is a completely different way to attack the problem, just to give you ideas.
(lambda (list)
(loop
;; memories
for px = nil then x
for pdx = nil then dx
;; current element
for x in list
;; first and second "derivatives" (signs only)
for dx = 1 then (signum (- x px))
for ddx = 0 then (signum (- dx pdx))
;; checks
sum ddx into total
always (and (<= dx 0) (<= -1 total 0))
finally (return (= total -1))))
I'm new to lisp, and have been trying to learn Common Lisp by diving in and writing some code. I've read plenty of documentation on the subject, but it's taking a while to really sink in.
I have written a couple of macros (? and ??) for performing unit tests, but I'm having some difficulty. The code is at the end of the post, to avoid cluttering the actual question.
Here is an example of usage:
(??
(? "Arithmetic tests"
(? "Addition"
(= (+ 1 2) 3)
(= (+ 1 2 3) 6)
(= (+ -1 -3) -4))))
And an example of output:
[Arithmetic tests]
[Addition]
(PASS) '(= (+ 1 2) 3)'
(PASS) '(= (+ 1 2 3) 6)'
(PASS) '(= (+ -1 -3) -4)'
Results: 3 tests passed, 0 tests failed
Now, the existing code works. Unfortunately, the (? ...) macro is ugly, verbose, resistant to change - and I'm pretty sure also badly structured. For example, do I really have to use a list to store pieces of output code and then emit the contents at the end?
I'd like to modify the macro to permit description strings (or symbols) to optionally follow each test, whereupon it would replace the test literal in the output, thus:
(??
(? "Arithmetic tests"
(? "Addition"
(= (+ 1 2) 3) "Adding 1 and 2 results in 3"
(= (+ 1 2 3) 6)
(= (+ -1 -3) -4))))
Output:
[Arithmetic tests]
[Addition]
(PASS) Adding 1 and 2 results in 3
(PASS) '(= (+ 1 2 3) 6)'
(PASS) '(= (+ -1 -3) -4)'
But unfortunately I can't find a sensible place in the macro to insert this change. Depending on where I put it, I get errors like you're not inside a backquote expression, label is not defined or body-forms is not defined. I know what these errors mean, but I can't find a way to avoid them.
Also, I'll be wanting to handle exceptions in the test, and treat that as a failure. Currently, there is no exception handling code - the test result is merely tested against nil. Again, it is not clear how I should add this functionality.
I'm thinking that maybe this macro is over-complex, due to my inexperience in writing macros; and perhaps if I simplify it, modification will be easier. I don't really want to separate it out into several smaller macros without good reason; but maybe there's a terser way to write it?
Can anyone help me out here, please?
A complete code listing follows:
(defmacro with-gensyms ((&rest names) &body body)
`(let ,(loop for n in names collect `(,n (gensym)))
,#body))
(defmacro while (condition &body body)
`(loop while ,condition do (progn ,#body)))
(defun flatten (L)
"Converts a list to single level."
(if (null L)
nil
(if (atom (first L))
(cons (first L) (flatten (rest L)))
(append (flatten (first L)) (flatten (rest L))))))
(defun starts-with-p (str1 str2)
"Determine whether `str1` starts with `str2`"
(let ((p (search str2 str1)))
(and p (= 0 p))))
(defmacro pop-first-char (string)
`(with-gensyms (c)
(if (> (length ,string) 0)
(progn
(setf c (schar ,string 0))
(if (> (length ,string) 1)
(setf ,string (subseq ,string 1))
(setf ,string ""))))
c))
(defmacro pop-chars (string count)
`(with-gensyms (result)
(setf result ())
(dotimes (index ,count)
(push (pop-first-char ,string) result))
result))
(defun format-ansi-codes (text)
(let ((result ()))
(while (> (length text) 0)
(cond
((starts-with-p text "\\e")
(push (code-char #o33) result)
(pop-chars text 2)
)
((starts-with-p text "\\r")
(push (code-char 13) result)
(pop-chars text 2)
)
(t (push (pop-first-char text) result))
))
(setf result (nreverse result))
(coerce result 'string)))
(defun kv-lookup (values key)
"Like getf, but works with 'keys as well as :keys, in both the list and the supplied key"
(setf key (if (typep key 'cons) (nth 1 key) key))
(while values
(let ((k (pop values)) (v (pop values)))
(setf k (if (typep k 'cons) (nth 1 k) k))
(if (eql (symbol-name key) (symbol-name k))
(return v)))))
(defun make-ansi-escape (ansi-name)
(let ((ansi-codes '( :normal "\\e[00m" :white "\\e[1;37m" :light-grey "\\e[0;37m" :dark-grey "\\e[1;30m"
:red "\\e[0;31m" :light-red "\\e[1;31m" :green "\\e[0;32m" :blue "\\e[1;34m" :dark-blue "\\e[1;34m"
:cyan "\\e[1;36m" :magenta "\\e[1;35m" :yellow "\\e[0;33m"
:bg-dark-grey "\\e[100m"
:bold "\\e[1m" :underline "\\e[4m"
:start-of-line "\\r" :clear-line "\\e[2K" :move-up "\\e[1A")))
(format-ansi-codes (kv-lookup ansi-codes ansi-name))
))
(defun format-ansi-escaped-arg (out-stream arg)
(cond
((typep arg 'symbol) (format out-stream "~a" (make-ansi-escape arg)))
((typep arg 'string) (format out-stream arg))
(t (format out-stream "~a" arg))
))
(defun format-ansi-escaped (out-stream &rest args)
(while args
(let ((arg (pop args)))
(if (typep arg 'list)
(let ((first-arg (eval (first arg))))
(format out-stream first-arg (second arg))
)
(format-ansi-escaped-arg out-stream arg)
))
))
(defmacro while-pop ((var sequence &optional result-form) &rest forms)
(with-gensyms (seq)
`(let (,var)
(progn
(do () ((not ,sequence))
(setf ,var (pop ,sequence))
(progn ,#forms))
,result-form))))
(defun report-start (form)
(format t "( ) '~a'~%" form))
(defun report-result (result form)
(format-ansi-escaped t "(" (if result :green :red) `("~:[FAIL~;PASS~]" ,result) :normal `(") '~a'~%" ,form))
result)
(defmacro ? (name &body body-forms)
"Run any number of test forms, optionally nested within further (?) calls, and print the results of each test"
(with-gensyms (result indent indent-string)
(if (not body-forms)
:empty
(progn
(setf result () indent 0 indent-string " ")
(cond
((typep (first body-forms) 'integer)
(setf indent (pop body-forms))))
`(progn
(format t "~v#{~A~:*~}" ,indent ,indent-string)
(format-ansi-escaped t "[" :white ,name :normal "]~%")
(with-gensyms (test-results)
(setf test-results ())
,(while-pop (body-form body-forms `(progn ,#(nreverse result)))
(cond
( (EQL (first body-form) '?)
(push `(progn
(setf test-results (append test-results (? ',(nth 1 body-form) ,(1+ indent) ,#(nthcdr 2 body-form))))
(format t "~%")
test-results
) result)
)
(t
(push `(progn
(format t "~v#{~A~:*~}" ,(1+ indent) ,indent-string)
(report-start ',body-form)
(with-gensyms (result label)
(setf result ,body-form)
(format-ansi-escaped t :move-up :start-of-line :clear-line)
(format t "~v#{~A~:*~}" ,(1+ indent) ,indent-string)
(push (report-result result ',body-form) test-results)
test-results
)) result))))))))))
(defun ?? (&rest results)
"Run any number of tests, and print a summary afterward"
(setf results (flatten results))
(format-ansi-escaped t "~&" :white "Results: " :green `("~a test~:p passed" ,(count t results)) :normal ", "
(if (find NIL results) :red :normal) `("~a test~:p failed" ,(count NIL results))
:yellow `("~[~:;, ~:*~a test~:p not run~]" ,(count :skip results))
:brown `("~[~:;, ~:*~a empty test group~:p skipped~]" ,(count :empty results))
:normal "~%"))
For my part, the ? macro is rather technical and it's hard to follow the logic behind the formatting functions. So instead of tracking errors I'd like to suggest my own attempt, perhaps it'll be of use.
I think that actually your ?? doesn't want to evaluate anything, but rather to treat its body as individual tests or sections. If the body includes a list starting with ?, this list represents a section; other elements are test forms optionally followed by descriptions. So in my implementation ?? will be a macro, and ? will be just a symbol.
I start with wishful thinking. I suppose I can create individual tests using a function make-test-item and test sections using a function make-test-section (their implementation is unimportant for now), that I can display them using an auxiliary function display-test and compute results using the function results, which returns two values: the total number of tests and the number of passed ones. Then I'd like the code
(??
(? "Arithmetic tests"
(? "Addition"
(= (+ 1 2) 3) "Adding 1 and 2 results in 3"
(= (+ 1 2 3) 6)
(= (+ -1 -3) 4))
(? "Subtraction"
(= (- 1 2) 1)))
(= (sin 0) 0) "Sine of 0 equals 0")
to expand into something like
(let ((tests (list (make-test-section :header "Arithmetic tests"
:items (list (make-test-section :header "Addition"
:items (list (make-test-item :form '(= (+ 1 2) 3)
:description "Adding 1 and 2 results in 3"
:passp (= (+ 1 2) 3))
(make-test-item :form '(= (+ 1 2 3) 6)
:passp (= (+ 1 2 3) 6))
(make-test-item :form '(= (+ -1 -3) 4)
:passp (= (+ -1 -3) 4))))
(make-test-section :header "Subtraction"
:items (list (make-test-item :form '(= (- 1 2) 1)
:passp (= (- 1 2) 1))))))
(make-test-item :form '(= (sin 0) 0)
:passp (= (sin 0) 0)
:description "Sine of 0 equals 0"))))
(loop for test in tests
with total = 0
with passed = 0
do (display-test test 0 t)
do (multiple-value-bind (ttl p) (results test)
(incf total ttl)
(incf passed p))
finally (display-result total passed t)))
Here a list of tests is created; then we traverse it printing each test (0 denotes the zero level of indentation and t is as in format) and keeping track of the results, finally displaying the total results. I don't think explicit eval is needed here.
It may not be the most exquisite piece of code ever, but it seems manageable. I supply missing definitions below, they are rather trivial (and can be improved) and have nothing to do with macros.
Now we pass on to the macros. Consider both pieces of code as data, then we want a list processing function which would turn the first one into the second. A few auxiliary functions would come in handy.
The major task is to parse the body of ?? and generate the list of test to go inside the let.
(defun test-item-form (form description)
`(make-test-item :form ',form :description ,description :passp ,form))
(defun test-section-form (header items)
`(make-test-section :header ,header :items (list ,#items)))
(defun parse-test (forms)
(let (new-forms)
(loop
(when (null forms)
(return (nreverse new-forms)))
(let ((f (pop forms)))
(cond ((and (listp f) (eq (first f) '?))
(push (test-section-form (second f) (parse-test (nthcdr 2 f))) new-forms))
((stringp (first forms))
(push (test-item-form f (pop forms)) new-forms))
(t (push (test-item-form f nil) new-forms)))))))
Here parse-test essentially absorbs the syntax of ??. Each iteration consumes one or two forms and collects corresponding make-... forms. The functions can be easily tested in REPL (and, of course, I did test them while writing).
Now the macro becomes quite simple:
(defmacro ?? (&body body)
`(let ((tests (list ,#(parse-test body))))
(loop for test in tests
with total = 0
with passed = 0
do (display-test test 0 t)
do (multiple-value-bind (ttl p) (results test)
(incf total ttl)
(incf passed p))
finally (display-result total passed t))))
It captures a few symbols, both in the variable name space and in the function one (the expansion may contain make-test-item and make-test-section). A clean solution with gensyms would be cumbersome, so I'd suggest just moving all the definitions in a separate package and exporting only ?? and ?.
For completeness, here is an implementation of the test API. Actually, it's what I started coding with and proceeded until I made sure the big let-form works; then I passed on to the macro part. This implementation is fairly sloppy; in particular, it doesn't support terminal colours and display-test can't even output a section into a string.
(defstruct test-item form description passp)
(defstruct test-section header items)
(defun results (test)
(etypecase test
(test-item (if (test-item-passp test)
(values 1 1)
(values 1 0)))
(test-section (let ((items-count 0)
(passed-count 0))
(dolist (i (test-section-items test) (values items-count passed-count))
(multiple-value-bind (i p) (results i)
(incf items-count i)
(incf passed-count p)))))))
(defparameter *test-indent* 2)
(defun display-test-item (i level stream)
(format stream "~V,0T~:[(FAIL)~;(PASS)~] ~:['~S'~;~:*~A~]~%"
(* level *test-indent*)
(test-item-passp i)
(test-item-description i)
(test-item-form i)))
(defun display-test-section-header (s level stream)
(format stream "~V,0T[~A]~%"
(* level *test-indent*)
(test-section-header s)))
(defun display-test (test level stream)
(etypecase test
(test-item (display-test-item test level stream))
(test-section
(display-test-section-header test level stream)
(dolist (i (test-section-items test))
(display-test i (1+ level) stream)))))
(defun display-result (total passed stream)
(format stream "Results: ~D test~:P passed, ~D test~:P failed.~%" passed (- total passed)))
All the code is licenced under WTFPL.
I have a problem with some part of my lisp code. It is a sudoku table generator. It works fine until this part:
(loop for e in entries do
(if (and (not (member e sub))
(not (member e col)))
(progn (setq choices (nconc choices (list e)))
(print choices)))
(if (= (length choices) 1)
(setq pick (car choices))
(if (not (= (length choices) 0))
(setq pick (nth (random (+ 0 (length choices))) choices))))
Basically, I am on a row x and a column y, and I need to insert an element. I watch the submatrix for this element and for column, and I choose the number that isn't appearing in any of the above and put it there. That's the "pick" variable. The problem is that sometimes the "choices" variable gets NIL value although in entries loop it has the right value. When it gets NIL, the pick value stays the same as it was in last loop (I am looping in columns and rows, above this snippet), making my final table have invalidated output (double values in a row, for example). How can I track where the choices variable changes? I work with it only in this snippet and I don't understand why it changes suddenly to nil.
For instance, I usually have:
in entries loop: choices (5)
Out of entries loop: choices (5)
in entries loop: choices (6 7)
Out of entries loop: choices (6 7) and after that this one:
in entries loop: choices nil.
Thank you.
First, some reformatting:
(loop for e in entries do
(if (and (not (member e sub))
(not (member e col)))
(progn (setq choices (nconc choices (list e)))
(print choices)))
(if (= (length choices) 1)
(setq pick (car choices))
(if (not (= (length choices) 0))
(setq pick (nth (random (+ 0 (length choices))) choices))))
Then, if you don't need the alternative clause of if, but want a progn, you can use when:
(loop for e in entries do
(when (and (not (member e sub))
(not (member e col)))
(setq choices (nconc choices (list e)))
(print choices))
(if (= (length choices) 1)
(setq pick (car choices))
(if (not (= (length choices) 0))
(setq pick (nth (random (+ 0 (length choices))) choices))))
The last two if clauses are mutually exclusive, so either cond or case would be appropriate (I'll use cond for now):
(loop for e in entries do
(when (and (not (member e sub))
(not (member e col)))
(setq choices (nconc choices (list e)))
(print choices))
(cond ((= (length choices) 1)
(setq pick (car choices)))
((not (= (length choices) 0))
(setq pick (nth (random (+ 0 (length choices))) choices))))
There is a zerop predicate:
(loop for e in entries do
(when (and (not (member e sub))
(not (member e col)))
(setq choices (nconc choices (list e)))
(print choices))
(cond ((= (length choices) 1)
(setq pick (car choices)))
((not (zerop (length choices)))
(setq pick (nth (random (+ 0 (length choices))) choices))))
I don't see what adding 0 to some value should accomplish:
(loop for e in entries do
(when (and (not (member e sub))
(not (member e col)))
(setq choices (nconc choices (list e)))
(print choices))
(cond ((= (length choices) 1)
(setq pick (car choices)))
((not (zerop (length choices)))
(setq pick (nth (random (length choices)) choices))))
Unless you are sure that pick is set to a sensible default to begin with, you should perhaps have a default case (this may be one of your problems):
(loop for e in entries do
(when (and (not (member e sub))
(not (member e col)))
(setq choices (nconc choices (list e)))
(print choices))
(cond ((= (length choices) 1)
(setq pick (car choices)))
((not (zerop (length choices)))
(setq pick (nth (random (length choices)) choices)))
(t
(setq pick nil))
Instead of using setq and nconc, you can use push (this puts the new element at the start of the list, but since you pick randomly anyway, this shouldn't be a concern):
(loop for e in entries do
(when (and (not (member e sub))
(not (member e col)))
(push e choices)
(print choices))
(cond ((= (length choices) 1)
(setq pick (car choices)))
((not (zerop (length choices)))
(setq pick (nth (random (length choices)) choices)))
(t
(setq pick nil))
I suspect that at the start of this snippet, choices is supposed to be (), that you don't need choices after this snippet, and that printing choices is just for debugging, so you could do this in a different way by using remove-if and changing the condition:
(let ((choices (remove-if (lambda (e)
(or (member e sub)
(member e col)))
entries)))
(print choices)
(cond ((= (length choices) 1)
(setq pick (car choices)))
((not (zerop (length choices)))
(setq pick (nth (random (length choices)) choices)))
(t
(setq pick nil)))
If choices is printed as () now, it means that there are no choices left here, so you will have to do some backtracking then (or whatever your algorithm does when a dead end is reached).
Finally, since (length choices) can only be non-negative integers, you can use case instead of cond if you test the cases in different order:
(let ((choices (remove-if (lambda (e)
(or (member e sub)
(member e col)))
entries)))
(print choices)
(case (length choices)
(0 (setq pick nil))
(1 (setq pick (car choices)))
(otherwise (setq pick (nth (random (length choices)) choices)))))
Update by request.
As Rainer points out, this is basically the body of a pick function, so we can get rid of all the free variables. Also, instead of car, you can use the (for lists) more descriptive name first:
(defun pick (entries sub col)
(let ((choices (remove-if (lambda (e)
(or (member e sub)
(member e col)))
entries)))
(print choices)
(case (length choices)
(0 nil)
(1 (first choices))
(otherwise (nth (random (length choices)) choices)))))
This function would be defined elsewhere, and in the snippet's place, it would be called like this:
(pick entries sub col)
In order not to compute (length choices) twice, we can put that into the let (which needs to become let* for serial evaluation):
(defun pick (entries sub col)
(let* ((choices (remove-if (lambda (e)
(or (member e sub)
(member e col)))
entries))
(choices-length (length choices)))
(print choices)
(case choices-length
(0 nil)
(1 (first choices))
(otherwise (nth (random choices-length) choices)))))
A final step (really optional, but perhaps you discover that you have more sequences reducing your choices, e.g. row) would be a little generalization:
(defun pick (entries &rest exclusion-sequences)
(let* ((choices (remove-if (lambda (e)
(some #'identity
(mapcar (lambda (seq)
(member e seq))
exclusion-sequences)))
entries))
(choices-length (length choices)))
(print choices)
(case choices-length
(0 nil)
(1 (first choices))
(otherwise (nth (random choices-length) choices)))))
The call to this function is of the same shape, but you can now use any number of exclusion sequences:
(pick entries col sub row ver ima fou)
A potential source of trouble is NCONC.
nconc is destructively modifying the first list. If that is unwanted, use APPEND instead.
A second source of problem with NCONC is the use of literal lists.
Example:
(defun foo (bar) (let ((l '(1 2 3))) ...))
Here '(1 2 3) is a literal list. The effects of destructively modifying such a list is undefined in Common Lisp. Thus it should be avoided. What to do instead?
cons the list: (list 1 2 3)
copy the literal list: (copy-list l)
use non destructive operations (APPEND instead of NCONC, ...)
My Lisp is quite rusty, but I don't see any backtracking there... and I think you cannot just start putting numbers randomly and expect that they will make a proper sudoku game.
It seems that the list is nil because there are no possible options and thus is not created. You should handle that.
This is not a proper answer, but I did fix the indentation to make the code a bit more legible to myself and other answerers:
(loop for e in entries do
(if (and (not (member e sub)) (not (member e col)))
(progn (setq choices (nconc choices (list e)))
(print choices) ))
(if (= (length choices) 1) (setq pick (car choices))
(if (not (=(length choices) 0))
(setq pick (nth (random (+ 0 (length choices))) choices))))
Questions:
Is entries a list of lists? Does each list represent a row?
What are the values 'sub' and 'col' set to?