/usr/share/common-lisp/source/slime/nregex.lisp is in cl-swank 2:2.10.1-3.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 | ;;;
;;; This code was written by:
;;;
;;; Lawrence E. Freil <lef@freil.com>
;;; National Science Center Foundation
;;; Augusta, Georgia 30909
;;;
;;; This program was released into the public domain on 2005-08-31.
;;; (See the slime-devel mailing list archive for details.)
;;;
;;; nregex.lisp - My 4/8/92 attempt at a Lisp based regular expression
;;; parser.
;;;
;;; This regular expression parser operates by taking a
;;; regular expression and breaking it down into a list
;;; consisting of lisp expressions and flags. The list
;;; of lisp expressions is then taken in turned into a
;;; lambda expression that can be later applied to a
;;; string argument for parsing.
;;;;
;;;; Modifications made 6 March 2001 By Chris Double (chris@double.co.nz)
;;;; to get working with Corman Lisp 1.42, add package statement and export
;;;; relevant functions.
;;;;
(in-package :cl-user)
;; Renamed to slime-nregex avoid name clashes with other versions of
;; this file. -- he
;;;; CND - 6/3/2001
(defpackage slime-nregex
(:use #:common-lisp)
(:export
#:regex
#:regex-compile
))
;;;; CND - 6/3/2001
(in-package :slime-nregex)
;;;
;;; First we create a copy of macros to help debug the beast
(eval-when (:compile-toplevel :load-toplevel :execute)
(defvar *regex-debug* nil) ; Set to nil for no debugging code
)
(defmacro info (message &rest args)
(if *regex-debug*
`(format *standard-output* ,message ,@args)))
;;;
;;; Declare the global variables for storing the paren index list.
;;;
(defvar *regex-groups* (make-array 10))
(defvar *regex-groupings* 0)
;;;
;;; Declare a simple interface for testing. You probably wouldn't want
;;; to use this interface unless you were just calling this once.
;;;
(defun regex (expression string)
"Usage: (regex <expression> <string)
This function will call regex-compile on the expression and then apply
the string to the returned lambda list."
(let ((findit (cond ((stringp expression)
(regex-compile expression))
((listp expression)
expression)))
(result nil))
(if (not (funcall (if (functionp findit)
findit
(eval `(function ,findit))) string))
(return-from regex nil))
(if (= *regex-groupings* 0)
(return-from regex t))
(dotimes (i *regex-groupings*)
(push (funcall 'subseq
string
(car (aref *regex-groups* i))
(cadr (aref *regex-groups* i)))
result))
(reverse result)))
;;;
;;; Declare some simple macros to make the code more readable.
;;;
(defvar *regex-special-chars* "?*+.()[]\\${}")
(defmacro add-exp (list)
"Add an item to the end of expression"
`(setf expression (append expression ,list)))
;;;
;;; Define a function that will take a quoted character and return
;;; what the real character should be plus how much of the source
;;; string was used. If the result is a set of characters, return an
;;; array of bits indicating which characters should be set. If the
;;; expression is one of the sub-group matches return a
;;; list-expression that will provide the match.
;;;
(defun regex-quoted (char-string &optional (invert nil))
"Usage: (regex-quoted <char-string> &optional invert)
Returns either the quoted character or a simple bit vector of bits set for
the matching values"
(let ((first (char char-string 0))
(result (char char-string 0))
(used-length 1))
(cond ((eql first #\n)
(setf result #\NewLine))
((eql first #\c)
(setf result #\Return))
((eql first #\t)
(setf result #\Tab))
((eql first #\d)
(setf result #*0000000000000000000000000000000000000000000000001111111111000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000))
((eql first #\D)
(setf result #*1111111111111111111111111111111111111111111111110000000000111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111))
((eql first #\w)
(setf result #*0000000000000000000000000000000000000000000000001111111111000000011111111111111111111111111000010111111111111111111111111110000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000))
((eql first #\W)
(setf result #*1111111111111111111111111111111111111111111111110000000000111111100000000000000000000000000111101000000000000000000000000001111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111))
((eql first #\b)
(setf result #*0000000001000000000000000000000011000000000010100000000000100000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000))
((eql first #\B)
(setf result #*1111111110111111111111111111111100111111111101011111111111011111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111))
((eql first #\s)
(setf result #*0000000001100000000000000000000010000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000))
((eql first #\S)
(setf result #*1111111110011111111111111111111101111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111))
((and (>= (char-code first) (char-code #\0))
(<= (char-code first) (char-code #\9)))
(if (and (> (length char-string) 2)
(and (>= (char-code (char char-string 1)) (char-code #\0))
(<= (char-code (char char-string 1)) (char-code #\9))
(>= (char-code (char char-string 2)) (char-code #\0))
(<= (char-code (char char-string 2)) (char-code #\9))))
;;
;; It is a single character specified in octal
;;
(progn
(setf result (do ((x 0 (1+ x))
(return 0))
((= x 2) return)
(setf return (+ (* return 8)
(- (char-code (char char-string x))
(char-code #\0))))))
(setf used-length 3))
;;
;; We have a group number replacement.
;;
(let ((group (- (char-code first) (char-code #\0))))
(setf result `((let ((nstring (subseq string (car (aref *regex-groups* ,group))
(cadr (aref *regex-groups* ,group)))))
(if (< length (+ index (length nstring)))
(return-from compare nil))
(if (not (string= string nstring
:start1 index
:end1 (+ index (length nstring))))
(return-from compare nil)
(incf index (length nstring)))))))))
(t
(setf result first)))
(if (and (vectorp result) invert)
(bit-xor result #*1111111110011111111111111111111101111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111 t))
(values result used-length)))
;;;
;;; Now for the main regex compiler routine.
;;;
(defun regex-compile (source &key (anchored nil))
"Usage: (regex-compile <expression> [ :anchored (t/nil) ])
This function take a regular expression (supplied as source) and
compiles this into a lambda list that a string argument can then
be applied to. It is also possible to compile this lambda list
for better performance or to save it as a named function for later
use"
(info "Now entering regex-compile with \"~A\"~%" source)
;;
;; This routine works in two parts.
;; The first pass take the regular expression and produces a list of
;; operators and lisp expressions for the entire regular expression.
;; The second pass takes this list and produces the lambda expression.
(let ((expression '()) ; holder for expressions
(group 1) ; Current group index
(group-stack nil) ; Stack of current group endings
(result nil) ; holder for built expression.
(fast-first nil)) ; holder for quick unanchored scan
;;
;; If the expression was an empty string then it alway
;; matches (so lets leave early)
;;
(if (= (length source) 0)
(return-from regex-compile
'(lambda (&rest args)
(declare (ignore args))
t)))
;;
;; If the first character is a caret then set the anchored
;; flags and remove if from the expression string.
;;
(cond ((eql (char source 0) #\^)
(setf source (subseq source 1))
(setf anchored t)))
;;
;; If the first sequence is .* then also set the anchored flags.
;; (This is purely for optimization, it will work without this).
;;
(if (>= (length source) 2)
(if (string= source ".*" :start1 0 :end1 2)
(setf anchored t)))
;;
;; Also, If this is not an anchored search and the first character is
;; a literal, then do a quick scan to see if it is even in the string.
;; If not then we can issue a quick nil,
;; otherwise we can start the search at the matching character to skip
;; the checks of the non-matching characters anyway.
;;
;; If I really wanted to speed up this section of code it would be
;; easy to recognize the case of a fairly long multi-character literal
;; and generate a Boyer-Moore search for the entire literal.
;;
;; I generate the code to do a loop because on CMU Lisp this is about
;; twice as fast a calling position.
;;
(if (and (not anchored)
(not (position (char source 0) *regex-special-chars*))
(not (and (> (length source) 1)
(position (char source 1) *regex-special-chars*))))
(setf fast-first `((if (not (dotimes (i length nil)
(if (eql (char string i)
,(char source 0))
(return (setf start i)))))
(return-from final-return nil)))))
;;
;; Generate the very first expression to save the starting index
;; so that group 0 will be the entire string matched always
;;
(add-exp '((setf (aref *regex-groups* 0)
(list index nil))))
;;
;; Loop over each character in the regular expression building the
;; expression list as we go.
;;
(do ((eindex 0 (1+ eindex)))
((= eindex (length source)))
(let ((current (char source eindex)))
(info "Now processing character ~A index = ~A~%" current eindex)
(case current
((#\.)
;;
;; Generate code for a single wild character
;;
(add-exp '((if (>= index length)
(return-from compare nil)
(incf index)))))
((#\$)
;;
;; If this is the last character of the expression then
;; anchor the end of the expression, otherwise let it slide
;; as a standard character (even though it should be quoted).
;;
(if (= eindex (1- (length source)))
(add-exp '((if (not (= index length))
(return-from compare nil))))
(add-exp '((if (not (and (< index length)
(eql (char string index) #\$)))
(return-from compare nil)
(incf index))))))
((#\*)
(add-exp '(ASTRISK)))
((#\+)
(add-exp '(PLUS)))
((#\?)
(add-exp '(QUESTION)))
((#\()
;;
;; Start a grouping.
;;
(incf group)
(push group group-stack)
(add-exp `((setf (aref *regex-groups* ,(1- group))
(list index nil))))
(add-exp `(,group)))
((#\))
;;
;; End a grouping
;;
(let ((group (pop group-stack)))
(add-exp `((setf (cadr (aref *regex-groups* ,(1- group)))
index)))
(add-exp `(,(- group)))))
((#\[)
;;
;; Start of a range operation.
;; Generate a bit-vector that has one bit per possible character
;; and then on each character or range, set the possible bits.
;;
;; If the first character is carat then invert the set.
(let* ((invert (eql (char source (1+ eindex)) #\^))
(bitstring (make-array 256 :element-type 'bit
:initial-element
(if invert 1 0)))
(set-char (if invert 0 1)))
(if invert (incf eindex))
(do ((x (1+ eindex) (1+ x)))
((eql (char source x) #\]) (setf eindex x))
(info "Building range with character ~A~%" (char source x))
(cond ((and (eql (char source (1+ x)) #\-)
(not (eql (char source (+ x 2)) #\])))
(if (>= (char-code (char source x))
(char-code (char source (+ 2 x))))
(error "Invalid range \"~A-~A\". Ranges must be in acending order"
(char source x) (char source (+ 2 x))))
(do ((j (char-code (char source x)) (1+ j)))
((> j (char-code (char source (+ 2 x))))
(incf x 2))
(info "Setting bit for char ~A code ~A~%" (code-char j) j)
(setf (sbit bitstring j) set-char)))
(t
(cond ((not (eql (char source x) #\]))
(let ((char (char source x)))
;;
;; If the character is quoted then find out what
;; it should have been
;;
(if (eql (char source x) #\\ )
(let ((length))
(multiple-value-setq (char length)
(regex-quoted (subseq source x) invert))
(incf x length)))
(info "Setting bit for char ~A code ~A~%" char (char-code char))
(if (not (vectorp char))
(setf (sbit bitstring (char-code (char source x))) set-char)
(bit-ior bitstring char t))))))))
(add-exp `((let ((range ,bitstring))
(if (>= index length)
(return-from compare nil))
(if (= 1 (sbit range (char-code (char string index))))
(incf index)
(return-from compare nil)))))))
((#\\ )
;;
;; Intreprete the next character as a special, range, octal, group or
;; just the character itself.
;;
(let ((length)
(value))
(multiple-value-setq (value length)
(regex-quoted (subseq source (1+ eindex)) nil))
(cond ((listp value)
(add-exp value))
((characterp value)
(add-exp `((if (not (and (< index length)
(eql (char string index)
,value)))
(return-from compare nil)
(incf index)))))
((vectorp value)
(add-exp `((let ((range ,value))
(if (>= index length)
(return-from compare nil))
(if (= 1 (sbit range (char-code (char string index))))
(incf index)
(return-from compare nil)))))))
(incf eindex length)))
(t
;;
;; We have a literal character.
;; Scan to see how many we have and if it is more than one
;; generate a string= verses as single eql.
;;
(let* ((lit "")
(term (dotimes (litindex (- (length source) eindex) nil)
(let ((litchar (char source (+ eindex litindex))))
(if (position litchar *regex-special-chars*)
(return litchar)
(progn
(info "Now adding ~A index ~A to lit~%" litchar
litindex)
(setf lit (concatenate 'string lit
(string litchar)))))))))
(if (= (length lit) 1)
(add-exp `((if (not (and (< index length)
(eql (char string index) ,current)))
(return-from compare nil)
(incf index))))
;;
;; If we have a multi-character literal then we must
;; check to see if the next character (if there is one)
;; is an astrisk or a plus or a question mark. If so then we must not use this
;; character in the big literal.
(progn
(if (or (eql term #\*)
(eql term #\+)
(eql term #\?))
(setf lit (subseq lit 0 (1- (length lit)))))
(add-exp `((if (< length (+ index ,(length lit)))
(return-from compare nil))
(if (not (string= string ,lit :start1 index
:end1 (+ index ,(length lit))))
(return-from compare nil)
(incf index ,(length lit)))))))
(incf eindex (1- (length lit))))))))
;;
;; Plug end of list to return t. If we made it this far then
;; We have matched!
(add-exp '((setf (cadr (aref *regex-groups* 0))
index)))
(add-exp '((return-from final-return t)))
;;
;;; (print expression)
;;
;; Now take the expression list and turn it into a lambda expression
;; replacing the special flags with lisp code.
;; For example: A BEGIN needs to be replace by an expression that
;; saves the current index, then evaluates everything till it gets to
;; the END then save the new index if it didn't fail.
;; On an ASTRISK I need to take the previous expression and wrap
;; it in a do that will evaluate the expression till an error
;; occurs and then another do that encompases the remainder of the
;; regular expression and iterates decrementing the index by one
;; of the matched expression sizes and then returns nil. After
;; the last expression insert a form that does a return t so that
;; if the entire nested sub-expression succeeds then the loop
;; is broken manually.
;;
(setf result (copy-tree nil))
;;
;; Reversing the current expression makes building up the
;; lambda list easier due to the nexting of expressions when
;; and astrisk has been encountered.
(setf expression (reverse expression))
(do ((elt 0 (1+ elt)))
((>= elt (length expression)))
(let ((piece (nth elt expression)))
;;
;; Now check for PLUS, if so then ditto the expression and then let the
;; ASTRISK below handle the rest.
;;
(cond ((eql piece 'PLUS)
(cond ((listp (nth (1+ elt) expression))
(setf result (append (list (nth (1+ elt) expression))
result)))
;;
;; duplicate the entire group
;; NOTE: This hasn't been implemented yet!!
(t
(error "GROUP repeat hasn't been implemented yet~%")))))
(cond ((listp piece) ;Just append the list
(setf result (append (list piece) result)))
((eql piece 'QUESTION) ; Wrap it in a block that won't fail
(cond ((listp (nth (1+ elt) expression))
(setf result
(append `((progn (block compare
,(nth (1+ elt)
expression))
t))
result))
(incf elt))
;;
;; This is a QUESTION on an entire group which
;; hasn't been implemented yet!!!
;;
(t
(error "Optional groups not implemented yet~%"))))
((or (eql piece 'ASTRISK) ; Do the wild thing!
(eql piece 'PLUS))
(cond ((listp (nth (1+ elt) expression))
;;
;; This is a single character wild card so
;; do the simple form.
;;
(setf result
`((let ((oindex index))
(block compare
(do ()
(nil)
,(nth (1+ elt) expression)))
(do ((start index (1- start)))
((< start oindex) nil)
(let ((index start))
(block compare
,@result))))))
(incf elt))
(t
;;
;; This is a subgroup repeated so I must build
;; the loop using several values.
;;
))
)
(t t)))) ; Just ignore everything else.
;;
;; Now wrap the result in a lambda list that can then be
;; invoked or compiled, however the user wishes.
;;
(if anchored
(setf result
`(lambda (string &key (start 0) (end (length string)))
(setf *regex-groupings* ,group)
(block final-return
(block compare
(let ((index start)
(length end))
,@result)))))
(setf result
`(lambda (string &key (start 0) (end (length string)))
(setf *regex-groupings* ,group)
(block final-return
(let ((length end))
,@fast-first
(do ((marker start (1+ marker)))
((> marker end) nil)
(let ((index marker))
(if (block compare
,@result)
(return t)))))))))))
;; (provide 'nregex)
|