1 const char pcrs_rcs[] = "$Id: pcrs.c,v 2.3 2002/10/08 16:25:30 oes Exp $";
3 /*********************************************************************
5 * File : $Source: /cvsroot/ijbswa/current/src/pcrs.c,v $
7 * Purpose : pcrs is a supplement to the pcre library by Philip Hazel
8 * <ph10@cam.ac.uk> and adds Perl-style substitution. That
9 * is, it mimics Perl's 's' operator. See pcrs(3) for details.
12 * Copyright : Written and Copyright (C) 2000, 2001 by Andreas S. Oesterhelt
13 * <andreas@oesterhelt.org>
15 * This program is free software; you can redistribute it
16 * and/or modify it under the terms of the GNU Lesser
17 * General Public License (LGPL), version 2.1, which should
18 * be included in this distribution (see LICENSE.txt), with
19 * the exception that the permission to replace that license
20 * with the GNU General Public License (GPL) given in section
21 * 3 is restricted to version 2 of the GPL.
23 * This program is distributed in the hope that it will
24 * be useful, but WITHOUT ANY WARRANTY; without even the
25 * implied warranty of MERCHANTABILITY or FITNESS FOR A
26 * PARTICULAR PURPOSE. See the license for more details.
28 * The GNU Lesser General Public License should be included
29 * with this file. If not, you can view it at
30 * http://www.gnu.org/licenses/lgpl.html
31 * or write to the Free Software Foundation, Inc., 59
32 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
36 * Revision 2.3 2002/10/08 16:25:30 oes
37 * Bugfix: Need to check validity of backreferences explicitly, because when max_matches are reached and matches is expanded, realloc() does not zero the memory. Fixes Bug # 606227
39 * Revision 2.2 2002/09/04 15:52:02 oes
40 * Synced with the stable branch:
41 * Revision 1.19.2.1 2002/08/10 11:23:40 oes
42 * Include prce.h via project.h, where the appropriate
43 * source will have been selected
45 * Revision 2.1 2002/08/26 11:18:24 sarantis
48 * Revision 2.0 2002/06/04 14:34:21 jongfoster
49 * Moving source files to src/
51 * Revision 1.19 2002/03/08 14:47:48 oes
54 * Revision 1.18 2002/03/08 14:17:14 oes
55 * Fixing -Wconversion warnings
57 * Revision 1.17 2002/03/08 13:45:48 oes
58 * Hiding internal functions
60 * Revision 1.16 2001/11/30 21:32:14 jongfoster
61 * Fixing signed/unsigned comparison (Andreas please check this!)
64 * Revision 1.15 2001/09/20 16:11:06 steudten
66 * Add casting for some string functions.
68 * Revision 1.14 2001/09/09 21:41:57 oes
69 * Fixing yet another silly bug
71 * Revision 1.13 2001/09/06 14:05:59 oes
74 * Revision 1.12 2001/08/18 11:35:00 oes
75 * - Introduced pcrs_strerror()
76 * - made some NULL arguments non-fatal
77 * - added support for \n \r \e \b \t \f \a \0 in substitute
78 * - made quoting adhere to standard rules
79 * - added warning for bad backrefs
80 * - added pcrs_execute_list()
82 * - bugfix & cosmetics
84 * Revision 1.11 2001/08/15 15:32:03 oes
85 * - Added support for Perl's special variables $+, $' and $`
86 * - Improved the substitute parser
87 * - Replaced the hard limit for the maximum number of matches
88 * by dynamic reallocation
90 * Revision 1.10 2001/08/05 13:13:11 jongfoster
91 * Making parameters "const" where possible.
93 * Revision 1.9 2001/07/18 17:27:00 oes
94 * Changed interface; Cosmetics
96 * Revision 1.8 2001/06/29 21:45:41 oes
97 * Indentation, CRLF->LF, Tab-> Space
99 * Revision 1.7 2001/06/29 13:33:04 oes
100 * - Cleaned up, renamed and reordered functions,
102 * - Removed my_strsep
103 * - Replaced globalflag with a general flags int
104 * that holds PCRS_GLOBAL, PCRS_SUCCESS, and PCRS_TRIVIAL
105 * - Introduced trivial option that will prevent pcrs
106 * from honouring backreferences in the substitute,
107 * which is useful for large substitutes that are
108 * red in from somewhere and saves the pain of escaping
110 * - Introduced convenience function pcrs_free_joblist()
111 * - Split pcrs_make_job() into pcrs_compile(), which still
112 * takes a complete s/// comand as argument and parses it,
113 * and a new function pcrs_make_job, which takes the
114 * three separate components. This should make for a
115 * much friendlier frontend.
116 * - Removed create_pcrs_job() which was useless
117 * - Fixed a bug in pcrs_execute
118 * - Success flag is now handled by pcrs instead of user
120 * Revision 1.6 2001/06/03 19:12:45 oes
123 * Revision 1.5 2001/05/29 09:50:24 jongfoster
124 * (Fixed one int -> size_t)
126 * Revision 1.4 2001/05/25 14:12:40 oes
127 * Fixed bug: Empty substitutes now detected
129 * Revision 1.3 2001/05/25 11:03:55 oes
130 * Added sanity check for NULL jobs to pcrs_exec_substitution
132 * Revision 1.2 2001/05/22 18:46:04 oes
134 * Added support for PCRE_UNGREEDY behaviour to pcrs,
135 * which is selected by the (nonstandard and therefore
136 * capital) letter 'U' in the option string.
137 * It causes the quantifiers to be ungreedy by default.
138 * Appending a ? turns back to greedy (!).
140 * Revision 1.1.1.1 2001/05/15 13:59:02 oes
141 * Initial import of version 2.9.3 source tree
144 *********************************************************************/
148 * Include project.h just so that the right pcre.h gets
149 * included from there
158 const char pcrs_h_rcs[] = PCRS_H_VERSION;
161 * Internal prototypes
164 static int pcrs_parse_perl_options(const char *optstring, int *flags);
165 static pcrs_substitute *pcrs_compile_replacement(const char *replacement, int trivialflag,
166 int capturecount, int *errptr);
168 /*********************************************************************
170 * Function : pcrs_strerror
172 * Description : Return a string describing a given error code.
175 * 1 : error = the error code
177 * Returns : char * to the descriptive string
179 *********************************************************************/
180 const char *pcrs_strerror(const int error)
186 /* Passed-through PCRE error: */
187 case PCRE_ERROR_NOMEMORY: return "(pcre:) No memory";
189 /* Shouldn't happen unless PCRE or PCRS bug, or user messed with compiled job: */
190 case PCRE_ERROR_NULL: return "(pcre:) NULL code or subject or ovector";
191 case PCRE_ERROR_BADOPTION: return "(pcre:) Unrecognized option bit";
192 case PCRE_ERROR_BADMAGIC: return "(pcre:) Bad magic number in code";
193 case PCRE_ERROR_UNKNOWN_NODE: return "(pcre:) Bad node in pattern";
195 /* Can't happen / not passed: */
196 case PCRE_ERROR_NOSUBSTRING: return "(pcre:) Fire in power supply";
197 case PCRE_ERROR_NOMATCH: return "(pcre:) Water in power supply";
200 case PCRS_ERR_NOMEM: return "(pcrs:) No memory";
201 case PCRS_ERR_CMDSYNTAX: return "(pcrs:) Syntax error while parsing command";
202 case PCRS_ERR_STUDY: return "(pcrs:) PCRE error while studying the pattern";
203 case PCRS_ERR_BADJOB: return "(pcrs:) Bad job - NULL job, pattern or substitute";
204 case PCRS_WARN_BADREF: return "(pcrs:) Backreference out of range";
207 default: return "Unknown error";
210 /* error >= 0: No error */
211 return "(pcrs:) Everything's just fine. Thanks for asking.";
216 /*********************************************************************
218 * Function : pcrs_parse_perl_options
220 * Description : This function parses a string containing the options to
221 * Perl's s/// operator. It returns an integer that is the
222 * pcre equivalent of the symbolic optstring.
223 * Since pcre doesn't know about Perl's 'g' (global) or pcrs',
224 * 'T' (trivial) options but pcrs needs them, the corresponding
225 * flags are set if 'g'or 'T' is encountered.
226 * Note: The 'T' and 'U' options do not conform to Perl.
229 * 1 : optstring = string with options in perl syntax
230 * 2 : flags = see description
232 * Returns : option integer suitable for pcre
234 *********************************************************************/
235 static int pcrs_parse_perl_options(const char *optstring, int *flags)
241 if (NULL == optstring) return 0;
243 for (i = 0; i < strlen(optstring); i++)
247 case 'e': break; /* ToDo ;-) */
248 case 'g': *flags |= PCRS_GLOBAL; break;
249 case 'i': rc |= PCRE_CASELESS; break;
250 case 'm': rc |= PCRE_MULTILINE; break;
252 case 's': rc |= PCRE_DOTALL; break;
253 case 'x': rc |= PCRE_EXTENDED; break;
254 case 'U': rc |= PCRE_UNGREEDY; break;
255 case 'T': *flags |= PCRS_TRIVIAL; break;
264 /*********************************************************************
266 * Function : pcrs_compile_replacement
268 * Description : This function takes a Perl-style replacement (2nd argument
269 * to the s/// operator and returns a compiled pcrs_substitute,
270 * or NULL if memory allocation for the substitute structure
274 * 1 : replacement = replacement part of s/// operator
276 * 2 : trivialflag = Flag that causes backreferences to be
278 * 3 : capturecount = Number of capturing subpatterns in
279 * the pattern. Needed for $+ handling.
280 * 4 : errptr = pointer to an integer in which error
281 * conditions can be returned.
283 * Returns : pcrs_substitute data structure, or NULL if an
284 * error is encountered. In that case, *errptr has
287 *********************************************************************/
288 static pcrs_substitute *pcrs_compile_replacement(const char *replacement, int trivialflag, int capturecount, int *errptr)
295 i = k = l = quoted = 0;
300 if (NULL == replacement)
308 if (NULL == (r = (pcrs_substitute *)malloc(sizeof(pcrs_substitute))))
310 *errptr = PCRS_ERR_NOMEM;
313 memset(r, '\0', sizeof(pcrs_substitute));
315 length = strlen(replacement);
317 if (NULL == (text = (char *)malloc(length + 1)))
320 *errptr = PCRS_ERR_NOMEM;
323 memset(text, '\0', length + 1);
327 * In trivial mode, just copy the substitute text
331 text = strncpy(text, replacement, length + 1);
336 * Else, parse, cut out and record all backreferences
340 while (i < (int)length)
343 if (replacement[i] == '\\')
347 text[k++] = replacement[i++];
352 if (replacement[i+1] && strchr("tnrfae0", replacement[i+1]))
354 switch (replacement[++i])
390 if (replacement[i] == '$' && !quoted && i < (int)(length - 1))
392 char *symbol, symbols[] = "'`+&";
393 r->block_length[l] = k - r->block_offset[l];
395 /* Numerical backreferences */
396 if (isdigit((int)replacement[i + 1]))
398 while (i < (int)length && isdigit((int)replacement[++i]))
400 r->backref[l] = r->backref[l] * 10 + replacement[i] - 48;
402 if (r->backref[l] > capturecount)
404 *errptr = PCRS_WARN_BADREF;
408 /* Symbolic backreferences: */
409 else if (NULL != (symbol = strchr(symbols, replacement[i + 1])))
412 if (symbol - symbols == 2) /* $+ */
414 r->backref[l] = capturecount;
416 else if (symbol - symbols == 3) /* $& */
422 r->backref[l] = PCRS_MAX_SUBMATCHES + 1 - (symbol - symbols);
427 /* Invalid backref -> plain '$' */
433 /* Valid and in range? -> record */
434 if (r->backref[l] < PCRS_MAX_SUBMATCHES + 2)
436 r->backref_count[r->backref[l]] += 1;
437 r->block_offset[++l] = k;
441 *errptr = PCRS_WARN_BADREF;
447 /* Plain chars are copied */
448 text[k++] = replacement[i++];
451 } /* -END- if (!trivialflag) */
458 r->block_length[l] = k - r->block_offset[l];
465 /*********************************************************************
467 * Function : pcrs_free_job
469 * Description : Frees the memory used by a pcrs_job struct and its
470 * dependant structures.
473 * 1 : job = pointer to the pcrs_job structure to be freed
475 * Returns : a pointer to the next job, if there was any, or
478 *********************************************************************/
479 pcrs_job *pcrs_free_job(pcrs_job *job)
490 if (job->pattern != NULL) free(job->pattern);
491 if (job->hints != NULL) free(job->hints);
492 if (job->substitute != NULL)
494 if (job->substitute->text != NULL) free(job->substitute->text);
495 free(job->substitute);
504 /*********************************************************************
506 * Function : pcrs_free_joblist
508 * Description : Iterates through a chained list of pcrs_job's and
509 * frees them using pcrs_free_job.
512 * 1 : joblist = pointer to the first pcrs_job structure to
517 *********************************************************************/
518 void pcrs_free_joblist(pcrs_job *joblist)
520 while ( NULL != (joblist = pcrs_free_job(joblist)) ) {};
527 /*********************************************************************
529 * Function : pcrs_compile_command
531 * Description : Parses a string with a Perl-style s/// command,
532 * calls pcrs_compile, and returns a corresponding
533 * pcrs_job, or NULL if parsing or compiling the job
537 * 1 : command = string with perl-style s/// command
538 * 2 : errptr = pointer to an integer in which error
539 * conditions can be returned.
541 * Returns : a corresponding pcrs_job data structure, or NULL
542 * if an error was encountered. In that case, *errptr
545 *********************************************************************/
546 pcrs_job *pcrs_compile_command(const char *command, int *errptr)
548 int i, k, l, quoted = FALSE;
557 * Tokenize the perl command
559 limit = strlen(command);
562 *errptr = PCRS_ERR_CMDSYNTAX;
567 delimiter = command[1];
570 tokens[l] = (char *) malloc(limit + 1);
572 for (i = 0; i <= (int)limit; i++)
575 if (command[i] == delimiter && !quoted)
582 tokens[0][k++] = '\0';
583 tokens[++l] = tokens[0] + k;
587 else if (command[i] == '\\' && !quoted)
590 if (command[i+1] == delimiter) continue;
596 tokens[0][k++] = command[i];
604 *errptr = PCRS_ERR_CMDSYNTAX;
609 newjob = pcrs_compile(tokens[1], tokens[2], tokens[3], errptr);
616 /*********************************************************************
618 * Function : pcrs_compile
620 * Description : Takes the three arguments to a perl s/// command
621 * and compiles a pcrs_job structure from them.
624 * 1 : pattern = string with perl-style pattern
625 * 2 : substitute = string with perl-style substitute
626 * 3 : options = string with perl-style options
627 * 4 : errptr = pointer to an integer in which error
628 * conditions can be returned.
630 * Returns : a corresponding pcrs_job data structure, or NULL
631 * if an error was encountered. In that case, *errptr
634 *********************************************************************/
635 pcrs_job *pcrs_compile(const char *pattern, const char *substitute, const char *options, int *errptr)
645 * Handle NULL arguments
647 if (pattern == NULL) pattern = "";
648 if (substitute == NULL) substitute = "";
652 * Get and init memory
654 if (NULL == (newjob = (pcrs_job *)malloc(sizeof(pcrs_job))))
656 *errptr = PCRS_ERR_NOMEM;
659 memset(newjob, '\0', sizeof(pcrs_job));
663 * Evaluate the options
665 newjob->options = pcrs_parse_perl_options(options, &flags);
666 newjob->flags = flags;
670 * Compile the pattern
672 newjob->pattern = pcre_compile(pattern, newjob->options, &error, errptr, NULL);
673 if (newjob->pattern == NULL)
675 pcrs_free_job(newjob);
681 * Generate hints. This has little overhead, since the
682 * hints will be NULL for a boring pattern anyway.
684 newjob->hints = pcre_study(newjob->pattern, 0, &error);
687 *errptr = PCRS_ERR_STUDY;
688 pcrs_free_job(newjob);
694 * Determine the number of capturing subpatterns.
695 * This is needed for handling $+ in the substitute.
697 if (0 > (*errptr = pcre_fullinfo(newjob->pattern, newjob->hints, PCRE_INFO_CAPTURECOUNT, &capturecount)))
699 pcrs_free_job(newjob);
705 * Compile the substitute
707 if (NULL == (newjob->substitute = pcrs_compile_replacement(substitute, newjob->flags & PCRS_TRIVIAL, capturecount, errptr)))
709 pcrs_free_job(newjob);
718 /*********************************************************************
720 * Function : pcrs_execute_list
722 * Description : This is a multiple job wrapper for pcrs_execute().
723 * Apply the regular substitutions defined by the jobs in
724 * the joblist to the subject.
725 * The subject itself is left untouched, memory for the result
726 * is malloc()ed and it is the caller's responsibility to free
727 * the result when it's no longer needed.
730 * 1 : joblist = the chained list of pcrs_jobs to be executed
731 * 2 : subject = the subject string
732 * 3 : subject_length = the subject's length
733 * INCLUDING the terminating zero, if string!
734 * 4 : result = char** for returning the result
735 * 5 : result_length = size_t* for returning the result's length
737 * Returns : On success, the number of substitutions that were made.
738 * May be > 1 if job->flags contained PCRS_GLOBAL
739 * On failure, the (negative) pcre error code describing the
740 * failure, which may be translated to text using pcrs_strerror().
742 *********************************************************************/
743 int pcrs_execute_list(pcrs_job *joblist, char *subject, size_t subject_length, char **result, size_t *result_length)
746 char *old_item, *new_item;
747 int hits, total_hits;
750 *result_length = subject_length;
751 hits = total_hits = 0;
753 for (job = joblist; job != NULL; job = job->next)
755 hits = pcrs_execute(job, old_item, *result_length, &new_item, result_length);
757 if (old_item != subject) free(old_item);
776 /*********************************************************************
778 * Function : pcrs_execute
780 * Description : Apply the regular substitution defined by the job to the
782 * The subject itself is left untouched, memory for the result
783 * is malloc()ed and it is the caller's responsibility to free
784 * the result when it's no longer needed.
787 * 1 : job = the pcrs_job to be executed
788 * 2 : subject = the subject (== original) string
789 * 3 : subject_length = the subject's length
790 * INCLUDING the terminating zero, if string!
791 * 4 : result = char** for returning the result
792 * 5 : result_length = size_t* for returning the result's length
794 * Returns : On success, the number of substitutions that were made.
795 * May be > 1 if job->flags contained PCRS_GLOBAL
796 * On failure, the (negative) pcre error code describing the
797 * failure, which may be translated to text using pcrs_strerror().
799 *********************************************************************/
800 int pcrs_execute(pcrs_job *job, char *subject, size_t subject_length, char **result, size_t *result_length)
802 int offsets[3 * PCRS_MAX_SUBMATCHES],
807 max_matches = PCRS_MAX_MATCH_INIT;
809 pcrs_match *matches, *dummy;
815 * Sanity check & memory allocation
817 if (job == NULL || job->pattern == NULL || job->substitute == NULL)
820 return(PCRS_ERR_BADJOB);
823 if (NULL == (matches = (pcrs_match *)malloc(max_matches * sizeof(pcrs_match))))
826 return(PCRS_ERR_NOMEM);
828 memset(matches, '\0', max_matches * sizeof(pcrs_match));
832 * Find the pattern and calculate the space
833 * requirements for the result
835 newsize = subject_length;
837 while ((submatches = pcre_exec(job->pattern, job->hints, subject, (int)subject_length, offset, 0, offsets, 3 * PCRS_MAX_SUBMATCHES)) > 0)
839 job->flags |= PCRS_SUCCESS;
840 matches[i].submatches = submatches;
842 for (k = 0; k < submatches; k++)
844 matches[i].submatch_offset[k] = offsets[2 * k];
846 /* Note: Non-found optional submatches have length -1-(-1)==0 */
847 matches[i].submatch_length[k] = offsets[2 * k + 1] - offsets[2 * k];
849 /* reserve mem for each submatch as often as it is ref'd */
850 newsize += matches[i].submatch_length[k] * job->substitute->backref_count[k];
852 /* plus replacement text size minus match text size */
853 newsize += strlen(job->substitute->text) - matches[i].submatch_length[0];
855 /* chunk before match */
856 matches[i].submatch_offset[PCRS_MAX_SUBMATCHES] = 0;
857 matches[i].submatch_length[PCRS_MAX_SUBMATCHES] = offsets[0];
858 newsize += offsets[0] * job->substitute->backref_count[PCRS_MAX_SUBMATCHES];
860 /* chunk after match */
861 matches[i].submatch_offset[PCRS_MAX_SUBMATCHES + 1] = offsets[1];
862 matches[i].submatch_length[PCRS_MAX_SUBMATCHES + 1] = subject_length - offsets[1] - 1;
863 newsize += (subject_length - offsets[1]) * job->substitute->backref_count[PCRS_MAX_SUBMATCHES + 1];
865 /* Storage for matches exhausted? -> Extend! */
866 if (++i >= max_matches)
868 max_matches = (int)(max_matches * PCRS_MAX_MATCH_GROW);
869 if (NULL == (dummy = (pcrs_match *)realloc(matches, max_matches * sizeof(pcrs_match))))
873 return(PCRS_ERR_NOMEM);
878 /* Non-global search or limit reached? */
879 if (!(job->flags & PCRS_GLOBAL)) break;
881 /* Don't loop on empty matches */
882 if (offsets[1] == offset)
883 if ((size_t)offset < subject_length)
887 /* Go find the next one */
891 /* Pass pcre error through if (bad) failure */
892 if (submatches < PCRE_ERROR_NOMATCH)
901 * Get memory for the result
903 if ((*result = (char *)malloc(newsize)) == NULL) /* must be free()d by caller */
906 return PCRS_ERR_NOMEM;
914 result_offset = *result;
916 for (i = 0; i < matches_found; i++)
918 /* copy the chunk preceding the match */
919 memcpy(result_offset, subject + offset, (size_t)matches[i].submatch_offset[0] - offset);
920 result_offset += matches[i].submatch_offset[0] - offset;
922 /* For every segment of the substitute.. */
923 for (k = 0; k <= job->substitute->backrefs; k++)
925 /* ...copy its text.. */
926 memcpy(result_offset, job->substitute->text + job->substitute->block_offset[k], job->substitute->block_length[k]);
927 result_offset += job->substitute->block_length[k];
929 /* ..plus, if it's not the last chunk, i.e.: There *is* a backref.. */
930 if (k != job->substitute->backrefs
931 /* ..in legal range.. */
932 && job->substitute->backref[k] < PCRS_MAX_SUBMATCHES + 2
933 /* ..and referencing a real submatch.. */
934 && job->substitute->backref[k] < matches[i].submatches
935 /* ..that is nonempty.. */
936 && matches[i].submatch_length[job->substitute->backref[k]] > 0)
938 /* ..copy the submatch that is ref'd. */
941 subject + matches[i].submatch_offset[job->substitute->backref[k]],
942 matches[i].submatch_length[job->substitute->backref[k]]
944 result_offset += matches[i].submatch_length[job->substitute->backref[k]];
947 offset = matches[i].submatch_offset[0] + matches[i].submatch_length[0];
951 memcpy(result_offset, subject + offset, subject_length - offset);
953 *result_length = newsize;
955 return matches_found;