1 const char pcrs_rcs[] = "$Id: pcrs.c,v 1.19.2.1 2002/08/10 11:23:40 oes Exp $";
3 /*********************************************************************
5 * File : $Source: /cvsroot/ijbswa/current/Attic/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 1.19.2.1 2002/08/10 11:23:40 oes
37 * Include prce.h via project.h, where the appropriate
38 * source will have been selected
40 * Revision 1.19 2002/03/08 14:47:48 oes
43 * Revision 1.18 2002/03/08 14:17:14 oes
44 * Fixing -Wconversion warnings
46 * Revision 1.17 2002/03/08 13:45:48 oes
47 * Hiding internal functions
49 * Revision 1.16 2001/11/30 21:32:14 jongfoster
50 * Fixing signed/unsigned comparison (Andreas please check this!)
53 * Revision 1.15 2001/09/20 16:11:06 steudten
55 * Add casting for some string functions.
57 * Revision 1.14 2001/09/09 21:41:57 oes
58 * Fixing yet another silly bug
60 * Revision 1.13 2001/09/06 14:05:59 oes
63 * Revision 1.12 2001/08/18 11:35:00 oes
64 * - Introduced pcrs_strerror()
65 * - made some NULL arguments non-fatal
66 * - added support for \n \r \e \b \t \f \a \0 in substitute
67 * - made quoting adhere to standard rules
68 * - added warning for bad backrefs
69 * - added pcrs_execute_list()
71 * - bugfix & cosmetics
73 * Revision 1.11 2001/08/15 15:32:03 oes
74 * - Added support for Perl's special variables $+, $' and $`
75 * - Improved the substitute parser
76 * - Replaced the hard limit for the maximum number of matches
77 * by dynamic reallocation
79 * Revision 1.10 2001/08/05 13:13:11 jongfoster
80 * Making parameters "const" where possible.
82 * Revision 1.9 2001/07/18 17:27:00 oes
83 * Changed interface; Cosmetics
85 * Revision 1.8 2001/06/29 21:45:41 oes
86 * Indentation, CRLF->LF, Tab-> Space
88 * Revision 1.7 2001/06/29 13:33:04 oes
89 * - Cleaned up, renamed and reordered functions,
92 * - Replaced globalflag with a general flags int
93 * that holds PCRS_GLOBAL, PCRS_SUCCESS, and PCRS_TRIVIAL
94 * - Introduced trivial option that will prevent pcrs
95 * from honouring backreferences in the substitute,
96 * which is useful for large substitutes that are
97 * red in from somewhere and saves the pain of escaping
99 * - Introduced convenience function pcrs_free_joblist()
100 * - Split pcrs_make_job() into pcrs_compile(), which still
101 * takes a complete s/// comand as argument and parses it,
102 * and a new function pcrs_make_job, which takes the
103 * three separate components. This should make for a
104 * much friendlier frontend.
105 * - Removed create_pcrs_job() which was useless
106 * - Fixed a bug in pcrs_execute
107 * - Success flag is now handled by pcrs instead of user
109 * Revision 1.6 2001/06/03 19:12:45 oes
112 * Revision 1.5 2001/05/29 09:50:24 jongfoster
113 * (Fixed one int -> size_t)
115 * Revision 1.4 2001/05/25 14:12:40 oes
116 * Fixed bug: Empty substitutes now detected
118 * Revision 1.3 2001/05/25 11:03:55 oes
119 * Added sanity check for NULL jobs to pcrs_exec_substitution
121 * Revision 1.2 2001/05/22 18:46:04 oes
123 * Added support for PCRE_UNGREEDY behaviour to pcrs,
124 * which is selected by the (nonstandard and therefore
125 * capital) letter 'U' in the option string.
126 * It causes the quantifiers to be ungreedy by default.
127 * Appending a ? turns back to greedy (!).
129 * Revision 1.1.1.1 2001/05/15 13:59:02 oes
130 * Initial import of version 2.9.3 source tree
133 *********************************************************************/
137 * Include project.h just so that the right pcre.h gets
138 * included from there
147 const char pcrs_h_rcs[] = PCRS_H_VERSION;
150 * Internal prototypes
153 static int pcrs_parse_perl_options(const char *optstring, int *flags);
154 static pcrs_substitute *pcrs_compile_replacement(const char *replacement, int trivialflag,
155 int capturecount, int *errptr);
157 /*********************************************************************
159 * Function : pcrs_strerror
161 * Description : Return a string describing a given error code.
164 * 1 : error = the error code
166 * Returns : char * to the descriptive string
168 *********************************************************************/
169 const char *pcrs_strerror(const int error)
175 /* Passed-through PCRE error: */
176 case PCRE_ERROR_NOMEMORY: return "(pcre:) No memory";
178 /* Shouldn't happen unless PCRE or PCRS bug, or user messed with compiled job: */
179 case PCRE_ERROR_NULL: return "(pcre:) NULL code or subject or ovector";
180 case PCRE_ERROR_BADOPTION: return "(pcre:) Unrecognized option bit";
181 case PCRE_ERROR_BADMAGIC: return "(pcre:) Bad magic number in code";
182 case PCRE_ERROR_UNKNOWN_NODE: return "(pcre:) Bad node in pattern";
184 /* Can't happen / not passed: */
185 case PCRE_ERROR_NOSUBSTRING: return "(pcre:) Fire in power supply";
186 case PCRE_ERROR_NOMATCH: return "(pcre:) Water in power supply";
189 case PCRS_ERR_NOMEM: return "(pcrs:) No memory";
190 case PCRS_ERR_CMDSYNTAX: return "(pcrs:) Syntax error while parsing command";
191 case PCRS_ERR_STUDY: return "(pcrs:) PCRE error while studying the pattern";
192 case PCRS_ERR_BADJOB: return "(pcrs:) Bad job - NULL job, pattern or substitute";
193 case PCRS_WARN_BADREF: return "(pcrs:) Backreference out of range";
196 default: return "Unknown error";
199 /* error >= 0: No error */
200 return "(pcrs:) Everything's just fine. Thanks for asking.";
205 /*********************************************************************
207 * Function : pcrs_parse_perl_options
209 * Description : This function parses a string containing the options to
210 * Perl's s/// operator. It returns an integer that is the
211 * pcre equivalent of the symbolic optstring.
212 * Since pcre doesn't know about Perl's 'g' (global) or pcrs',
213 * 'T' (trivial) options but pcrs needs them, the corresponding
214 * flags are set if 'g'or 'T' is encountered.
215 * Note: The 'T' and 'U' options do not conform to Perl.
218 * 1 : optstring = string with options in perl syntax
219 * 2 : flags = see description
221 * Returns : option integer suitable for pcre
223 *********************************************************************/
224 static int pcrs_parse_perl_options(const char *optstring, int *flags)
230 if (NULL == optstring) return 0;
232 for (i = 0; i < strlen(optstring); i++)
236 case 'e': break; /* ToDo ;-) */
237 case 'g': *flags |= PCRS_GLOBAL; break;
238 case 'i': rc |= PCRE_CASELESS; break;
239 case 'm': rc |= PCRE_MULTILINE; break;
241 case 's': rc |= PCRE_DOTALL; break;
242 case 'x': rc |= PCRE_EXTENDED; break;
243 case 'U': rc |= PCRE_UNGREEDY; break;
244 case 'T': *flags |= PCRS_TRIVIAL; break;
253 /*********************************************************************
255 * Function : pcrs_compile_replacement
257 * Description : This function takes a Perl-style replacement (2nd argument
258 * to the s/// operator and returns a compiled pcrs_substitute,
259 * or NULL if memory allocation for the substitute structure
263 * 1 : replacement = replacement part of s/// operator
265 * 2 : trivialflag = Flag that causes backreferences to be
267 * 3 : capturecount = Number of capturing subpatterns in
268 * the pattern. Needed for $+ handling.
269 * 4 : errptr = pointer to an integer in which error
270 * conditions can be returned.
272 * Returns : pcrs_substitute data structure, or NULL if an
273 * error is encountered. In that case, *errptr has
276 *********************************************************************/
277 static pcrs_substitute *pcrs_compile_replacement(const char *replacement, int trivialflag, int capturecount, int *errptr)
284 i = k = l = quoted = 0;
289 if (NULL == replacement)
297 if (NULL == (r = (pcrs_substitute *)malloc(sizeof(pcrs_substitute))))
299 *errptr = PCRS_ERR_NOMEM;
302 memset(r, '\0', sizeof(pcrs_substitute));
304 length = strlen(replacement);
306 if (NULL == (text = (char *)malloc(length + 1)))
309 *errptr = PCRS_ERR_NOMEM;
312 memset(text, '\0', length + 1);
316 * In trivial mode, just copy the substitute text
320 text = strncpy(text, replacement, length + 1);
325 * Else, parse, cut out and record all backreferences
329 while (i < (int)length)
332 if (replacement[i] == '\\')
336 text[k++] = replacement[i++];
341 if (replacement[i+1] && strchr("tnrfae0", replacement[i+1]))
343 switch (replacement[++i])
379 if (replacement[i] == '$' && !quoted && i < (int)(length - 1))
381 char *symbol, symbols[] = "'`+&";
382 r->block_length[l] = k - r->block_offset[l];
384 /* Numerical backreferences */
385 if (isdigit((int)replacement[i + 1]))
387 while (i < (int)length && isdigit((int)replacement[++i]))
389 r->backref[l] = r->backref[l] * 10 + replacement[i] - 48;
391 if (r->backref[l] > capturecount)
393 *errptr = PCRS_WARN_BADREF;
397 /* Symbolic backreferences: */
398 else if (NULL != (symbol = strchr(symbols, replacement[i + 1])))
401 if (symbol - symbols == 2) /* $+ */
403 r->backref[l] = capturecount;
405 else if (symbol - symbols == 3) /* $& */
411 r->backref[l] = PCRS_MAX_SUBMATCHES + 1 - (symbol - symbols);
416 /* Invalid backref -> plain '$' */
422 /* Valid and in range? -> record */
423 if (r->backref[l] < PCRS_MAX_SUBMATCHES + 2)
425 r->backref_count[r->backref[l]] += 1;
426 r->block_offset[++l] = k;
430 *errptr = PCRS_WARN_BADREF;
436 /* Plain chars are copied */
437 text[k++] = replacement[i++];
440 } /* -END- if (!trivialflag) */
447 r->block_length[l] = k - r->block_offset[l];
454 /*********************************************************************
456 * Function : pcrs_free_job
458 * Description : Frees the memory used by a pcrs_job struct and its
459 * dependant structures.
462 * 1 : job = pointer to the pcrs_job structure to be freed
464 * Returns : a pointer to the next job, if there was any, or
467 *********************************************************************/
468 pcrs_job *pcrs_free_job(pcrs_job *job)
479 if (job->pattern != NULL) free(job->pattern);
480 if (job->hints != NULL) free(job->hints);
481 if (job->substitute != NULL)
483 if (job->substitute->text != NULL) free(job->substitute->text);
484 free(job->substitute);
493 /*********************************************************************
495 * Function : pcrs_free_joblist
497 * Description : Iterates through a chained list of pcrs_job's and
498 * frees them using pcrs_free_job.
501 * 1 : joblist = pointer to the first pcrs_job structure to
506 *********************************************************************/
507 void pcrs_free_joblist(pcrs_job *joblist)
509 while ( NULL != (joblist = pcrs_free_job(joblist)) ) {};
516 /*********************************************************************
518 * Function : pcrs_compile_command
520 * Description : Parses a string with a Perl-style s/// command,
521 * calls pcrs_compile, and returns a corresponding
522 * pcrs_job, or NULL if parsing or compiling the job
526 * 1 : command = string with perl-style s/// command
527 * 2 : errptr = pointer to an integer in which error
528 * conditions can be returned.
530 * Returns : a corresponding pcrs_job data structure, or NULL
531 * if an error was encountered. In that case, *errptr
534 *********************************************************************/
535 pcrs_job *pcrs_compile_command(const char *command, int *errptr)
537 int i, k, l, quoted = FALSE;
546 * Tokenize the perl command
548 limit = strlen(command);
551 *errptr = PCRS_ERR_CMDSYNTAX;
556 delimiter = command[1];
559 tokens[l] = (char *) malloc(limit + 1);
561 for (i = 0; i <= (int)limit; i++)
564 if (command[i] == delimiter && !quoted)
571 tokens[0][k++] = '\0';
572 tokens[++l] = tokens[0] + k;
576 else if (command[i] == '\\' && !quoted)
579 if (command[i+1] == delimiter) continue;
585 tokens[0][k++] = command[i];
593 *errptr = PCRS_ERR_CMDSYNTAX;
598 newjob = pcrs_compile(tokens[1], tokens[2], tokens[3], errptr);
605 /*********************************************************************
607 * Function : pcrs_compile
609 * Description : Takes the three arguments to a perl s/// command
610 * and compiles a pcrs_job structure from them.
613 * 1 : pattern = string with perl-style pattern
614 * 2 : substitute = string with perl-style substitute
615 * 3 : options = string with perl-style options
616 * 4 : errptr = pointer to an integer in which error
617 * conditions can be returned.
619 * Returns : a corresponding pcrs_job data structure, or NULL
620 * if an error was encountered. In that case, *errptr
623 *********************************************************************/
624 pcrs_job *pcrs_compile(const char *pattern, const char *substitute, const char *options, int *errptr)
634 * Handle NULL arguments
636 if (pattern == NULL) pattern = "";
637 if (substitute == NULL) substitute = "";
641 * Get and init memory
643 if (NULL == (newjob = (pcrs_job *)malloc(sizeof(pcrs_job))))
645 *errptr = PCRS_ERR_NOMEM;
648 memset(newjob, '\0', sizeof(pcrs_job));
652 * Evaluate the options
654 newjob->options = pcrs_parse_perl_options(options, &flags);
655 newjob->flags = flags;
659 * Compile the pattern
661 newjob->pattern = pcre_compile(pattern, newjob->options, &error, errptr, NULL);
662 if (newjob->pattern == NULL)
664 pcrs_free_job(newjob);
670 * Generate hints. This has little overhead, since the
671 * hints will be NULL for a boring pattern anyway.
673 newjob->hints = pcre_study(newjob->pattern, 0, &error);
676 *errptr = PCRS_ERR_STUDY;
677 pcrs_free_job(newjob);
683 * Determine the number of capturing subpatterns.
684 * This is needed for handling $+ in the substitute.
686 if (0 > (*errptr = pcre_fullinfo(newjob->pattern, newjob->hints, PCRE_INFO_CAPTURECOUNT, &capturecount)))
688 pcrs_free_job(newjob);
694 * Compile the substitute
696 if (NULL == (newjob->substitute = pcrs_compile_replacement(substitute, newjob->flags & PCRS_TRIVIAL, capturecount, errptr)))
698 pcrs_free_job(newjob);
707 /*********************************************************************
709 * Function : pcrs_execute_list
711 * Description : This is a multiple job wrapper for pcrs_execute().
712 * Apply the regular substitutions defined by the jobs in
713 * the joblist to the subject.
714 * The subject itself is left untouched, memory for the result
715 * is malloc()ed and it is the caller's responsibility to free
716 * the result when it's no longer needed.
719 * 1 : joblist = the chained list of pcrs_jobs to be executed
720 * 2 : subject = the subject string
721 * 3 : subject_length = the subject's length
722 * INCLUDING the terminating zero, if string!
723 * 4 : result = char** for returning the result
724 * 5 : result_length = size_t* for returning the result's length
726 * Returns : On success, the number of substitutions that were made.
727 * May be > 1 if job->flags contained PCRS_GLOBAL
728 * On failiure, the (negative) pcre error code describing the
729 * failiure, which may be translated to text using pcrs_strerror().
731 *********************************************************************/
732 int pcrs_execute_list(pcrs_job *joblist, char *subject, size_t subject_length, char **result, size_t *result_length)
736 int hits, total_hits;
739 *result_length = subject_length;
740 hits = total_hits = 0;
742 for (job = joblist; job != NULL; job = job->next)
744 hits = pcrs_execute(job, old, *result_length, &new, result_length);
746 if (old != subject) free(old);
765 /*********************************************************************
767 * Function : pcrs_execute
769 * Description : Apply the regular substitution defined by the job to the
771 * The subject itself is left untouched, memory for the result
772 * is malloc()ed and it is the caller's responsibility to free
773 * the result when it's no longer needed.
776 * 1 : job = the pcrs_job to be executed
777 * 2 : subject = the subject (== original) string
778 * 3 : subject_length = the subject's length
779 * INCLUDING the terminating zero, if string!
780 * 4 : result = char** for returning the result
781 * 5 : result_length = size_t* for returning the result's length
783 * Returns : On success, the number of substitutions that were made.
784 * May be > 1 if job->flags contained PCRS_GLOBAL
785 * On failiure, the (negative) pcre error code describing the
786 * failiure, which may be translated to text using pcrs_strerror().
788 *********************************************************************/
789 int pcrs_execute(pcrs_job *job, char *subject, size_t subject_length, char **result, size_t *result_length)
791 int offsets[3 * PCRS_MAX_SUBMATCHES],
796 max_matches = PCRS_MAX_MATCH_INIT;
798 pcrs_match *matches, *dummy;
804 * Sanity check & memory allocation
806 if (job == NULL || job->pattern == NULL || job->substitute == NULL)
809 return(PCRS_ERR_BADJOB);
812 if (NULL == (matches = (pcrs_match *)malloc(max_matches * sizeof(pcrs_match))))
815 return(PCRS_ERR_NOMEM);
817 memset(matches, '\0', max_matches * sizeof(pcrs_match));
821 * Find the pattern and calculate the space
822 * requirements for the result
824 newsize = subject_length;
826 while ((submatches = pcre_exec(job->pattern, job->hints, subject, (int)subject_length, offset, 0, offsets, 3 * PCRS_MAX_SUBMATCHES)) > 0)
828 job->flags |= PCRS_SUCCESS;
829 matches[i].submatches = submatches;
831 for (k = 0; k < submatches; k++)
833 matches[i].submatch_offset[k] = offsets[2 * k];
835 /* Note: Non-found optional submatches have length -1-(-1)==0 */
836 matches[i].submatch_length[k] = offsets[2 * k + 1] - offsets[2 * k];
838 /* reserve mem for each submatch as often as it is ref'd */
839 newsize += matches[i].submatch_length[k] * job->substitute->backref_count[k];
841 /* plus replacement text size minus match text size */
842 newsize += strlen(job->substitute->text) - matches[i].submatch_length[0];
844 /* chunk before match */
845 matches[i].submatch_offset[PCRS_MAX_SUBMATCHES] = 0;
846 matches[i].submatch_length[PCRS_MAX_SUBMATCHES] = offsets[0];
847 newsize += offsets[0] * job->substitute->backref_count[PCRS_MAX_SUBMATCHES];
849 /* chunk after match */
850 matches[i].submatch_offset[PCRS_MAX_SUBMATCHES + 1] = offsets[1];
851 matches[i].submatch_length[PCRS_MAX_SUBMATCHES + 1] = subject_length - offsets[1] - 1;
852 newsize += (subject_length - offsets[1]) * job->substitute->backref_count[PCRS_MAX_SUBMATCHES + 1];
854 /* Storage for matches exhausted? -> Extend! */
855 if (++i >= max_matches)
857 max_matches = (int)(max_matches * PCRS_MAX_MATCH_GROW);
858 if (NULL == (dummy = (pcrs_match *)realloc(matches, max_matches * sizeof(pcrs_match))))
862 return(PCRS_ERR_NOMEM);
867 /* Non-global search or limit reached? */
868 if (!(job->flags & PCRS_GLOBAL)) break;
870 /* Don't loop on empty matches */
871 if (offsets[1] == offset)
872 if ((size_t)offset < subject_length)
876 /* Go find the next one */
880 /* Pass pcre error through if (bad) failiure */
881 if (submatches < PCRE_ERROR_NOMATCH)
890 * Get memory for the result
892 if ((*result = (char *)malloc(newsize)) == NULL) /* must be free()d by caller */
895 return PCRS_ERR_NOMEM;
903 result_offset = *result;
905 for (i = 0; i < matches_found; i++)
907 /* copy the chunk preceding the match */
908 memcpy(result_offset, subject + offset, (size_t)matches[i].submatch_offset[0] - offset);
909 result_offset += matches[i].submatch_offset[0] - offset;
911 /* For every segment of the substitute.. */
912 for (k = 0; k <= job->substitute->backrefs; k++)
914 /* ...copy its text.. */
915 memcpy(result_offset, job->substitute->text + job->substitute->block_offset[k], job->substitute->block_length[k]);
916 result_offset += job->substitute->block_length[k];
918 /* ..plus, if it's not the last chunk, i.e.: There *is* a backref.. */
919 if (k != job->substitute->backrefs
920 /* ..in legal range.. */
921 && job->substitute->backref[k] < PCRS_MAX_SUBMATCHES + 2
922 /* ..and referencing a real submatch.. */
923 && job->substitute->backref[k] < matches[i].submatches
924 /* ..that is nonempty.. */
925 && matches[i].submatch_length[job->substitute->backref[k]] > 0)
927 /* ..copy the submatch that is ref'd. */
930 subject + matches[i].submatch_offset[job->substitute->backref[k]],
931 matches[i].submatch_length[job->substitute->backref[k]]
933 result_offset += matches[i].submatch_length[job->substitute->backref[k]];
936 offset = matches[i].submatch_offset[0] + matches[i].submatch_length[0];
940 memcpy(result_offset, subject + offset, subject_length - offset);
942 *result_length = newsize;
944 return matches_found;