Mercurial > emacs
view lisp/=cl.el @ 28418:5b7b3b5d54df
* configure.in: Line up "--help" output a little better.
author | Ken Raeburn <raeburn@raeburn.org> |
---|---|
date | Thu, 30 Mar 2000 10:06:34 +0000 |
parents | 507f64624555 |
children |
line wrap: on
line source
;;; cl.el --- Common-Lisp extensions for GNU Emacs Lisp. ;; Copyright (C) 1987, 1988, 1989, 1992 Free Software Foundation, Inc. ;; Author: Cesar Quiroz <quiroz@cs.rochester.edu> ;; Keywords: extensions (defvar cl-version "3.0 07-February-1993") ;; This file is part of GNU Emacs. ;; GNU Emacs is distributed in the hope that it will be useful, ;; but WITHOUT ANY WARRANTY. No author or distributor ;; accepts responsibility to anyone for the consequences of using it ;; or for whether it serves any particular purpose or works at all, ;; unless he says so in writing. Refer to the GNU Emacs General Public ;; License for full details. ;; Everyone is granted permission to copy, modify and redistribute ;; GNU Emacs, but only under the conditions described in the ;; GNU Emacs General Public License. A copy of this license is ;; supposed to have been given to you along with GNU Emacs so you ;; can know your rights and responsibilities. It should be in a ;; file named COPYING. Among other things, the copyright notice ;; and this notice must be preserved on all copies. ;;; Commentary: ;;; Notes from Rob Austein on his mods ;; yaya:/usr/u/sra/cl/cl.el, 5-May-1991 16:01:34, sra ;; ;; Slightly hacked copy of cl.el 2.0 beta 27. ;; ;; Various minor performance improvements: ;; a) Don't use MAPCAR when we're going to discard its results. ;; b) Make various macros a little more clever about optimizing ;; generated code in common cases. ;; c) Fix DEFSETF to expand to the right code at compile-time. ;; d) Make various macros cleverer about generating reasonable ;; code when compiled, particularly forms like DEFSTRUCT which ;; are usually used at top-level and thus are only compiled if ;; you use Hallvard Furuseth's hacked bytecomp.el. ;; ;; New features: GETF, REMF, and REMPROP. ;; ;; Notes: ;; 1) I'm sceptical about the FBOUNDP checks in SETF. Why should ;; the SETF expansion fail because the SETF method isn't defined ;; at compile time? Lisp is going to check for a binding at run-time ;; anyway, so maybe we should just assume the user's right here. ;;;; These are extensions to Emacs Lisp that provide some form of ;;;; Common Lisp compatibility, beyond what is already built-in ;;;; in Emacs Lisp. ;;;; ;;;; When developing them, I had the code spread among several files. ;;;; This file 'cl.el' is a concatenation of those original files, ;;;; minus some declarations that became redundant. The marks between ;;;; the original files can be found easily, as they are lines that ;;;; begin with four semicolons (as this does). The names of the ;;;; original parts follow the four semicolons in uppercase, those ;;;; names are GLOBAL, SYMBOLS, LISTS, SEQUENCES, CONDITIONALS, ;;;; ITERATIONS, MULTIPLE VALUES, ARITH, SETF and DEFSTRUCT. If you ;;;; add functions to this file, you might want to put them in a place ;;;; that is compatible with the division above (or invent your own ;;;; categories). ;;;; ;;;; To compile this file, make sure you load it first. This is ;;;; because many things are implemented as macros and now that all ;;;; the files are concatenated together one cannot ensure that ;;;; declaration always precedes use. ;;;; ;;;; Bug reports, suggestions and comments, ;;;; to quiroz@cs.rochester.edu ;;;; GLOBAL ;;;; This file provides utilities and declarations that are global ;;;; to Common Lisp and so might be used by more than one of the ;;;; other libraries. Especially, I intend to keep here some ;;;; utilities that help parsing/destructuring some difficult calls. ;;;; ;;;; ;;;; Cesar Quiroz @ UofR DofCSc - Dec. 1986 ;;;; (quiroz@cs.rochester.edu) ;;; Too many pieces of the rest of this package use psetq. So it is unwise to ;;; use here anything but plain Emacs Lisp! There is a neater recursive form ;;; for the algorithm that deals with the bodies. ;;; Code: ;;; This version is due to Hallvard Furuseth (hallvard@ifi.uio.no, 6 Jul 91) (defmacro psetq (&rest args) "(psetq {VARIABLE VALUE}...): In parallel, set each VARIABLE to its VALUE. All the VALUEs are evaluated, and then all the VARIABLEs are set. Aside from order of evaluation, this is the same as `setq'." ;; check there is a reasonable number of forms (if (/= (% (length args) 2) 0) (error "Odd number of arguments to `psetq'")) (setq args (copy-sequence args)) ;for safety below (prog1 (cons 'setq args) (while (progn (if (not (symbolp (car args))) (error "`psetq' expected a symbol, found '%s'." (prin1-to-string (car args)))) (cdr (cdr args))) (setcdr args (list (list 'prog1 (nth 1 args) (cons 'setq (setq args (cdr (cdr args)))))))))) ;;; utilities ;;; ;;; pair-with-newsyms takes a list and returns a list of lists of the ;;; form (newsym form), such that a let* can then bind the evaluation ;;; of the forms to the newsyms. The idea is to guarantee correct ;;; order of evaluation of the subforms of a setf. It also returns a ;;; list of the newsyms generated, in the corresponding order. (defun pair-with-newsyms (oldforms) "PAIR-WITH-NEWSYMS OLDFORMS The top-level components of the list oldforms are paired with fresh symbols, the pairings list and the newsyms list are returned." (do ((ptr oldforms (cdr ptr)) (bindings '()) (newsyms '())) ((endp ptr) (values (nreverse bindings) (nreverse newsyms))) (let ((newsym (gentemp))) (setq bindings (cons (list newsym (car ptr)) bindings)) (setq newsyms (cons newsym newsyms))))) (defun zip-lists (evens odds) "Merge two lists EVENS and ODDS, taking elts from each list alternatingly. EVENS and ODDS are two lists. ZIP-LISTS constructs a new list, whose even numbered elements (0,2,...) come from EVENS and whose odd numbered elements (1,3,...) come from ODDS. The construction stops when the shorter list is exhausted." (do* ((p0 evens (cdr p0)) (p1 odds (cdr p1)) (even (car p0) (car p0)) (odd (car p1) (car p1)) (result '())) ((or (endp p0) (endp p1)) (nreverse result)) (setq result (cons odd (cons even result))))) (defun unzip-list (list) "Extract even and odd elements of LIST into two separate lists. The argument LIST is separated in two strands, the even and the odd numbered elements. Numbering starts with 0, so the first element belongs in EVENS. No check is made that there is an even number of elements to start with." (do* ((ptr list (cddr ptr)) (this (car ptr) (car ptr)) (next (cadr ptr) (cadr ptr)) (evens '()) (odds '())) ((endp ptr) (values (nreverse evens) (nreverse odds))) (setq evens (cons this evens)) (setq odds (cons next odds)))) (defun reassemble-argslists (argslists) "(reassemble-argslists ARGSLISTS) => a list of lists ARGSLISTS is a list of sequences. Return a list of lists, the first sublist being all the entries coming from ELT 0 of the original sublists, the next those coming from ELT 1 and so on, until the shortest list is exhausted." (let* ((minlen (apply 'min (mapcar 'length argslists))) (result '())) (dotimes (i minlen (nreverse result)) ;; capture all the elements at index i (setq result (cons (mapcar (function (lambda (sublist) (elt sublist i))) argslists) result))))) ;;; Checking that a list of symbols contains no duplicates is a common ;;; task when checking the legality of some macros. The check for 'eq ;;; pairs can be too expensive, as it is quadratic on the length of ;;; the list. I use a 4-pass, linear, counting approach. It surely ;;; loses on small lists (less than 5 elements?), but should win for ;;; larger lists. The fourth pass could be eliminated. ;;; 10 dec 1986. Emacs Lisp has no REMPROP, so I just eliminated the ;;; 4th pass. ;;; ;;; [22 April 1991, sra] REMPROP now in library, so restored 4th pass. (defun duplicate-symbols-p (list) "Find all symbols appearing more than once in LIST. Return a list of all such duplicates; `nil' if there are no duplicates." (let ((duplicates '()) ;result built here (propname (gensym)) ;we use a fresh property ) ;; check validity (unless (and (listp list) (every 'symbolp list)) (error "a list of symbols is needed")) ;; pass 1: mark (dolist (x list) (put x propname 0)) ;; pass 2: count (dolist (x list) (put x propname (1+ (get x propname)))) ;; pass 3: collect (dolist (x list) (if (> (get x propname) 1) (setq duplicates (cons x duplicates)))) ;; pass 4: unmark. (dolist (x list) (remprop x propname)) ;; return result duplicates)) ;;;; end of cl-global.el ;;;; SYMBOLS ;;;; This file provides the gentemp function, which generates fresh ;;;; symbols, plus some other minor Common Lisp symbol tools. ;;;; ;;;; Cesar Quiroz @ UofR DofCSc - Dec. 1986 ;;;; (quiroz@cs.rochester.edu) ;;; Keywords. There are no packages in Emacs Lisp, so this is only a ;;; kludge around to let things be "as if" a keyword package was around. (defmacro defkeyword (x &optional docstring) "Make symbol X a keyword (symbol whose value is itself). Optional second argument is a documentation string for it." (cond ((symbolp x) (list 'defconst x (list 'quote x) docstring)) (t (error "`%s' is not a symbol" (prin1-to-string x))))) (defun keywordp (sym) "t if SYM is a keyword." (if (and (symbolp sym) (char-equal (aref (symbol-name sym) 0) ?\:)) ;; looks like one, make sure value is right (set sym sym) nil)) (defun keyword-of (sym) "Return a keyword that is naturally associated with symbol SYM. If SYM is keyword, the value is SYM. Otherwise it is a keyword whose name is `:' followed by SYM's name." (cond ((keywordp sym) sym) ((symbolp sym) (let ((newsym (intern (concat ":" (symbol-name sym))))) (set newsym newsym))) (t (error "expected a symbol, not `%s'" (prin1-to-string sym))))) ;;; Temporary symbols. ;;; (defvar *gentemp-index* 0 "Integer used by gentemp to produce new names.") (defvar *gentemp-prefix* "T$$_" "Names generated by gentemp begin with this string by default.") (defun gentemp (&optional prefix oblist) "Generate a fresh interned symbol. There are 2 optional arguments, PREFIX and OBLIST. PREFIX is the string that begins the new name, OBLIST is the obarray used to search for old names. The defaults are just right, YOU SHOULD NEVER NEED THESE ARGUMENTS IN YOUR OWN CODE." (if (null prefix) (setq prefix *gentemp-prefix*)) (if (null oblist) (setq oblist obarray)) ;default for the intern functions (let ((newsymbol nil) (newname)) (while (not newsymbol) (setq newname (concat prefix *gentemp-index*)) (setq *gentemp-index* (+ *gentemp-index* 1)) (if (not (intern-soft newname oblist)) (setq newsymbol (intern newname oblist)))) newsymbol)) (defvar *gensym-index* 0 "Integer used by gensym to produce new names.") (defvar *gensym-prefix* "G$$_" "Names generated by gensym begin with this string by default.") (defun gensym (&optional prefix) "Generate a fresh uninterned symbol. There is an optional argument, PREFIX. PREFIX is the string that begins the new name. Most people take just the default, except when debugging needs suggest otherwise." (if (null prefix) (setq prefix *gensym-prefix*)) (let ((newsymbol nil) (newname "")) (while (not newsymbol) (setq newname (concat prefix *gensym-index*)) (setq *gensym-index* (+ *gensym-index* 1)) (if (not (intern-soft newname)) (setq newsymbol (make-symbol newname)))) newsymbol)) ;;;; end of cl-symbols.el ;;;; CONDITIONALS ;;;; This file provides some of the conditional constructs of ;;;; Common Lisp. Total compatibility is again impossible, as the ;;;; 'if' form is different in both languages, so only a good ;;;; approximation is desired. ;;;; ;;;; Cesar Quiroz @ UofR DofCSc - Dec. 1986 ;;;; (quiroz@cs.rochester.edu) ;;; indentation info (put 'case 'lisp-indent-hook 1) (put 'ecase 'lisp-indent-hook 1) (put 'when 'lisp-indent-hook 1) (put 'unless 'lisp-indent-hook 1) ;;; WHEN and UNLESS ;;; These two forms are simplified ifs, with a single branch. (defmacro when (condition &rest body) "(when CONDITION . BODY) => evaluate BODY if CONDITION is true." (list* 'if (list 'not condition) '() body)) (defmacro unless (condition &rest body) "(unless CONDITION . BODY) => evaluate BODY if CONDITION is false." (list* 'if condition '() body)) ;;; CASE and ECASE ;;; CASE selects among several clauses, based on the value (evaluated) ;;; of a expression and a list of (unevaluated) key values. ECASE is ;;; the same, but signals an error if no clause is activated. (defmacro case (expr &rest cases) "(case EXPR . CASES) => evals EXPR, chooses from CASES on that value. EXPR -> any form CASES -> list of clauses, non empty CLAUSE -> HEAD . BODY HEAD -> t = catch all, must be last clause -> otherwise = same as t -> nil = illegal -> atom = activated if (eql EXPR HEAD) -> list of atoms = activated if (memq EXPR HEAD) BODY -> list of forms, implicit PROGN is built around it. EXPR is evaluated only once." (let* ((newsym (gentemp)) (clauses (case-clausify cases newsym))) ;; convert case into a cond inside a let (list 'let (list (list newsym expr)) (list* 'cond (nreverse clauses))))) (defmacro ecase (expr &rest cases) "(ecase EXPR . CASES) => like `case', but error if no case fits. `t'-clauses are not allowed." (let* ((newsym (gentemp)) (clauses (case-clausify cases newsym))) ;; check that no 't clause is present. ;; case-clausify would put one such at the beginning of clauses (if (eq (caar clauses) t) (error "no clause-head should be `t' or `otherwise' for `ecase'")) ;; insert error-catching clause (setq clauses (cons (list 't (list 'error "ecase on %s = %s failed to take any branch" (list 'quote expr) (list 'prin1-to-string newsym))) clauses)) ;; generate code as usual (list 'let (list (list newsym expr)) (list* 'cond (nreverse clauses))))) (defun case-clausify (cases newsym) "CASE-CLAUSIFY CASES NEWSYM => clauses for a 'cond' Converts the CASES of a [e]case macro into cond clauses to be evaluated inside a let that binds NEWSYM. Returns the clauses in reverse order." (do* ((currentpos cases (cdr currentpos)) (nextpos (cdr cases) (cdr nextpos)) (curclause (car cases) (car currentpos)) (result '())) ((endp currentpos) result) (let ((head (car curclause)) (body (cdr curclause))) ;; construct a cond-clause according to the head (cond ((null head) (error "case clauses cannot have null heads: `%s'" (prin1-to-string curclause))) ((or (eq head 't) (eq head 'otherwise)) ;; check it is the last clause (if (not (endp nextpos)) (error "clause with `t' or `otherwise' head must be last")) ;; accept this clause as a 't' for cond (setq result (cons (cons 't body) result))) ((atom head) (setq result (cons (cons (list 'eql newsym (list 'quote head)) body) result))) ((listp head) (setq result (cons (cons (list 'memq newsym (list 'quote head)) body) result))) (t ;; catch-all for this parser (error "don't know how to parse case clause `%s'" (prin1-to-string head))))))) ;;;; end of cl-conditionals.el ;;;; ITERATIONS ;;;; This file provides simple iterative macros (a la Common Lisp) ;;;; constructed on the basis of let, let* and while, which are the ;;;; primitive binding/iteration constructs of Emacs Lisp ;;;; ;;;; The Common Lisp iterations use to have a block named nil ;;;; wrapped around them, and allow declarations at the beginning ;;;; of their bodies and you can return a value using (return ...). ;;;; Nothing of the sort exists in Emacs Lisp, so I haven't tried ;;;; to imitate these behaviors. ;;;; ;;;; Other than the above, the semantics of Common Lisp are ;;;; correctly reproduced to the extent this was reasonable. ;;;; ;;;; Cesar Quiroz @ UofR DofCSc - Dec. 1986 ;;;; (quiroz@cs.rochester.edu) ;;; some lisp-indentation information (put 'do 'lisp-indent-hook 2) (put 'do* 'lisp-indent-hook 2) (put 'dolist 'lisp-indent-hook 1) (put 'dotimes 'lisp-indent-hook 1) (put 'do-symbols 'lisp-indent-hook 1) (put 'do-all-symbols 'lisp-indent-hook 1) (defmacro do (stepforms endforms &rest body) "(do STEPFORMS ENDFORMS . BODY): Iterate BODY, stepping some local variables. STEPFORMS must be a list of symbols or lists. In the second case, the lists must start with a symbol and contain up to two more forms. In the STEPFORMS, a symbol is the same as a (symbol). The other 2 forms are the initial value (def. NIL) and the form to step (def. itself). The values used by initialization and stepping are computed in parallel. The ENDFORMS are a list (CONDITION . ENDBODY). If the CONDITION evaluates to true in any iteration, ENDBODY is evaluated and the last form in it is returned. The BODY (which may be empty) is evaluated at every iteration, with the symbols of the STEPFORMS bound to the initial or stepped values." ;; check the syntax of the macro (and (check-do-stepforms stepforms) (check-do-endforms endforms)) ;; construct emacs-lisp equivalent (let ((initlist (extract-do-inits stepforms)) (steplist (extract-do-steps stepforms)) (endcond (car endforms)) (endbody (cdr endforms))) (cons 'let (cons initlist (cons (cons 'while (cons (list 'not endcond) (append body steplist))) (append endbody)))))) (defmacro do* (stepforms endforms &rest body) "`do*' is to `do' as `let*' is to `let'. STEPFORMS must be a list of symbols or lists. In the second case, the lists must start with a symbol and contain up to two more forms. In the STEPFORMS, a symbol is the same as a (symbol). The other 2 forms are the initial value (def. NIL) and the form to step (def. itself). Initializations and steppings are done in the sequence they are written. The ENDFORMS are a list (CONDITION . ENDBODY). If the CONDITION evaluates to true in any iteration, ENDBODY is evaluated and the last form in it is returned. The BODY (which may be empty) is evaluated at every iteration, with the symbols of the STEPFORMS bound to the initial or stepped values." ;; check the syntax of the macro (and (check-do-stepforms stepforms) (check-do-endforms endforms)) ;; construct emacs-lisp equivalent (let ((initlist (extract-do-inits stepforms)) (steplist (extract-do*-steps stepforms)) (endcond (car endforms)) (endbody (cdr endforms))) (cons 'let* (cons initlist (cons (cons 'while (cons (list 'not endcond) (append body steplist))) (append endbody)))))) ;;; DO and DO* share the syntax checking functions that follow. (defun check-do-stepforms (forms) "True if FORMS is a valid stepforms for the do[*] macro (q.v.)" (if (nlistp forms) (error "init/step form for do[*] should be a list, not `%s'" (prin1-to-string forms)) (mapcar (function (lambda (entry) (if (not (or (symbolp entry) (and (listp entry) (symbolp (car entry)) (< (length entry) 4)))) (error "init/step must be %s, not `%s'" "symbol or (symbol [init [step]])" (prin1-to-string entry))))) forms))) (defun check-do-endforms (forms) "True if FORMS is a valid endforms for the do[*] macro (q.v.)" (if (nlistp forms) (error "termination form for do macro should be a list, not `%s'" (prin1-to-string forms)))) (defun extract-do-inits (forms) "Returns a list of the initializations (for do) in FORMS --a stepforms, see the do macro--. FORMS is assumed syntactically valid." (mapcar (function (lambda (entry) (cond ((symbolp entry) (list entry nil)) ((listp entry) (list (car entry) (cadr entry)))))) forms)) ;;; There used to be a reason to deal with DO differently than with ;;; DO*. The writing of PSETQ has made it largely unnecessary. (defun extract-do-steps (forms) "EXTRACT-DO-STEPS FORMS => an s-expr FORMS is the stepforms part of a DO macro (q.v.). This function constructs an s-expression that does the stepping at the end of an iteration." (list (cons 'psetq (select-stepping-forms forms)))) (defun extract-do*-steps (forms) "EXTRACT-DO*-STEPS FORMS => an s-expr FORMS is the stepforms part of a DO* macro (q.v.). This function constructs an s-expression that does the stepping at the end of an iteration." (list (cons 'setq (select-stepping-forms forms)))) (defun select-stepping-forms (forms) "Separate only the forms that cause stepping." (let ((result '()) ;ends up being (... var form ...) (ptr forms) ;to traverse the forms entry ;to explore each form in turn ) (while ptr ;(not (endp entry)) might be safer (setq entry (car ptr)) (cond ((and (listp entry) (= (length entry) 3)) (setq result (append ;append in reverse order! (list (caddr entry) (car entry)) result)))) (setq ptr (cdr ptr))) ;step in the list of forms (nreverse result))) ;;; Other iterative constructs (defmacro dolist (stepform &rest body) "(dolist (VAR LIST [RESULTFORM]) . BODY): do BODY for each elt of LIST. The RESULTFORM defaults to nil. The VAR is bound to successive elements of the value of LIST and remains bound (to the nil value) when the RESULTFORM is evaluated." ;; check sanity (cond ((nlistp stepform) (error "stepform for `dolist' should be (VAR LIST [RESULT]), not `%s'" (prin1-to-string stepform))) ((not (symbolp (car stepform))) (error "first component of stepform should be a symbol, not `%s'" (prin1-to-string (car stepform)))) ((> (length stepform) 3) (error "too many components in stepform `%s'" (prin1-to-string stepform)))) ;; generate code (let* ((var (car stepform)) (listform (cadr stepform)) (resultform (caddr stepform)) (listsym (gentemp))) (nconc (list 'let (list var (list listsym listform)) (nconc (list 'while listsym (list 'setq var (list 'car listsym) listsym (list 'cdr listsym))) body)) (and resultform (cons (list 'setq var nil) (list resultform)))))) (defmacro dotimes (stepform &rest body) "(dotimes (VAR COUNTFORM [RESULTFORM]) . BODY): Repeat BODY, counting in VAR. The COUNTFORM should return a positive integer. The VAR is bound to successive integers from 0 to COUNTFORM-1 and the BODY is repeated for each of them. At the end, the RESULTFORM is evaluated and its value returned. During this last evaluation, the VAR is still bound, and its value is the number of times the iteration occurred. An omitted RESULTFORM defaults to nil." ;; check sanity (cond ((nlistp stepform) (error "stepform for `dotimes' should be (VAR COUNT [RESULT]), not `%s'" (prin1-to-string stepform))) ((not (symbolp (car stepform))) (error "first component of stepform should be a symbol, not `%s'" (prin1-to-string (car stepform)))) ((> (length stepform) 3) (error "too many components in stepform `%s'" (prin1-to-string stepform)))) ;; generate code (let* ((var (car stepform)) (countform (cadr stepform)) (resultform (caddr stepform)) (testsym (if (consp countform) (gentemp) countform))) (nconc (list 'let (cons (list var -1) (and (not (eq countform testsym)) (list (list testsym countform)))) (nconc (list 'while (list '< (list 'setq var (list '1+ var)) testsym)) body)) (and resultform (list resultform))))) (defmacro do-symbols (stepform &rest body) "(do_symbols (VAR [OBARRAY [RESULTFORM]]) . BODY) The VAR is bound to each of the symbols in OBARRAY (def. obarray) and the BODY is repeatedly performed for each of those bindings. At the end, RESULTFORM (def. nil) is evaluated and its value returned. During this last evaluation, the VAR is still bound and its value is nil. See also the function `mapatoms'." ;; check sanity (cond ((nlistp stepform) (error "stepform for `do-symbols' should be (VAR OBARRAY [RESULT]), not `%s'" (prin1-to-string stepform))) ((not (symbolp (car stepform))) (error "first component of stepform should be a symbol, not `%s'" (prin1-to-string (car stepform)))) ((> (length stepform) 3) (error "too many components in stepform `%s'" (prin1-to-string stepform)))) ;; generate code (let* ((var (car stepform)) (oblist (cadr stepform)) (resultform (caddr stepform))) (list 'progn (list 'mapatoms (list 'function (cons 'lambda (cons (list var) body))) oblist) (list 'let (list (list var nil)) resultform)))) (defmacro do-all-symbols (stepform &rest body) "(do-all-symbols (VAR [RESULTFORM]) . BODY) Is the same as (do-symbols (VAR obarray RESULTFORM) . BODY)." (list* 'do-symbols (list (car stepform) 'obarray (cadr stepform)) body)) (defmacro loop (&rest body) "(loop . BODY) repeats BODY indefinitely and does not return. Normally BODY uses `throw' or `signal' to cause an exit. The forms in BODY should be lists, as non-lists are reserved for new features." ;; check that the body doesn't have atomic forms (if (nlistp body) (error "body of `loop' should be a list of lists or nil") ;; ok, it is a list, check for atomic components (mapcar (function (lambda (component) (if (nlistp component) (error "components of `loop' should be lists")))) body) ;; build the infinite loop (cons 'while (cons 't body)))) ;;;; end of cl-iterations.el ;;;; LISTS ;;;; This file provides some of the lists machinery of Common-Lisp ;;;; in a way compatible with Emacs Lisp. Especially, see the the ;;;; typical c[ad]*r functions. ;;;; ;;;; Cesar Quiroz @ UofR DofCSc - Dec. 1986 ;;;; (quiroz@cs.rochester.edu) ;;; Synonyms for list functions (defsubst first (x) "Synonym for `car'" (car x)) (defsubst second (x) "Return the second element of the list LIST." (nth 1 x)) (defsubst third (x) "Return the third element of the list LIST." (nth 2 x)) (defsubst fourth (x) "Return the fourth element of the list LIST." (nth 3 x)) (defsubst fifth (x) "Return the fifth element of the list LIST." (nth 4 x)) (defsubst sixth (x) "Return the sixth element of the list LIST." (nth 5 x)) (defsubst seventh (x) "Return the seventh element of the list LIST." (nth 6 x)) (defsubst eighth (x) "Return the eighth element of the list LIST." (nth 7 x)) (defsubst ninth (x) "Return the ninth element of the list LIST." (nth 8 x)) (defsubst tenth (x) "Return the tenth element of the list LIST." (nth 9 x)) (defsubst rest (x) "Synonym for `cdr'" (cdr x)) (defsubst endp (x) "t if X is nil, nil if X is a cons; error otherwise." (if (listp x) (null x) (error "endp received a non-cons, non-null argument `%s'" (prin1-to-string x)))) (defun last (x) "Returns the last link in the list LIST." (if (nlistp x) (error "arg to `last' must be a list")) (do ((current-cons x (cdr current-cons)) (next-cons (cdr x) (cdr next-cons))) ((endp next-cons) current-cons))) (defun list-length (x) ;taken from CLtL sect. 15.2 "Returns the length of a non-circular list, or `nil' for a circular one." (do ((n 0) ;counter (fast x (cddr fast)) ;fast pointer, leaps by 2 (slow x (cdr slow)) ;slow pointer, leaps by 1 (ready nil)) ;indicates termination (ready n) (cond ((endp fast) (setq ready t)) ;return n ((endp (cdr fast)) (setq n (+ n 1)) (setq ready t)) ;return n+1 ((and (eq fast slow) (> n 0)) (setq n nil) (setq ready t)) ;return nil (t (setq n (+ n 2)))))) ;just advance counter (defun butlast (list &optional n) "Return a new list like LIST but sans the last N elements. N defaults to 1. If the list doesn't have N elements, nil is returned." (if (null n) (setq n 1)) (nreverse (nthcdr n (reverse list)))) ;optim. due to macrakis@osf.org ;;; This version due to Aaron Larson (alarson@src.honeywell.com, 26 Jul 91) (defun list* (arg &rest others) "Return a new list containing the first arguments consed onto the last arg. Thus, (list* 1 2 3 '(a b)) returns (1 2 3 a b)." (if (null others) arg (let* ((others (cons arg (copy-sequence others))) (a others)) (while (cdr (cdr a)) (setq a (cdr a))) (setcdr a (car (cdr a))) others))) (defun adjoin (item list) "Return a list which contains ITEM but is otherwise like LIST. If ITEM occurs in LIST, the value is LIST. Otherwise it is (cons ITEM LIST). When comparing ITEM against elements, `eql' is used." (if (memq item list) list (cons item list))) (defun ldiff (list sublist) "Return a new list like LIST but sans SUBLIST. SUBLIST must be one of the links in LIST; otherwise the value is LIST itself." (do ((result '()) (curcons list (cdr curcons))) ((or (endp curcons) (eq curcons sublist)) (reverse result)) (setq result (cons (car curcons) result)))) ;;; The popular c[ad]*r functions and other list accessors. ;;; To implement this efficiently, a new byte compile handler is used to ;;; generate the minimal code, saving one function call. (defsubst caar (X) "Return the car of the car of X." (car (car X))) (defsubst cadr (X) "Return the car of the cdr of X." (car (cdr X))) (defsubst cdar (X) "Return the cdr of the car of X." (cdr (car X))) (defsubst cddr (X) "Return the cdr of the cdr of X." (cdr (cdr X))) (defsubst caaar (X) "Return the car of the car of the car of X." (car (car (car X)))) (defsubst caadr (X) "Return the car of the car of the cdr of X." (car (car (cdr X)))) (defsubst cadar (X) "Return the car of the cdr of the car of X." (car (cdr (car X)))) (defsubst cdaar (X) "Return the cdr of the car of the car of X." (cdr (car (car X)))) (defsubst caddr (X) "Return the car of the cdr of the cdr of X." (car (cdr (cdr X)))) (defsubst cdadr (X) "Return the cdr of the car of the cdr of X." (cdr (car (cdr X)))) (defsubst cddar (X) "Return the cdr of the cdr of the car of X." (cdr (cdr (car X)))) (defsubst cdddr (X) "Return the cdr of the cdr of the cdr of X." (cdr (cdr (cdr X)))) (defsubst caaaar (X) "Return the car of the car of the car of the car of X." (car (car (car (car X))))) (defsubst caaadr (X) "Return the car of the car of the car of the cdr of X." (car (car (car (cdr X))))) (defsubst caadar (X) "Return the car of the car of the cdr of the car of X." (car (car (cdr (car X))))) (defsubst cadaar (X) "Return the car of the cdr of the car of the car of X." (car (cdr (car (car X))))) (defsubst cdaaar (X) "Return the cdr of the car of the car of the car of X." (cdr (car (car (car X))))) (defsubst caaddr (X) "Return the car of the car of the cdr of the cdr of X." (car (car (cdr (cdr X))))) (defsubst cadadr (X) "Return the car of the cdr of the car of the cdr of X." (car (cdr (car (cdr X))))) (defsubst cdaadr (X) "Return the cdr of the car of the car of the cdr of X." (cdr (car (car (cdr X))))) (defsubst caddar (X) "Return the car of the cdr of the cdr of the car of X." (car (cdr (cdr (car X))))) (defsubst cdadar (X) "Return the cdr of the car of the cdr of the car of X." (cdr (car (cdr (car X))))) (defsubst cddaar (X) "Return the cdr of the cdr of the car of the car of X." (cdr (cdr (car (car X))))) (defsubst cadddr (X) "Return the car of the cdr of the cdr of the cdr of X." (car (cdr (cdr (cdr X))))) (defsubst cddadr (X) "Return the cdr of the cdr of the car of the cdr of X." (cdr (cdr (car (cdr X))))) (defsubst cdaddr (X) "Return the cdr of the car of the cdr of the cdr of X." (cdr (car (cdr (cdr X))))) (defsubst cdddar (X) "Return the cdr of the cdr of the cdr of the car of X." (cdr (cdr (cdr (car X))))) (defsubst cddddr (X) "Return the cdr of the cdr of the cdr of the cdr of X." (cdr (cdr (cdr (cdr X))))) ;;; some inverses of the accessors are needed for setf purposes (defsubst setnth (n list newval) "Set (nth N LIST) to NEWVAL. Returns NEWVAL." (rplaca (nthcdr n list) newval)) (defun setnthcdr (n list newval) "(setnthcdr N LIST NEWVAL) => NEWVAL As a side effect, sets the Nth cdr of LIST to NEWVAL." (when (< n 0) (error "N must be 0 or greater, not %d" n)) (while (> n 0) (setq list (cdr list) n (- n 1))) ;; here only if (zerop n) (rplaca list (car newval)) (rplacd list (cdr newval)) newval) ;;; A-lists machinery (defsubst acons (key item alist) "Return a new alist with KEY paired with ITEM; otherwise like ALIST. Does not copy ALIST." (cons (cons key item) alist)) (defun pairlis (keys data &optional alist) "Return a new alist with each elt of KEYS paired with an elt of DATA; optional 3rd arg ALIST is nconc'd at the end. KEYS and DATA must have the same length." (unless (= (length keys) (length data)) (error "keys and data should be the same length")) (do* ;;collect keys and data in front of alist ((kptr keys (cdr kptr)) ;traverses the keys (dptr data (cdr dptr)) ;traverses the data (key (car kptr) (car kptr)) ;current key (item (car dptr) (car dptr)) ;current data item (result alist)) ((endp kptr) result) (setq result (acons key item result)))) ;;;; end of cl-lists.el ;;;; SEQUENCES ;;;; Emacs Lisp provides many of the 'sequences' functionality of ;;;; Common Lisp. This file provides a few things that were left out. ;;;; (defkeyword :test "Used to designate positive (selection) tests.") (defkeyword :test-not "Used to designate negative (rejection) tests.") (defkeyword :key "Used to designate component extractions.") (defkeyword :predicate "Used to define matching of sequence components.") (defkeyword :start "Inclusive low index in sequence") (defkeyword :end "Exclusive high index in sequence") (defkeyword :start1 "Inclusive low index in first of two sequences.") (defkeyword :start2 "Inclusive low index in second of two sequences.") (defkeyword :end1 "Exclusive high index in first of two sequences.") (defkeyword :end2 "Exclusive high index in second of two sequences.") (defkeyword :count "Number of elements to affect.") (defkeyword :from-end "T when counting backwards.") (defkeyword :initial-value "For the syntax of #'reduce") (defun some (pred seq &rest moreseqs) "Test PREDICATE on each element of SEQUENCE; is it ever non-nil? Extra args are additional sequences; PREDICATE gets one arg from each sequence and we advance down all the sequences together in lock-step. A sequence means either a list or a vector." (let ((args (reassemble-argslists (list* seq moreseqs)))) (do* ((ready nil) ;flag: return when t (result nil) ;resulting value (applyval nil) ;result of applying pred once (remaining args (cdr remaining)) ;remaining argument sets (current (car remaining) ;current argument set (car remaining))) ((or ready (endp remaining)) result) (setq applyval (apply pred current)) (when applyval (setq ready t) (setq result applyval))))) (defun every (pred seq &rest moreseqs) "Test PREDICATE on each element of SEQUENCE; is it always non-nil? Extra args are additional sequences; PREDICATE gets one arg from each sequence and we advance down all the sequences together in lock-step. A sequence means either a list or a vector." (let ((args (reassemble-argslists (list* seq moreseqs)))) (do* ((ready nil) ;flag: return when t (result t) ;resulting value (applyval nil) ;result of applying pred once (remaining args (cdr remaining)) ;remaining argument sets (current (car remaining) ;current argument set (car remaining))) ((or ready (endp remaining)) result) (setq applyval (apply pred current)) (unless applyval (setq ready t) (setq result nil))))) (defun notany (pred seq &rest moreseqs) "Test PREDICATE on each element of SEQUENCE; is it always nil? Extra args are additional sequences; PREDICATE gets one arg from each sequence and we advance down all the sequences together in lock-step. A sequence means either a list or a vector." (let ((args (reassemble-argslists (list* seq moreseqs)))) (do* ((ready nil) ;flag: return when t (result t) ;resulting value (applyval nil) ;result of applying pred once (remaining args (cdr remaining)) ;remaining argument sets (current (car remaining) ;current argument set (car remaining))) ((or ready (endp remaining)) result) (setq applyval (apply pred current)) (when applyval (setq ready t) (setq result nil))))) (defun notevery (pred seq &rest moreseqs) "Test PREDICATE on each element of SEQUENCE; is it sometimes nil? Extra args are additional sequences; PREDICATE gets one arg from each sequence and we advance down all the sequences together in lock-step. A sequence means either a list or a vector." (let ((args (reassemble-argslists (list* seq moreseqs)))) (do* ((ready nil) ;flag: return when t (result nil) ;resulting value (applyval nil) ;result of applying pred once (remaining args (cdr remaining)) ;remaining argument sets (current (car remaining) ;current argument set (car remaining))) ((or ready (endp remaining)) result) (setq applyval (apply pred current)) (unless applyval (setq ready t) (setq result t))))) ;;; More sequence functions that don't need keyword arguments (defun concatenate (type &rest sequences) "(concatenate TYPE &rest SEQUENCES) => a sequence The sequence returned is of type TYPE (must be 'list, 'string, or 'vector) and contains the concatenation of the elements of all the arguments, in the order given." (let ((sequences (append sequences '(())))) (case type (list (apply (function append) sequences)) (string (apply (function concat) sequences)) (vector (apply (function vector) (apply (function append) sequences))) (t (error "type for concatenate `%s' not 'list, 'string or 'vector" (prin1-to-string type)))))) (defun map (type function &rest sequences) "(map TYPE FUNCTION &rest SEQUENCES) => a sequence The FUNCTION is called on each set of elements from the SEQUENCES \(stopping when the shortest sequence is terminated\) and the results are possibly returned in a sequence of type TYPE \(one of 'list, 'vector, 'string, or nil\) giving NIL for TYPE gets rid of the values." (if (not (memq type (list 'list 'string 'vector nil))) (error "type for map `%s' not 'list, 'string, 'vector or nil" (prin1-to-string type))) (let ((argslists (reassemble-argslists sequences)) results) (if (null type) (while argslists ;don't bother accumulating (apply function (car argslists)) (setq argslists (cdr argslists))) (setq results (mapcar (function (lambda (args) (apply function args))) argslists)) (case type (list results) (string (funcall (function concat) results)) (vector (apply (function vector) results)))))) ;;; an inverse of elt is needed for setf purposes (defun setelt (seq n newval) "In SEQUENCE, set the Nth element to NEWVAL. Returns NEWVAL. A sequence means either a list or a vector." (let ((l (length seq))) (if (or (< n 0) (>= n l)) (error "N(%d) should be between 0 and %d" n l) ;; only two cases need be considered valid, as strings are arrays (cond ((listp seq) (setnth n seq newval)) ((arrayp seq) (aset seq n newval)) (t (error "SEQ should be a sequence, not `%s'" (prin1-to-string seq))))))) ;;; Testing with keyword arguments. ;;; ;;; Many of the sequence functions use keywords to denote some stylized ;;; form of selecting entries in a sequence. The involved arguments ;;; are collected with a &rest marker (as Emacs Lisp doesn't have a &key ;;; marker), then they are passed to build-klist, who ;;; constructs an association list. That association list is used to ;;; test for satisfaction and matching. ;;; DON'T USE MEMBER, NOR ANY FUNCTION THAT COULD TAKE KEYWORDS HERE!!! (defun build-klist (argslist acceptable &optional allow-other-keys) "Decode a keyword argument list ARGSLIST for keywords in ACCEPTABLE. ARGSLIST is a list, presumably the &rest argument of a call, whose even numbered elements must be keywords. ACCEPTABLE is a list of keywords, the only ones that are truly acceptable. The result is an alist containing the arguments named by the keywords in ACCEPTABLE, or an error is signalled, if something failed. If the third argument (an optional) is non-nil, other keys are acceptable." ;; check legality of the arguments, then destructure them (unless (and (listp argslist) (evenp (length argslist))) (error "build-klist: odd number of keyword-args")) (unless (and (listp acceptable) (every 'keywordp acceptable)) (error "build-klist: second arg should be a list of keywords")) (multiple-value-bind (keywords forms) (unzip-list argslist) (unless (every 'keywordp keywords) (error "build-klist: expected keywords, found `%s'" (prin1-to-string keywords))) (unless (or allow-other-keys (every (function (lambda (keyword) (memq keyword acceptable))) keywords)) (error "bad keyword[s]: %s not in %s" (prin1-to-string (mapcan (function (lambda (keyword) (if (memq keyword acceptable) nil (list keyword)))) keywords)) (prin1-to-string acceptable))) (do* ;;pick up the pieces ((auxlist ;auxiliary a-list, may (pairlis keywords forms)) ;contain repetitions and junk (ptr acceptable (cdr ptr)) ;pointer in acceptable (this (car ptr) (car ptr)) ;current acceptable keyword (auxval nil) ;used to move values around (alist '())) ;used to build the result ((endp ptr) alist) ;; if THIS appears in auxlist, use its value (when (setq auxval (assq this auxlist)) (setq alist (cons auxval alist)))))) (defun extract-from-klist (klist key &optional default) "(extract-from-klist KLIST KEY [DEFAULT]) => value of KEY or DEFAULT Extract value associated with KEY in KLIST (return DEFAULT if nil)." (let ((retrieved (cdr (assq key klist)))) (or retrieved default))) (defun keyword-argument-supplied-p (klist key) "(keyword-argument-supplied-p KLIST KEY) => nil or something NIL if KEY (a keyword) does not appear in the KLIST." (assq key klist)) (defun add-to-klist (key item klist) "(ADD-TO-KLIST KEY ITEM KLIST) => new KLIST Add association (KEY . ITEM) to KLIST." (setq klist (acons key item klist))) (defun elt-satisfies-test-p (item elt klist) "(elt-satisfies-test-p ITEM ELT KLIST) => t or nil KLIST encodes a keyword-arguments test, as in CH. 14 of CLtL. True if the given ITEM and ELT satisfy the test." (let ((test (extract-from-klist klist :test)) (test-not (extract-from-klist klist :test-not)) (keyfn (extract-from-klist klist :key 'identity))) (cond (test (funcall test item (funcall keyfn elt))) (test-not (not (funcall test-not item (funcall keyfn elt)))) (t ;should never happen (error "neither :test nor :test-not in `%s'" (prin1-to-string klist)))))) (defun elt-satisfies-if-p (item klist) "(elt-satisfies-if-p ITEM KLIST) => t or nil True if an -if style function was called and ITEM satisfies the predicate under :predicate in KLIST." (let ((predicate (extract-from-klist klist :predicate)) (keyfn (extract-from-klist klist :key 'identity))) (funcall predicate (funcall keyfn item)))) (defun elt-satisfies-if-not-p (item klist) "(elt-satisfies-if-not-p ITEM KLIST) => t or nil KLIST encodes a keyword-arguments test, as in CH. 14 of CLtL. True if an -if-not style function was called and ITEM does not satisfy the predicate under :predicate in KLIST." (let ((predicate (extract-from-klist klist :predicate)) (keyfn (extract-from-klist klist :key 'identity))) (not (funcall predicate (funcall keyfn item))))) (defun elts-match-under-klist-p (e1 e2 klist) "(elts-match-under-klist-p E1 E2 KLIST) => t or nil KLIST encodes a keyword-arguments test, as in CH. 14 of CLtL. True if elements E1 and E2 match under the tests encoded in KLIST." (let ((test (extract-from-klist klist :test)) (test-not (extract-from-klist klist :test-not)) (keyfn (extract-from-klist klist :key 'identity))) (if (and test test-not) (error "both :test and :test-not in `%s'" (prin1-to-string klist))) (cond (test (funcall test (funcall keyfn e1) (funcall keyfn e2))) (test-not (not (funcall test-not (funcall keyfn e1) (funcall keyfn e2)))) (t ;should never happen (error "neither :test nor :test-not in `%s'" (prin1-to-string klist)))))) ;;; This macro simplifies using keyword args. It is less clumsy than using ;;; the primitives build-klist, etc... For instance, member could be written ;;; this way: ;;; (defun member (item list &rest kargs) ;;; (with-keyword-args kargs (test test-not (key 'identity)) ;;; ...)) ;;; Suggested by Robert Potter (potter@cs.rochester.edu, 15 Nov 1989) (defmacro with-keyword-args (keyargslist vardefs &rest body) "(WITH-KEYWORD-ARGS KEYARGSLIST VARDEFS . BODY) KEYARGSLIST can be either a symbol or a list of one or two symbols. In the second case, the second symbol is either T or NIL, indicating whether keywords other than the mentioned ones are tolerable. VARDEFS is a list. Each entry is either a VAR (symbol) or matches \(VAR [DEFAULT [KEYWORD]]). Just giving VAR is the same as giving \(VAR nil :VAR). The BODY is executed in an environment where each VAR (a symbol) is bound to the value present in the KEYARGSLIST provided, or to the DEFAULT. The value is searched by using the keyword form of VAR (i.e., :VAR) or the optional keyword if provided. Notice that this macro doesn't distinguish between a default value given explicitly by the user and one provided by default. See also the more primitive functions build-klist, add-to-klist, extract-from-klist, keyword-argument-supplied-p, elt-satisfies-test-p, elt-satisfies-if-p, elt-satisfies-if-not-p, elts-match-under-klist-p. They provide more complete, if clumsier, control over this feature." (let (allow-other-keys) (if (listp keyargslist) (if (> (length keyargslist) 2) (error "`%s' should be SYMBOL, (SYMBOL), or (SYMBOL t-OR-nil)" (prin1-to-string keyargslist)) (setq allow-other-keys (cadr keyargslist) keyargslist (car keyargslist)) (if (not (and (symbolp keyargslist) (memq allow-other-keys '(t nil)))) (error "first subform should be SYMBOL, (SYMBOL), or (SYMBOL t-OR-nil)" ))) (if (symbolp keyargslist) (setq allow-other-keys nil) (error "first subform should be SYMBOL, (SYMBOL), or (SYMBOL t-OR-nil)"))) (let (vars defaults keywords forms (klistname (gensym "KLIST_"))) (mapcar (function (lambda (entry) (if (symbolp entry) ;defaulty case (setq entry (list entry nil (keyword-of entry)))) (let* ((l (length entry)) (v (car entry)) (d (cadr entry)) (k (caddr entry))) (if (or (< l 1) (> l 3)) (error "`%s' must match (VAR [DEFAULT [KEYWORD]])" (prin1-to-string entry))) (if (or (null v) (not (symbolp v))) (error "bad variable `%s': must be non-null symbol" (prin1-to-string v))) (setq vars (cons v vars)) (setq defaults (cons d defaults)) (if (< l 3) (setq k (keyword-of v))) (if (and (= l 3) (or (null k) (not (keywordp k)))) (error "bad keyword `%s'" (prin1-to-string k))) (setq keywords (cons k keywords)) (setq forms (cons (list v (list 'extract-from-klist klistname k d)) forms))))) vardefs) (append (list 'let* (nconc (list (list klistname (list 'build-klist keyargslist (list 'quote keywords) allow-other-keys))) (nreverse forms))) body)))) (put 'with-keyword-args 'lisp-indent-hook 1) ;;; REDUCE ;;; It is here mostly as an example of how to use KLISTs. ;;; ;;; First of all, you need to declare the keywords (done elsewhere in this ;;; file): ;;; (defkeyword :from-end "syntax of sequence functions") ;;; (defkeyword :start "syntax of sequence functions") ;;; etc... ;;; ;;; Then, you capture all the possible keyword arguments with a &rest ;;; argument. You can pass that list downward again, of course, but ;;; internally you need to parse it into a KLIST (an alist, really). One uses ;;; (build-klist REST-ARGS ACCEPTABLE-KEYWORDS [ALLOW-OTHER]). You can then ;;; test for presence by using (keyword-argument-supplied-p KLIST KEY) and ;;; extract a value with (extract-from-klist KLIST KEY [DEFAULT]). (defun reduce (function sequence &rest kargs) "Apply FUNCTION (a function of two arguments) to successive pairs of elements from SEQUENCE. Some keyword arguments are valid after FUNCTION and SEQUENCE: :from-end If non-nil, process the values backwards :initial-value If given, prefix it to the SEQUENCE. Suffix, if :from-end :start Restrict reduction to the subsequence from this index :end Restrict reduction to the subsequence BEFORE this index. If the sequence is empty and no :initial-value is given, the FUNCTION is called on zero (not two) arguments. Otherwise, if there is exactly one element in the combination of SEQUENCE and the initial value, that element is returned." (let* ((klist (build-klist kargs '(:from-end :start :end :initial-value))) (length (length sequence)) (from-end (extract-from-klist klist :from-end)) (initial-value-given (keyword-argument-supplied-p klist :initial-value)) (start (extract-from-klist kargs :start 0)) (end (extract-from-klist kargs :end length))) (setq sequence (cl$subseq-as-list sequence start end)) (if from-end (setq sequence (reverse sequence))) (if initial-value-given (setq sequence (cons (extract-from-klist klist :initial-value) sequence))) (if (null sequence) (funcall function) ;only use of 0 arguments (let* ((result (car sequence)) (sequence (cdr sequence))) (while sequence (setq result (if from-end (funcall function (car sequence) result) (funcall function result (car sequence))) sequence (cdr sequence))) result)))) (defun cl$subseq-as-list (sequence start end) "(cl$subseq-as-list SEQUENCE START END) => a list" (let ((list (append sequence nil)) (length (length sequence)) result) (if (< start 0) (error "start should be >= 0, not %d" start)) (if (> end length) (error "end should be <= %d, not %d" length end)) (if (and (zerop start) (= end length)) list (let ((i start) (vector (apply 'vector list))) (while (/= i end) (setq result (cons (elt vector i) result)) (setq i (+ i 1))) (nreverse result))))) ;;;; end of cl-sequences.el ;;;; Some functions with keyword arguments ;;;; ;;;; Both list and sequence functions are considered here together. This ;;;; doesn't fit any more with the original split of functions in files. (defun cl-member (item list &rest kargs) "Look for ITEM in LIST; return first tail of LIST the car of whose first cons cell tests the same as ITEM. Admits arguments :key, :test, and :test-not." (if (null kargs) ;treat this fast for efficiency (memq item list) (let* ((klist (build-klist kargs '(:test :test-not :key))) (test (extract-from-klist klist :test)) (testnot (extract-from-klist klist :test-not)) (key (extract-from-klist klist :key 'identity))) ;; another workaround allegedly for speed, BLAH (if (and (or (eq test 'eq) (eq test 'eql) (eq test (symbol-function 'eq)) (eq test (symbol-function 'eql))) (null testnot) (or (eq key 'identity) ;either by default or so given (eq key (function identity)) ;could this happen? (eq key (symbol-function 'identity)) ;sheer paranoia )) (memq item list) (if (and test testnot) (error ":test and :test-not both specified for member")) (if (not (or test testnot)) (setq test 'eql)) ;; final hack: remove the indirection through the function names (if testnot (if (symbolp testnot) (setq testnot (symbol-function testnot))) (if (symbolp test) (setq test (symbol-function test)))) (if (symbolp key) (setq key (symbol-function key))) ;; ok, go for it (let ((ptr list) (done nil) (result '())) (if testnot (while (not (or done (endp ptr))) (cond ((not (funcall testnot item (funcall key (car ptr)))) (setq done t) (setq result ptr))) (setq ptr (cdr ptr))) (while (not (or done (endp ptr))) (cond ((funcall test item (funcall key (car ptr))) (setq done t) (setq result ptr))) (setq ptr (cdr ptr)))) result))))) ;;;; MULTIPLE VALUES ;;;; This package approximates the behavior of the multiple-values ;;;; forms of Common Lisp. ;;;; ;;;; Cesar Quiroz @ UofR DofCSc - Dec. 1986 ;;;; (quiroz@cs.rochester.edu) ;;; Lisp indentation information (put 'multiple-value-bind 'lisp-indent-hook 2) (put 'multiple-value-setq 'lisp-indent-hook 2) (put 'multiple-value-list 'lisp-indent-hook nil) (put 'multiple-value-call 'lisp-indent-hook 1) (put 'multiple-value-prog1 'lisp-indent-hook 1) ;;; Global state of the package is kept here (defvar *mvalues-values* nil "Most recently returned multiple-values") (defvar *mvalues-count* nil "Count of multiple-values returned, or nil if the mechanism was not used") ;;; values is the standard multiple-value-return form. Must be the ;;; last thing evaluated inside a function. If the caller is not ;;; expecting multiple values, only the first one is passed. (values) ;;; is the same as no-values returned (unaware callers see nil). The ;;; alternative (values-list <list>) is just a convenient shorthand ;;; and complements multiple-value-list. (defun values (&rest val-forms) "Produce multiple values (zero or more). Each arg is one value. See also `multiple-value-bind', which is one way to examine the multiple values produced by a form. If the containing form or caller does not check specially to see multiple values, it will see only the first value." (setq *mvalues-values* val-forms) (setq *mvalues-count* (length *mvalues-values*)) (car *mvalues-values*)) (defun values-list (&optional val-forms) "Produce multiple values (zero or more). Each element of LIST is one value. This is equivalent to (apply 'values LIST)." (cond ((nlistp val-forms) (error "Argument to values-list must be a list, not `%s'" (prin1-to-string val-forms)))) (setq *mvalues-values* val-forms) (setq *mvalues-count* (length *mvalues-values*)) (car *mvalues-values*)) ;;; Callers that want to see the multiple values use these macros. (defmacro multiple-value-list (form) "Execute FORM and return a list of all the (multiple) values FORM produces. See `values' and `multiple-value-bind'." (list 'progn (list 'setq '*mvalues-count* nil) (list 'let (list (list 'it '(gensym))) (list 'set 'it form) (list 'if '*mvalues-count* (list 'copy-sequence '*mvalues-values*) (list 'progn (list 'setq '*mvalues-count* 1) (list 'setq '*mvalues-values* (list 'list (list 'symbol-value 'it))) (list 'copy-sequence '*mvalues-values*)))))) (defmacro multiple-value-call (function &rest args) "Call FUNCTION on all the values produced by the remaining arguments. (multiple-value-call '+ (values 1 2) (values 3 4)) is 10." (let* ((result (gentemp)) (arg (gentemp))) (list 'apply (list 'function (eval function)) (list 'let* (list (list result '())) (list 'dolist (list arg (list 'quote args) result) (list 'setq result (list 'append result (list 'multiple-value-list (list 'eval arg))))))))) (defmacro multiple-value-bind (vars form &rest body) "Bind VARS to the (multiple) values produced by FORM, then do BODY. VARS is a list of variables; each is bound to one of FORM's values. If FORM doesn't make enough values, the extra variables are bound to nil. (Ordinary forms produce only one value; to produce more, use `values'.) Extra values are ignored. BODY (zero or more forms) is executed with the variables bound, then the bindings are unwound." (let* ((vals (gentemp)) ;name for intermediate values (clauses (mv-bind-clausify ;convert into clauses usable vars vals))) ; in a let form (list* 'let* (cons (list vals (list 'multiple-value-list form)) clauses) body))) (defmacro multiple-value-setq (vars form) "Set VARS to the (multiple) values produced by FORM. VARS is a list of variables; each is set to one of FORM's values. If FORM doesn't make enough values, the extra variables are set to nil. (Ordinary forms produce only one value; to produce more, use `values'.) Extra values are ignored." (let* ((vals (gentemp)) ;name for intermediate values (clauses (mv-bind-clausify ;convert into clauses usable vars vals))) ; in a setq (after append). (list 'let* (list (list vals (list 'multiple-value-list form))) (cons 'setq (apply (function append) clauses))))) (defmacro multiple-value-prog1 (form &rest body) "Evaluate FORM, then BODY, then produce the same values FORM produced. Thus, (multiple-value-prog1 (values 1 2) (foobar)) produces values 1 and 2. This is like `prog1' except that `prog1' would produce only one value, which would be the first of FORM's values." (let* ((heldvalues (gentemp))) (cons 'let* (cons (list (list heldvalues (list 'multiple-value-list form))) (append body (list (list 'values-list heldvalues))))))) ;;; utility functions ;;; ;;; mv-bind-clausify makes the pairs needed to have the variables in ;;; the variable list correspond with the values returned by the form. ;;; vals is a fresh symbol that intervenes in all the bindings. (defun mv-bind-clausify (vars vals) "MV-BIND-CLAUSIFY VARS VALS => Auxiliary list Forms a list of pairs `(,(nth i vars) (nth i vals)) for i from 0 to the length of VARS (a list of symbols). VALS is just a fresh symbol." (if (or (nlistp vars) (notevery 'symbolp vars)) (error "expected a list of symbols, not `%s'" (prin1-to-string vars))) (let* ((nvars (length vars)) (clauses '())) (dotimes (n nvars clauses) (setq clauses (cons (list (nth n vars) (list 'nth n vals)) clauses))))) ;;;; end of cl-multiple-values.el ;;;; ARITH ;;;; This file provides integer arithmetic extensions. Although ;;;; Emacs Lisp doesn't really support anything but integers, that ;;;; has still to be made to look more or less standard. ;;;; ;;;; ;;;; Cesar Quiroz @ UofR DofCSc - Dec. 1986 ;;;; (quiroz@cs.rochester.edu) (defsubst plusp (number) "True if NUMBER is strictly greater than zero." (> number 0)) (defsubst minusp (number) "True if NUMBER is strictly less than zero." (< number 0)) (defsubst oddp (number) "True if INTEGER is not divisible by 2." (/= (% number 2) 0)) (defsubst evenp (number) "True if INTEGER is divisible by 2." (= (% number 2) 0)) (defsubst abs (number) "Return the absolute value of NUMBER." (if (< number 0) (- number) number)) (defsubst signum (number) "Return -1, 0 or 1 according to the sign of NUMBER." (cond ((< number 0) -1) ((> number 0) 1) (t ;exactly zero 0))) (defun gcd (&rest integers) "Return the greatest common divisor of all the arguments. The arguments must be integers. With no arguments, value is zero." (let ((howmany (length integers))) (cond ((= howmany 0) 0) ((= howmany 1) (abs (car integers))) ((> howmany 2) (apply (function gcd) (cons (gcd (nth 0 integers) (nth 1 integers)) (nthcdr 2 integers)))) (t ;howmany=2 ;; essentially the euclidean algorithm (when (zerop (* (nth 0 integers) (nth 1 integers))) (error "a zero argument is invalid for `gcd'")) (do* ((absa (abs (nth 0 integers))) ; better to operate only (absb (abs (nth 1 integers))) ;on positives. (dd (max absa absb)) ; setup correct order for the (ds (min absa absb)) ;successive divisions. ;; intermediate results (q 0) (r 0) ;; final results (done nil) ; flag: end of iterations (result 0)) ; final value (done result) (setq q (/ dd ds)) (setq r (% dd ds)) (cond ((zerop r) (setq done t) (setq result ds)) (t (setq dd ds) (setq ds r)))))))) (defun lcm (integer &rest more) "Return the least common multiple of all the arguments. The arguments must be integers and there must be at least one of them." (let ((howmany (length more)) (a integer) (b (nth 0 more)) prod ; intermediate product (yetmore (nthcdr 1 more))) (cond ((zerop howmany) (abs a)) ((> howmany 1) ; recursive case (apply (function lcm) (cons (lcm a b) yetmore))) (t ; base case, just 2 args (setq prod (* a b)) (cond ((zerop prod) 0) (t (/ (abs prod) (gcd a b)))))))) (defun isqrt (number) "Return the integer square root of NUMBER. NUMBER must not be negative. Result is largest integer less than or equal to the real square root of the argument." ;; The method used here is essentially the Newtonian iteration ;; x[n+1] <- (x[n] + Number/x[n]) / 2 ;; suitably adapted to integer arithmetic. ;; Thanks to Philippe Schnoebelen <phs@lifia.imag.fr> for suggesting the ;; termination condition. (cond ((minusp number) (error "argument to `isqrt' (%d) must not be negative" number)) ((zerop number) 0) (t ;so (>= number 0) (do* ((approx 1) ;any positive integer will do (new 0) ;init value irrelevant (done nil)) (done (if (> (* approx approx) number) (- approx 1) approx)) (setq new (/ (+ approx (/ number approx)) 2) done (or (= new approx) (= new (+ approx 1))) approx new))))) (defun cl-floor (number &optional divisor) "Divide DIVIDEND by DIVISOR, rounding toward minus infinity. DIVISOR defaults to 1. The remainder is produced as a second value." (cond ((and (null divisor) ; trivial case (numberp number)) (values number 0)) (t ; do the division (multiple-value-bind (q r s) (safe-idiv number divisor) (cond ((zerop s) (values 0 0)) ((plusp s) (values q r)) (t ;opposite-signs case (if (zerop r) (values (- q) 0) (let ((q (- (+ q 1)))) (values q (- number (* q divisor))))))))))) (defun cl-ceiling (number &optional divisor) "Divide DIVIDEND by DIVISOR, rounding toward plus infinity. DIVISOR defaults to 1. The remainder is produced as a second value." (cond ((and (null divisor) ; trivial case (numberp number)) (values number 0)) (t ; do the division (multiple-value-bind (q r s) (safe-idiv number divisor) (cond ((zerop s) (values 0 0)) ((plusp s) (values (+ q 1) (- r divisor))) (t (values (- q) (+ number (* q divisor))))))))) (defun cl-truncate (number &optional divisor) "Divide DIVIDEND by DIVISOR, rounding toward zero. DIVISOR defaults to 1. The remainder is produced as a second value." (cond ((and (null divisor) ; trivial case (numberp number)) (values number 0)) (t ; do the division (multiple-value-bind (q r s) (safe-idiv number divisor) (cond ((zerop s) (values 0 0)) ((plusp s) ;same as floor (values q r)) (t ;same as ceiling (values (- q) (+ number (* q divisor))))))))) (defun cl-round (number &optional divisor) "Divide DIVIDEND by DIVISOR, rounding to nearest integer. DIVISOR defaults to 1. The remainder is produced as a second value." (cond ((and (null divisor) ; trivial case (numberp number)) (values number 0)) (t ; do the division (multiple-value-bind (q r s) (safe-idiv number divisor) (setq r (abs r)) ;; adjust magnitudes first, and then signs (let ((other-r (- (abs divisor) r))) (cond ((> r other-r) (setq q (+ q 1))) ((and (= r other-r) (oddp q)) ;; round to even is mandatory (setq q (+ q 1)))) (setq q (* s q)) (setq r (- number (* q divisor))) (values q r)))))) ;;; These two functions access the implementation-dependent representation of ;;; the multiple value returns. (defun cl-mod (number divisor) "Return remainder of X by Y (rounding quotient toward minus infinity). That is, the remainder goes with the quotient produced by `cl-floor'. Emacs Lisp hint: If you know that both arguments are positive, use `%' instead for speed." (cl-floor number divisor) (cadr *mvalues-values*)) (defun rem (number divisor) "Return remainder of X by Y (rounding quotient toward zero). That is, the remainder goes with the quotient produced by `cl-truncate'. Emacs Lisp hint: If you know that both arguments are positive, use `%' instead for speed." (cl-truncate number divisor) (cadr *mvalues-values*)) ;;; internal utilities ;;; ;;; safe-idiv performs an integer division with positive numbers only. ;;; It is known that some machines/compilers implement weird remainder ;;; computations when working with negatives, so the idea here is to ;;; make sure we know what is coming back to the caller in all cases. ;;; Signum computation fixed by mad@math.keio.JUNET (MAEDA Atusi) (defun safe-idiv (a b) "SAFE-IDIV A B => Q R S Q=|A|/|B|, S is the sign of A/B, R is the rest A - S*Q*B." ;; (unless (and (numberp a) (numberp b)) ;; (error "arguments to `safe-idiv' must be numbers")) ;; (when (zerop b) ;; (error "cannot divide %d by zero" a)) (let* ((q (/ (abs a) (abs b))) (s (* (signum a) (signum b))) (r (- a (* s q b)))) (values q r s))) ;;;; end of cl-arith.el ;;;; SETF ;;;; This file provides the setf macro and friends. The purpose has ;;;; been modest, only the simplest defsetf forms are accepted. ;;;; Use it and enjoy. ;;;; ;;;; Cesar Quiroz @ UofR DofCSc - Dec. 1986 ;;;; (quiroz@cs.rochester.edu) (defkeyword :setf-update-fn "Property, its value is the function setf must invoke to update a generalized variable whose access form is a function call of the symbol that has this property.") (defkeyword :setf-update-doc "Property of symbols that have a `defsetf' update function on them, installed by the `defsetf' from its optional third argument.") (defmacro setf (&rest pairs) "Generalized `setq' that can set things other than variable values. A use of `setf' looks like (setf {PLACE VALUE}...). The behavior of (setf PLACE VALUE) is to access the generalized variable at PLACE and store VALUE there. It returns VALUE. If there is more than one PLACE and VALUE, each PLACE is set from its VALUE before the next PLACE is evaluated." (let ((nforms (length pairs))) ;; check the number of subforms (cond ((/= (% nforms 2) 0) (error "odd number of arguments to `setf'")) ((= nforms 0) nil) ((> nforms 2) ;; this is the recursive case (cons 'progn (do* ;collect the place-value pairs ((args pairs (cddr args)) (place (car args) (car args)) (value (cadr args) (cadr args)) (result '())) ((endp args) (nreverse result)) (setq result (cons (list 'setf place value) result))))) (t ;i.e., nforms=2 ;; this is the base case (SETF PLACE VALUE) (let* ((place (car pairs)) (value (cadr pairs)) (head nil) (updatefn nil)) ;; dispatch on the type of the PLACE (cond ((symbolp place) (list 'setq place value)) ((and (listp place) (setq head (car place)) (symbolp head) (setq updatefn (get head :setf-update-fn))) ;; dispatch on the type of update function (cond ((and (consp updatefn) (eq (car updatefn) 'lambda)) (cons 'funcall (cons (list 'function updatefn) (append (cdr place) (list value))))) ((and (symbolp updatefn) (fboundp updatefn) (let ((defn (symbol-function updatefn))) (or (subrp defn) (and (consp defn) (or (eq (car defn) 'lambda) (eq (car defn) 'macro)))))) (cons updatefn (append (cdr place) (list value)))) (t (multiple-value-bind (bindings newsyms) (pair-with-newsyms (append (cdr place) (list value))) ;; this let gets new symbols to ensure adequate ;; order of evaluation of the subforms. (list 'let bindings (cons updatefn newsyms)))))) (t (error "no `setf' update-function for `%s'" (prin1-to-string place))))))))) (defmacro defsetf (accessfn updatefn &optional docstring) "Define how `setf' works on a certain kind of generalized variable. A use of `defsetf' looks like (defsetf ACCESSFN UPDATEFN [DOCSTRING]). ACCESSFN is a symbol. UPDATEFN is a function or macro which takes one more argument than ACCESSFN does. DEFSETF defines the translation of (SETF (ACCESFN . ARGS) NEWVAL) to be a form like (UPDATEFN ARGS... NEWVAL). The function UPDATEFN must return its last arg, after performing the updating called for." ;; reject ill-formed requests. too bad one can't test for functionp ;; or macrop. (when (not (symbolp accessfn)) (error "first argument of `defsetf' must be a symbol, not `%s'" (prin1-to-string accessfn))) ;; update properties (list 'progn (list 'eval-and-compile (list 'put (list 'quote accessfn) :setf-update-fn (list 'function updatefn))) (list 'put (list 'quote accessfn) :setf-update-doc docstring) ;; any better thing to return? (list 'quote accessfn))) ;;; This section provides the "default" setfs for Common-Emacs-Lisp ;;; The user will not normally add anything to this, although ;;; defstruct will introduce new ones as a matter of fact. ;;; ;;; Apply is a special case. The Common Lisp ;;; standard makes the case of apply be useful when the user writes ;;; something like (apply #'name ...), Emacs Lisp doesn't have the # ;;; stuff, but it has (function ...). Notice that V18 includes a new ;;; apply: this file is compatible with V18 and pre-V18 Emacses. ;;; INCOMPATIBILITY: the SETF macro evaluates its arguments in the ;;; (correct) left to right sequence *before* checking for apply ;;; methods (which should really be an special case inside setf). Due ;;; to this, the lambda expression defsetf'd to apply will succeed in ;;; applying the right function even if the name was not quoted, but ;;; computed! That extension is not Common Lisp (nor is particularly ;;; useful, I think). (defsetf apply (lambda (&rest args) ;; disassemble the calling form ;; "(((quote fn) x1 x2 ... xn) val)" (function instead of quote, too) (let* ((fnform (car args)) ;functional form (applyargs (append ;arguments "to apply fnform" (apply 'list* (butlast (cdr args))) (last args))) (newupdater nil)) ; its update-fn, if any (if (and (symbolp fnform) (setq newupdater (get fnform :setf-update-fn))) (apply newupdater applyargs) (error "can't `setf' to `%s'" (prin1-to-string fnform))))) "`apply' is a special case for `setf'") (defsetf aref aset "`setf' inversion for `aref'") (defsetf nth setnth "`setf' inversion for `nth'") (defsetf nthcdr setnthcdr "`setf' inversion for `nthcdr'") (defsetf elt setelt "`setf' inversion for `elt'") (defsetf first (lambda (list val) (setnth 0 list val)) "`setf' inversion for `first'") (defsetf second (lambda (list val) (setnth 1 list val)) "`setf' inversion for `second'") (defsetf third (lambda (list val) (setnth 2 list val)) "`setf' inversion for `third'") (defsetf fourth (lambda (list val) (setnth 3 list val)) "`setf' inversion for `fourth'") (defsetf fifth (lambda (list val) (setnth 4 list val)) "`setf' inversion for `fifth'") (defsetf sixth (lambda (list val) (setnth 5 list val)) "`setf' inversion for `sixth'") (defsetf seventh (lambda (list val) (setnth 6 list val)) "`setf' inversion for `seventh'") (defsetf eighth (lambda (list val) (setnth 7 list val)) "`setf' inversion for `eighth'") (defsetf ninth (lambda (list val) (setnth 8 list val)) "`setf' inversion for `ninth'") (defsetf tenth (lambda (list val) (setnth 9 list val)) "`setf' inversion for `tenth'") (defsetf rest (lambda (list val) (setcdr list val)) "`setf' inversion for `rest'") (defsetf car setcar "Replace the car of a cons") (defsetf cdr setcdr "Replace the cdr of a cons") (defsetf caar (lambda (list val) (setcar (nth 0 list) val)) "`setf' inversion for `caar'") (defsetf cadr (lambda (list val) (setcar (cdr list) val)) "`setf' inversion for `cadr'") (defsetf cdar (lambda (list val) (setcdr (car list) val)) "`setf' inversion for `cdar'") (defsetf cddr (lambda (list val) (setcdr (cdr list) val)) "`setf' inversion for `cddr'") (defsetf caaar (lambda (list val) (setcar (caar list) val)) "`setf' inversion for `caaar'") (defsetf caadr (lambda (list val) (setcar (cadr list) val)) "`setf' inversion for `caadr'") (defsetf cadar (lambda (list val) (setcar (cdar list) val)) "`setf' inversion for `cadar'") (defsetf cdaar (lambda (list val) (setcdr (caar list) val)) "`setf' inversion for `cdaar'") (defsetf caddr (lambda (list val) (setcar (cddr list) val)) "`setf' inversion for `caddr'") (defsetf cdadr (lambda (list val) (setcdr (cadr list) val)) "`setf' inversion for `cdadr'") (defsetf cddar (lambda (list val) (setcdr (cdar list) val)) "`setf' inversion for `cddar'") (defsetf cdddr (lambda (list val) (setcdr (cddr list) val)) "`setf' inversion for `cdddr'") (defsetf caaaar (lambda (list val) (setcar (caaar list) val)) "`setf' inversion for `caaaar'") (defsetf caaadr (lambda (list val) (setcar (caadr list) val)) "`setf' inversion for `caaadr'") (defsetf caadar (lambda (list val) (setcar (cadar list) val)) "`setf' inversion for `caadar'") (defsetf cadaar (lambda (list val) (setcar (cdaar list) val)) "`setf' inversion for `cadaar'") (defsetf cdaaar (lambda (list val) (setcdr (caar list) val)) "`setf' inversion for `cdaaar'") (defsetf caaddr (lambda (list val) (setcar (caddr list) val)) "`setf' inversion for `caaddr'") (defsetf cadadr (lambda (list val) (setcar (cdadr list) val)) "`setf' inversion for `cadadr'") (defsetf cdaadr (lambda (list val) (setcdr (caadr list) val)) "`setf' inversion for `cdaadr'") (defsetf caddar (lambda (list val) (setcar (cddar list) val)) "`setf' inversion for `caddar'") (defsetf cdadar (lambda (list val) (setcdr (cadar list) val)) "`setf' inversion for `cdadar'") (defsetf cddaar (lambda (list val) (setcdr (cdaar list) val)) "`setf' inversion for `cddaar'") (defsetf cadddr (lambda (list val) (setcar (cdddr list) val)) "`setf' inversion for `cadddr'") (defsetf cddadr (lambda (list val) (setcdr (cdadr list) val)) "`setf' inversion for `cddadr'") (defsetf cdaddr (lambda (list val) (setcdr (caddr list) val)) "`setf' inversion for `cdaddr'") (defsetf cdddar (lambda (list val) (setcdr (cddar list) val)) "`setf' inversion for `cdddar'") (defsetf cddddr (lambda (list val) (setcdr (cddr list) val)) "`setf' inversion for `cddddr'") (defsetf get put "`setf' inversion for `get' is `put'") (defsetf symbol-function fset "`setf' inversion for `symbol-function' is `fset'") (defsetf symbol-plist setplist "`setf' inversion for `symbol-plist' is `setplist'") (defsetf symbol-value set "`setf' inversion for `symbol-value' is `set'") (defsetf point goto-char "To set (point) to N, use (goto-char N)") ;; how about defsetfing other Emacs forms? ;;; Modify macros ;;; ;;; It could be nice to implement define-modify-macro, but I don't ;;; think it really pays. (defmacro incf (ref &optional delta) "(incf REF [DELTA]) -> increment the g.v. REF by DELTA (default 1)" (if (null delta) (setq delta 1)) (list 'setf ref (list '+ ref delta))) (defmacro decf (ref &optional delta) "(decf REF [DELTA]) -> decrement the g.v. REF by DELTA (default 1)" (if (null delta) (setq delta 1)) (list 'setf ref (list '- ref delta))) (defmacro push (item ref) "(push ITEM REF) -> cons ITEM at the head of the g.v. REF (a list)" (list 'setf ref (list 'cons item ref))) (defmacro pushnew (item ref) "(pushnew ITEM REF): adjoin ITEM at the head of the g.v. REF (a list)" (list 'setf ref (list 'adjoin item ref))) (defmacro pop (ref) "(pop REF) -> (prog1 (car REF) (setf REF (cdr REF)))" (let ((listname (gensym))) (list 'let (list (list listname ref)) (list 'prog1 (list 'car listname) (list 'setf ref (list 'cdr listname)))))) ;;; PSETF ;;; ;;; Psetf is the generalized variable equivalent of psetq. The right ;;; hand sides are evaluated and assigned (via setf) to the left hand ;;; sides. The evaluations are done in an environment where they ;;; appear to occur in parallel. (defmacro psetf (&rest body) "(psetf {var value }...) => nil Like setf, but all the values are computed before any assignment is made." (let ((length (length body))) (cond ((/= (% length 2) 0) (error "psetf needs an even number of arguments, %d given" length)) ((null body) '()) (t (list 'prog1 nil (let ((setfs '()) (bodyforms (reverse body))) (while bodyforms (let* ((value (car bodyforms)) (place (cadr bodyforms))) (setq bodyforms (cddr bodyforms)) (if (null setfs) (setq setfs (list 'setf place value)) (setq setfs (list 'setf place (list 'prog1 value setfs)))))) setfs)))))) ;;; SHIFTF and ROTATEF ;;; (defmacro shiftf (&rest forms) "(shiftf PLACE1 PLACE2... NEWVALUE) Set PLACE1 to PLACE2, PLACE2 to PLACE3... Each PLACE is set to the old value of the following PLACE, and the last PLACE is set to the value NEWVALUE. Returns the old value of PLACE1." (unless (> (length forms) 1) (error "`shiftf' needs more than one argument")) (let ((places (butlast forms)) (newvalue (car (last forms)))) ;; the places are accessed to fresh symbols (multiple-value-bind (bindings newsyms) (pair-with-newsyms places) (list 'let bindings (cons 'setf (zip-lists places (append (cdr newsyms) (list newvalue)))) (car newsyms))))) (defmacro rotatef (&rest places) "(rotatef PLACE...) sets each PLACE to the old value of the following PLACE. The last PLACE is set to the old value of the first PLACE. Thus, the values rotate through the PLACEs. Returns nil." (if (null places) nil (multiple-value-bind (bindings newsyms) (pair-with-newsyms places) (list 'let bindings (cons 'setf (zip-lists places (append (cdr newsyms) (list (car newsyms))))) nil)))) ;;; GETF, REMF, and REMPROP ;;; (defun getf (place indicator &optional default) "Return PLACE's PROPNAME property, or DEFAULT if not present." (while (and place (not (eq (car place) indicator))) (setq place (cdr (cdr place)))) (if place (car (cdr place)) default)) (defmacro getf$setf$method (place indicator default &rest newval) "SETF method for GETF. Not for public use." (case (length newval) (0 (setq newval default default nil)) (1 (setq newval (car newval))) (t (error "Wrong number of arguments to (setf (getf ...)) form"))) (let ((psym (gentemp)) (isym (gentemp)) (vsym (gentemp))) (list 'let (list (list psym place) (list isym indicator) (list vsym newval)) (list 'while (list 'and psym (list 'not (list 'eq (list 'car psym) isym))) (list 'setq psym (list 'cdr (list 'cdr psym)))) (list 'if psym (list 'setcar (list 'cdr psym) vsym) (list 'setf place (list 'nconc place (list 'list isym newval)))) vsym))) (defsetf getf getf$setf$method) (defmacro remf (place indicator) "Remove from the property list at PLACE its PROPNAME property. Returns non-nil if and only if the property existed." (let ((psym (gentemp)) (isym (gentemp))) (list 'let (list (list psym place) (list isym indicator)) (list 'cond (list (list 'eq isym (list 'car psym)) (list 'setf place (list 'cdr (list 'cdr psym))) t) (list t (list 'setq psym (list 'cdr psym)) (list 'while (list 'and (list 'cdr psym) (list 'not (list 'eq (list 'car (list 'cdr psym)) isym))) (list 'setq psym (list 'cdr (list 'cdr psym)))) (list 'cond (list (list 'cdr psym) (list 'setcdr psym (list 'cdr (list 'cdr (list 'cdr psym)))) t))))))) (defun remprop (symbol indicator) "Remove SYMBOL's PROPNAME property, returning non-nil if it was present." (remf (symbol-plist symbol) indicator)) ;;;; STRUCTS ;;;; This file provides the structures mechanism. See the ;;;; documentation for Common-Lisp's defstruct. Mine doesn't ;;;; implement all the functionality of the standard, although some ;;;; more could be grafted if so desired. More details along with ;;;; the code. ;;;; ;;;; ;;;; Cesar Quiroz @ UofR DofCSc - Dec. 1986 ;;;; (quiroz@cs.rochester.edu) (defkeyword :include "Syntax of `defstruct'") (defkeyword :named "Syntax of `defstruct'") (defkeyword :conc-name "Syntax of `defstruct'") (defkeyword :copier "Syntax of `defstruct'") (defkeyword :predicate "Syntax of `defstruct'") (defkeyword :print-function "Syntax of `defstruct'") (defkeyword :type "Syntax of `defstruct'") (defkeyword :initial-offset "Syntax of `defstruct'") (defkeyword :structure-doc "Documentation string for a structure.") (defkeyword :structure-slotsn "Number of slots in structure") (defkeyword :structure-slots "List of the slot's names") (defkeyword :structure-indices "List of (KEYWORD-NAME . INDEX)") (defkeyword :structure-initforms "List of (KEYWORD-NAME . INITFORM)") (defkeyword :structure-includes "() or list of a symbol, that this struct includes") (defkeyword :structure-included-in "List of the structs that include this") (defmacro defstruct (&rest args) "(defstruct NAME [DOC-STRING] . SLOTS) define NAME as structure type. NAME must be a symbol, the name of the new structure. It could also be a list (NAME . OPTIONS). Each option is either a symbol, or a list of a keyword symbol taken from the list \{:conc-name, :copier, :constructor, :predicate, :include, :print-function, :type, :initial-offset\}. The meanings of these are as in CLtL, except that no BOA-constructors are provided, and the options \{:print-function, :type, :initial-offset\} are ignored quietly. All these structs are named, in the sense that their names can be used for type discrimination. The DOC-STRING is established as the `structure-doc' property of NAME. The SLOTS are one or more of the following: SYMBOL -- meaning the SYMBOL is the name of a SLOT of NAME list of SYMBOL and VALUE -- meaning that VALUE is the initial value of the slot. `defstruct' defines functions `make-NAME', `NAME-p', `copy-NAME' for the structure, and functions with the same name as the slots to access them. `setf' of the accessors sets their values." (multiple-value-bind (name options docstring slotsn slots initlist) (parse$defstruct$args args) ;; Names for the member functions come from the options. The ;; slots* stuff collects info about the slots declared explicitly. (multiple-value-bind (conc-name constructor copier predicate moreslotsn moreslots moreinits included) (parse$defstruct$options name options slots) ;; The moreslots* stuff refers to slots gained as a consequence ;; of (:include clauses). -- Oct 89: Only one :include tolerated (when (and (numberp moreslotsn) (> moreslotsn 0)) (setf slotsn (+ slotsn moreslotsn)) (setf slots (append moreslots slots)) (setf initlist (append moreinits initlist))) (unless (> slotsn 0) (error "%s needs at least one slot" (prin1-to-string name))) (let ((dups (duplicate-symbols-p slots))) (when dups (error "`%s' are duplicates" (prin1-to-string dups)))) (setq initlist (simplify$inits slots initlist)) (let (properties functions keywords accessors alterators returned) ;; compute properties of NAME (setq properties (append (list (list 'put (list 'quote name) :structure-doc docstring) (list 'put (list 'quote name) :structure-slotsn slotsn) (list 'put (list 'quote name) :structure-slots (list 'quote slots)) (list 'put (list 'quote name) :structure-initforms (list 'quote initlist)) (list 'put (list 'quote name) :structure-indices (list 'quote (extract$indices initlist)))) ;; If this definition :includes another defstruct, ;; modify both property lists. (cond (included (list (list 'put (list 'quote name) :structure-includes (list 'quote included)) (list 'pushnew (list 'quote name) (list 'get (list 'quote (car included)) :structure-included-in)))) (t (list (let ((old (gensym))) (list 'let (list (list old (list 'car (list 'get (list 'quote name) :structure-includes)))) (list 'when old (list 'put old :structure-included-in (list 'delq (list 'quote name) ;; careful with destructive ;;manipulation! (list 'append (list 'get old :structure-included-in) '()) ))))) (list 'put (list 'quote name) :structure-includes '())))) ;; If this definition used to be :included in another, warn ;; that things make break. On the other hand, the redefinition ;; may be trivial, so don't call it an error. (let ((old (gensym))) (list (list 'let (list (list old (list 'get (list 'quote name) :structure-included-in))) (list 'when old (list 'message "`%s' redefined. Should redefine `%s'?" (list 'quote name) (list 'prin1-to-string old)))))))) ;; Compute functions associated with NAME. This is not ;; handling BOA constructors yet, but here would be the place. (setq functions (list (list 'fset (list 'quote constructor) (list 'function (list 'lambda (list '&rest 'args) (list 'make$structure$instance (list 'quote name) 'args)))) (list 'fset (list 'quote copier) (list 'function 'copy-sequence)) (let ((typetag (gensym))) (list 'fset (list 'quote predicate) (list 'function (list 'lambda (list 'thing) (list 'and (list 'vectorp 'thing) (list 'let (list (list typetag (list 'elt 'thing 0))) (list 'or (list 'and (list 'eq typetag (list 'quote name)) (list '= (list 'length 'thing) (1+ slotsn))) (list 'memq typetag (list 'get (list 'quote name) :structure-included-in)))))) ))))) ;; compute accessors for NAME's slots (multiple-value-setq (accessors alterators keywords) (build$accessors$for name conc-name predicate slots slotsn)) ;; generate returned value -- not defined by the standard (setq returned (list (cons 'vector (mapcar (function (lambda (x) (list 'quote x))) (cons name slots))))) ;; generate code (cons 'progn (nconc properties functions keywords accessors alterators returned)))))) (defun parse$defstruct$args (args) "(parse$defstruct$args ARGS) => NAME OPTIONS DOCSTRING SLOTSN SLOTS INITLIST NAME=symbol, OPTIONS=list of, DOCSTRING=string, SLOTSN=count of slots, SLOTS=list of their names, INITLIST=alist (keyword . initform)." (let (name ;args=(symbol...) or ((symbol...)...) options ;args=((symbol . options) ...) (docstring "") ;args=(head docstring . slotargs) slotargs ;second or third cdr of args (slotsn 0) ;number of slots (slots '()) ;list of slot names (initlist '())) ;list of (slot keyword . initform) ;; extract name and options (cond ((symbolp (car args)) ;simple name (setq name (car args) options '())) ((and (listp (car args)) ;(name . options) (symbolp (caar args))) (setq name (caar args) options (cdar args))) (t (error "first arg to `defstruct' must be symbol or (symbol ...)"))) (setq slotargs (cdr args)) ;; is there a docstring? (when (stringp (car slotargs)) (setq docstring (car slotargs) slotargs (cdr slotargs))) ;; now for the slots (multiple-value-bind (slotsn slots initlist) (process$slots slotargs) (values name options docstring slotsn slots initlist)))) (defun process$slots (slots) "(process$slots SLOTS) => SLOTSN SLOTSLIST INITLIST Converts a list of symbols or lists of symbol and form into the last 3 values returned by PARSE$DEFSTRUCT$ARGS." (let ((slotsn (length slots)) ;number of slots slotslist ;(slot1 slot2 ...) initlist) ;((:slot1 . init1) ...) (do* ((ptr slots (cdr ptr)) (this (car ptr) (car ptr))) ((endp ptr)) (cond ((symbolp this) (setq slotslist (cons this slotslist)) (setq initlist (acons (keyword-of this) nil initlist))) ((and (listp this) (symbolp (car this))) (let ((name (car this)) (form (cadr this))) ;; this silently ignores any slot options. bad... (setq slotslist (cons name slotslist)) (setq initlist (acons (keyword-of name) form initlist)))) (t (error "slot should be symbol or (symbol ...), not `%s'" (prin1-to-string this))))) (values slotsn (nreverse slotslist) (nreverse initlist)))) (defun parse$defstruct$options (name options slots) "(parse$defstruct$options name OPTIONS SLOTS) => many values A defstruct named NAME, with options list OPTIONS, has already slots SLOTS. Parse the OPTIONS and return the updated form of the struct's slots and other information. The values returned are: CONC-NAME is the string to use as prefix/suffix in the methods, CONST is the name of the official constructor, COPIER is the name of the structure copier, PRED is the name of the type predicate, MORESLOTSN is the number of slots added by :include, MORESLOTS is the list of slots added by :include, MOREINITS is the list of initialization forms added by :include, INCLUDED is nil, or the list of the symbol added by :include" (let* ((namestring (symbol-name name)) ;; to build the return values (conc-name (concat namestring "-")) (const (intern (concat "make-" namestring))) (copier (intern (concat "copy-" namestring))) (pred (intern (concat namestring "-p"))) (moreslotsn 0) (moreslots '()) (moreinits '()) ;; auxiliaries option-head ;When an option is not a plain option-second ; keyword, it must be a list of option-rest ; the form (head second . rest) these-slotsn ;When :include is found, the these-slots ; info about the included these-inits ; structure is added here. included ;NIL or (list INCLUDED) ) ;; Values above are the defaults. Now we read the options themselves (dolist (option options) ;; 2 cases arise, as options must be a keyword or a list (cond ((keywordp option) (case option (:named ) ;ignore silently (t (error "can't recognize option `%s'" (prin1-to-string option))))) ((and (listp option) (keywordp (setq option-head (car option)))) (setq option-second (second option)) (setq option-rest (nthcdr 2 option)) (case option-head (:conc-name (setq conc-name (cond ((stringp option-second) option-second) ((null option-second) "") (t (error "`%s' is invalid as `conc-name'" (prin1-to-string option-second)))))) (:copier (setq copier (cond ((and (symbolp option-second) (null option-rest)) option-second) (t (error "can't recognize option `%s'" (prin1-to-string option)))))) (:constructor ;no BOA-constructors allowed (setq const (cond ((and (symbolp option-second) (null option-rest)) option-second) (t (error "can't recognize option `%s'" (prin1-to-string option)))))) (:predicate (setq pred (cond ((and (symbolp option-second) (null option-rest)) option-second) (t (error "can't recognize option `%s'" (prin1-to-string option)))))) (:include (unless (symbolp option-second) (error "arg to `:include' should be a symbol, not `%s'" (prin1-to-string option-second))) (setq these-slotsn (get option-second :structure-slotsn) these-slots (get option-second :structure-slots) these-inits (get option-second :structure-initforms)) (unless (and (numberp these-slotsn) (> these-slotsn 0)) (error "`%s' is not a valid structure" (prin1-to-string option-second))) (if included (error "`%s' already includes `%s', can't include `%s' too" name (car included) option-second) (push option-second included)) (multiple-value-bind (xtra-slotsn xtra-slots xtra-inits) (process$slots option-rest) (when (> xtra-slotsn 0) (dolist (xslot xtra-slots) (unless (memq xslot these-slots) (error "`%s' is not a slot of `%s'" (prin1-to-string xslot) (prin1-to-string option-second)))) (setq these-inits (append xtra-inits these-inits))) (setq moreslotsn (+ moreslotsn these-slotsn)) (setq moreslots (append these-slots moreslots)) (setq moreinits (append these-inits moreinits)))) ((:print-function :type :initial-offset) ) ;ignore silently (t (error "can't recognize option `%s'" (prin1-to-string option))))) (t (error "can't recognize option `%s'" (prin1-to-string option))))) ;; Return values found (values conc-name const copier pred moreslotsn moreslots moreinits included))) (defun simplify$inits (slots initlist) "(simplify$inits SLOTS INITLIST) => new INITLIST Removes from INITLIST - an ALIST - any shadowed bindings." (let ((result '()) ;built here key ;from the slot ) (dolist (slot slots) (setq key (keyword-of slot)) (setq result (acons key (cdr (assoc key initlist)) result))) (nreverse result))) (defun extract$indices (initlist) "(extract$indices INITLIST) => indices list Kludge. From a list of pairs (keyword . form) build a list of pairs of the form (keyword . position in list from 0). Useful to precompute some of the work of MAKE$STRUCTURE$INSTANCE." (let ((result '()) (index 0)) (dolist (entry initlist (nreverse result)) (setq result (acons (car entry) index result) index (+ index 1))))) (defun build$accessors$for (name conc-name predicate slots slotsn) "(build$accessors$for NAME PREDICATE SLOTS SLOTSN) => FSETS DEFSETFS KWDS Generate the code for accesors and defsetfs of a structure called NAME, whose slots are SLOTS. Also, establishes the keywords for the slots names." (do ((i 0 (1+ i)) (accessors '()) (alterators '()) (keywords '()) (canonic "")) ;slot name with conc-name prepended ((>= i slotsn) (values (nreverse accessors) (nreverse alterators) (nreverse keywords))) (setq canonic (intern (concat conc-name (symbol-name (nth i slots))))) (setq accessors (cons (list 'fset (list 'quote canonic) (list 'function (list 'lambda (list 'object) (list 'cond (list (list predicate 'object) (list 'aref 'object (1+ i))) (list 't (list 'error "`%s' is not a struct %s" (list 'prin1-to-string 'object) (list 'prin1-to-string (list 'quote name)))))))) accessors)) (setq alterators (cons (list 'defsetf canonic (list 'lambda (list 'object 'newval) (list 'cond (list (list predicate 'object) (list 'aset 'object (1+ i) 'newval)) (list 't (list 'error "`%s' not a `%s'" (list 'prin1-to-string 'object) (list 'prin1-to-string (list 'quote name))))))) alterators)) (setq keywords (cons (list 'defkeyword (keyword-of (nth i slots))) keywords)))) (defun make$structure$instance (name args) "(make$structure$instance NAME ARGS) => new struct NAME A struct of type NAME is created, some slots might be initialized according to ARGS (the &rest argument of MAKE-name)." (unless (symbolp name) (error "`%s' is not a possible name for a structure" (prin1-to-string name))) (let ((initforms (get name :structure-initforms)) (slotsn (get name :structure-slotsn)) (indices (get name :structure-indices)) initalist ;pairlis'd on initforms initializers ;definitive initializers ) ;; check sanity of the request (unless (and (numberp slotsn) (> slotsn 0)) (error "`%s' is not a defined structure" (prin1-to-string name))) (unless (evenp (length args)) (error "slot initializers `%s' not of even length" (prin1-to-string args))) ;; analyze the initializers provided by the call (multiple-value-bind (speckwds specvals) ;keywords and values given (unzip-list args) ; by the user ;; check that all the arguments are introduced by keywords (unless (every (function keywordp) speckwds) (error "all of the names in `%s' should be keywords" (prin1-to-string speckwds))) ;; check that all the keywords are known (dolist (kwd speckwds) (unless (numberp (cdr (assoc kwd indices))) (error "`%s' is not a valid slot name for %s" (prin1-to-string kwd) (prin1-to-string name)))) ;; update initforms (setq initalist (pairlis speckwds (do* ;;protect values from further evaluation ((ptr specvals (cdr ptr)) (val (car ptr) (car ptr)) (result '())) ((endp ptr) (nreverse result)) (setq result (cons (list 'quote val) result))) (copy-sequence initforms))) ;; compute definitive initializers (setq initializers (do* ;;gather the values of the most definitive forms ((ptr indices (cdr ptr)) (key (caar ptr) (caar ptr)) (result '())) ((endp ptr) (nreverse result)) (setq result (cons (eval (cdr (assoc key initalist))) result)))) ;; do real initialization (apply (function vector) (cons name initializers))))) ;;;; end of cl-structs.el ;;; For lisp-interaction mode, so that multiple values can be seen when passed ;;; back. Lies every now and then... (defvar - nil "form currently under evaluation") (defvar + nil "previous -") (defvar ++ nil "previous +") (defvar +++ nil "previous ++") (defvar / nil "list of values returned by +") (defvar // nil "list of values returned by ++") (defvar /// nil "list of values returned by +++") (defvar * nil "(first) value of +") (defvar ** nil "(first) value of ++") (defvar *** nil "(first) value of +++") (defun cl-eval-print-last-sexp () "Evaluate sexp before point; print value\(s\) into current buffer. If the evaled form returns multiple values, they are shown one to a line. The variables -, +, ++, +++, *, **, ***, /, //, /// have their usual meaning. It clears the multiple-value passing mechanism, and does not pass back multiple values. Use this only if you are debugging cl.el and understand well how the multiple-value stuff works, because it can be fooled into believing that multiple values have been returned when they actually haven't, for instance \(identity \(values nil 1\)\) However, even when this fails, you can trust the first printed value to be \(one of\) the returned value\(s\)." (interactive) ;; top level call, can reset mvalues (setq *mvalues-count* nil *mvalues-values* nil) (setq - (car (read-from-string (buffer-substring (let ((stab (syntax-table))) (unwind-protect (save-excursion (set-syntax-table emacs-lisp-mode-syntax-table) (forward-sexp -1) (point)) (set-syntax-table stab))) (point))))) (setq *** ** ** * * (eval -)) (setq /// // // / / *mvalues-values*) (setq +++ ++ ++ + + -) (cond ((or (null *mvalues-count*) ;mvalues mechanism not used (not (eq * (car *mvalues-values*)))) (print * (current-buffer))) ((null /) ;no values returned (terpri (current-buffer))) (t ;more than zero mvalues (terpri (current-buffer)) (mapcar (function (lambda (value) (prin1 value (current-buffer)) (terpri (current-buffer)))) /))) (setq *mvalues-count* nil ;make sure *mvalues-values* nil)) ;;;; More LISTS functions ;;;; ;;; Some mapping functions on lists, commonly useful. ;;; They take no extra sequences, to go along with Emacs Lisp's MAPCAR. (defun mapc (function list) "(MAPC FUNCTION LIST) => LIST Apply FUNCTION to each element of LIST, return LIST. Like mapcar, but called only for effect." (let ((args list)) (while args (funcall function (car args)) (setq args (cdr args)))) list) (defun maplist (function list) "(MAPLIST FUNCTION LIST) => list'ed results of FUNCTION on cdrs of LIST Apply FUNCTION to successive sublists of LIST, return the list of the results" (let ((args list) results '()) (while args (setq results (cons (funcall function args) results) args (cdr args))) (nreverse results))) (defun mapl (function list) "(MAPL FUNCTION LIST) => LIST Apply FUNCTION to successive cdrs of LIST, return LIST. Like maplist, but called only for effect." (let ((args list)) (while args (funcall function args) (setq args (cdr args))) list)) (defun mapcan (function list) "(MAPCAN FUNCTION LIST) => nconc'd results of FUNCTION on LIST Apply FUNCTION to each element of LIST, nconc the results. Beware: nconc destroys its first argument! See copy-list." (let ((args list) (results '())) (while args (setq results (nconc (funcall function (car args)) results) args (cdr args))) (nreverse results))) (defun mapcon (function list) "(MAPCON FUNCTION LIST) => nconc'd results of FUNCTION on cdrs of LIST Apply FUNCTION to successive sublists of LIST, nconc the results. Beware: nconc destroys its first argument! See copy-list." (let ((args list) (results '())) (while args (setq results (nconc (funcall function args) results) args (cdr args))) (nreverse results))) ;;; Copiers (defsubst copy-list (list) "Build a copy of LIST" (append list '())) (defun copy-tree (tree) "Build a copy of the tree of conses TREE The argument is a tree of conses, it is recursively copied down to non conses. Circularity and sharing of substructure are not necessarily preserved." (if (consp tree) (cons (copy-tree (car tree)) (copy-tree (cdr tree))) tree)) ;;; reversals, and destructive manipulations of a list's spine (defun revappend (x y) "does what (append (reverse X) Y) would, only faster" (if (endp x) y (revappend (cdr x) (cons (car x) y)))) (defun nreconc (x y) "does (nconc (nreverse X) Y) would, only faster Destructive on X, be careful." (if (endp x) y ;; reuse the first cons of x, making it point to y (nreconc (cdr x) (prog1 x (rplacd x y))))) (defun nbutlast (list &optional n) "Side-effected LIST truncated N+1 conses from the end. This is the destructive version of BUTLAST. Returns () and does not modify the LIST argument if the length of the list is not at least N." (when (null n) (setf n 1)) (let ((length (list-length list))) (cond ((null length) list) ((< length n) '()) (t (setnthcdr (- length n) list nil) list)))) ;;; Substitutions (defun subst (new old tree) "NEW replaces OLD in a copy of TREE Uses eql for the test." (subst-if new (function (lambda (x) (eql x old))) tree)) (defun subst-if-not (new test tree) "NEW replaces any subtree or leaf that fails TEST in a copy of TREE" ;; (subst-if new (function (lambda (x) (not (funcall test x)))) tree) (cond ((not (funcall test tree)) new) ((atom tree) tree) (t ;no match so far (let ((head (subst-if-not new test (car tree))) (tail (subst-if-not new test (cdr tree)))) ;; If nothing changed, return originals. Else use the new ;; components to assemble a new tree. (if (and (eql head (car tree)) (eql tail (cdr tree))) tree (cons head tail)))))) (defun subst-if (new test tree) "NEW replaces any subtree or leaf that satisfies TEST in a copy of TREE" (cond ((funcall test tree) new) ((atom tree) tree) (t ;no match so far (let ((head (subst-if new test (car tree))) (tail (subst-if new test (cdr tree)))) ;; If nothing changed, return originals. Else use the new ;; components to assemble a new tree. (if (and (eql head (car tree)) (eql tail (cdr tree))) tree (cons head tail)))))) (defun sublis (alist tree) "Use association list ALIST to modify a copy of TREE If a subtree or leaf of TREE is a key in ALIST, it is replaced by the associated value. Not exactly Common Lisp, but close in spirit and compatible with the native Emacs Lisp ASSOC, which uses EQUAL." (let ((toplevel (assoc tree alist))) (cond (toplevel ;Bingo at top (cdr toplevel)) ((atom tree) ;Give up on this tree) (t (let ((head (sublis alist (car tree))) (tail (sublis alist (cdr tree)))) (if (and (eql head (car tree)) (eql tail (cdr tree))) tree (cons head tail))))))) (defun member-if (predicate list) "PREDICATE is applied to the members of LIST. As soon as one of them returns true, that tail of the list if returned. Else NIL." (catch 'found-member-if (while (not (endp list)) (if (funcall predicate (car list)) (throw 'found-member-if list) (setq list (cdr list)))) nil)) (defun member-if-not (predicate list) "PREDICATE is applied to the members of LIST. As soon as one of them returns false, that tail of the list if returned. Else NIL." (catch 'found-member-if-not (while (not (endp list)) (if (funcall predicate (car list)) (setq list (cdr list)) (throw 'found-member-if-not list))) nil)) (defun tailp (sublist list) "(tailp SUBLIST LIST) => True if SUBLIST is a sublist of LIST." (catch 'tailp-found (while (not (endp list)) (if (eq sublist list) (throw 'tailp-found t) (setq list (cdr list)))) nil)) ;;; Suggestion of phr%widow.Berkeley.EDU@lilac.berkeley.edu (defmacro declare (&rest decls) "Ignore a Common-Lisp declaration." "declarations are ignored in this implementation") (defun proclaim (&rest decls) "Ignore a Common-Lisp proclamation." "declarations are ignored in this implementation") (defmacro the (type form) "(the TYPE FORM) macroexpands to FORM No checking is even attempted. This is just for compatibility with Common-Lisp codes." form) ;;; Due to Aaron Larson (alarson@src.honeywell.com, 26 Jul 91) (put 'progv 'common-lisp-indent-hook '(4 4 &body)) (defmacro progv (vars vals &rest body) "progv vars vals &body forms bind vars to vals then execute forms. If there are more vars than vals, the extra vars are unbound, if there are more vals than vars, the extra vals are just ignored." (` (progv$runtime (, vars) (, vals) (function (lambda () (,@ body)))))) ;;; To do this efficiently, it really needs to be a special form... (defun progv$runtime (vars vals body) (eval (let ((vars-n-vals nil) (unbind-forms nil)) (do ((r vars (cdr r)) (l vals (cdr l))) ((endp r)) (push (list (car r) (list 'quote (car l))) vars-n-vals) (if (null l) (push (` (makunbound '(, (car r)))) unbind-forms))) (` (let (, vars-n-vals) (,@ unbind-forms) (funcall '(, body))))))) (provide 'cl) ;;;; end of cl.el