1 const char pcrs_rcs[] = "$Id: pcrs.c,v 2.1 2002/08/26 11:18:24 sarantis 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 2.1 2002/08/26 11:18:24 sarantis
39 * Revision 2.0 2002/06/04 14:34:21 jongfoster
40 * Moving source files to src/
42 * Revision 1.19 2002/03/08 14:47:48 oes
45 * Revision 1.18 2002/03/08 14:17:14 oes
46 * Fixing -Wconversion warnings
48 * Revision 1.17 2002/03/08 13:45:48 oes
49 * Hiding internal functions
51 * Revision 1.16 2001/11/30 21:32:14 jongfoster
52 * Fixing signed/unsigned comparison (Andreas please check this!)
55 * Revision 1.15 2001/09/20 16:11:06 steudten
57 * Add casting for some string functions.
59 * Revision 1.14 2001/09/09 21:41:57 oes
60 * Fixing yet another silly bug
62 * Revision 1.13 2001/09/06 14:05:59 oes
65 * Revision 1.12 2001/08/18 11:35:00 oes
66 * - Introduced pcrs_strerror()
67 * - made some NULL arguments non-fatal
68 * - added support for \n \r \e \b \t \f \a \0 in substitute
69 * - made quoting adhere to standard rules
70 * - added warning for bad backrefs
71 * - added pcrs_execute_list()
73 * - bugfix & cosmetics
75 * Revision 1.11 2001/08/15 15:32:03 oes
76 * - Added support for Perl's special variables $+, $' and $`
77 * - Improved the substitute parser
78 * - Replaced the hard limit for the maximum number of matches
79 * by dynamic reallocation
81 * Revision 1.10 2001/08/05 13:13:11 jongfoster
82 * Making parameters "const" where possible.
84 * Revision 1.9 2001/07/18 17:27:00 oes
85 * Changed interface; Cosmetics
87 * Revision 1.8 2001/06/29 21:45:41 oes
88 * Indentation, CRLF->LF, Tab-> Space
90 * Revision 1.7 2001/06/29 13:33:04 oes
91 * - Cleaned up, renamed and reordered functions,
94 * - Replaced globalflag with a general flags int
95 * that holds PCRS_GLOBAL, PCRS_SUCCESS, and PCRS_TRIVIAL
96 * - Introduced trivial option that will prevent pcrs
97 * from honouring backreferences in the substitute,
98 * which is useful for large substitutes that are
99 * red in from somewhere and saves the pain of escaping
101 * - Introduced convenience function pcrs_free_joblist()
102 * - Split pcrs_make_job() into pcrs_compile(), which still
103 * takes a complete s/// comand as argument and parses it,
104 * and a new function pcrs_make_job, which takes the
105 * three separate components. This should make for a
106 * much friendlier frontend.
107 * - Removed create_pcrs_job() which was useless
108 * - Fixed a bug in pcrs_execute
109 * - Success flag is now handled by pcrs instead of user
111 * Revision 1.6 2001/06/03 19:12:45 oes
114 * Revision 1.5 2001/05/29 09:50:24 jongfoster
115 * (Fixed one int -> size_t)
117 * Revision 1.4 2001/05/25 14:12:40 oes
118 * Fixed bug: Empty substitutes now detected
120 * Revision 1.3 2001/05/25 11:03:55 oes
121 * Added sanity check for NULL jobs to pcrs_exec_substitution
123 * Revision 1.2 2001/05/22 18:46:04 oes
125 * Added support for PCRE_UNGREEDY behaviour to pcrs,
126 * which is selected by the (nonstandard and therefore
127 * capital) letter 'U' in the option string.
128 * It causes the quantifiers to be ungreedy by default.
129 * Appending a ? turns back to greedy (!).
131 * Revision 1.1.1.1 2001/05/15 13:59:02 oes
132 * Initial import of version 2.9.3 source tree
135 *********************************************************************/
139 * Include project.h just so that the right pcre.h gets
140 * included from there
149 const char pcrs_h_rcs[] = PCRS_H_VERSION;
152 * Internal prototypes
155 static int pcrs_parse_perl_options(const char *optstring, int *flags);
156 static pcrs_substitute *pcrs_compile_replacement(const char *replacement, int trivialflag,
157 int capturecount, int *errptr);
159 /*********************************************************************
161 * Function : pcrs_strerror
163 * Description : Return a string describing a given error code.
166 * 1 : error = the error code
168 * Returns : char * to the descriptive string
170 *********************************************************************/
171 const char *pcrs_strerror(const int error)
177 /* Passed-through PCRE error: */
178 case PCRE_ERROR_NOMEMORY: return "(pcre:) No memory";
180 /* Shouldn't happen unless PCRE or PCRS bug, or user messed with compiled job: */
181 case PCRE_ERROR_NULL: return "(pcre:) NULL code or subject or ovector";
182 case PCRE_ERROR_BADOPTION: return "(pcre:) Unrecognized option bit";
183 case PCRE_ERROR_BADMAGIC: return "(pcre:) Bad magic number in code";
184 case PCRE_ERROR_UNKNOWN_NODE: return "(pcre:) Bad node in pattern";
186 /* Can't happen / not passed: */
187 case PCRE_ERROR_NOSUBSTRING: return "(pcre:) Fire in power supply";
188 case PCRE_ERROR_NOMATCH: return "(pcre:) Water in power supply";
191 case PCRS_ERR_NOMEM: return "(pcrs:) No memory";
192 case PCRS_ERR_CMDSYNTAX: return "(pcrs:) Syntax error while parsing command";
193 case PCRS_ERR_STUDY: return "(pcrs:) PCRE error while studying the pattern";
194 case PCRS_ERR_BADJOB: return "(pcrs:) Bad job - NULL job, pattern or substitute";
195 case PCRS_WARN_BADREF: return "(pcrs:) Backreference out of range";
198 default: return "Unknown error";
201 /* error >= 0: No error */
202 return "(pcrs:) Everything's just fine. Thanks for asking.";
207 /*********************************************************************
209 * Function : pcrs_parse_perl_options
211 * Description : This function parses a string containing the options to
212 * Perl's s/// operator. It returns an integer that is the
213 * pcre equivalent of the symbolic optstring.
214 * Since pcre doesn't know about Perl's 'g' (global) or pcrs',
215 * 'T' (trivial) options but pcrs needs them, the corresponding
216 * flags are set if 'g'or 'T' is encountered.
217 * Note: The 'T' and 'U' options do not conform to Perl.
220 * 1 : optstring = string with options in perl syntax
221 * 2 : flags = see description
223 * Returns : option integer suitable for pcre
225 *********************************************************************/
226 static int pcrs_parse_perl_options(const char *optstring, int *flags)
232 if (NULL == optstring) return 0;
234 for (i = 0; i < strlen(optstring); i++)
238 case 'e': break; /* ToDo ;-) */
239 case 'g': *flags |= PCRS_GLOBAL; break;
240 case 'i': rc |= PCRE_CASELESS; break;
241 case 'm': rc |= PCRE_MULTILINE; break;
243 case 's': rc |= PCRE_DOTALL; break;
244 case 'x': rc |= PCRE_EXTENDED; break;
245 case 'U': rc |= PCRE_UNGREEDY; break;
246 case 'T': *flags |= PCRS_TRIVIAL; break;
255 /*********************************************************************
257 * Function : pcrs_compile_replacement
259 * Description : This function takes a Perl-style replacement (2nd argument
260 * to the s/// operator and returns a compiled pcrs_substitute,
261 * or NULL if memory allocation for the substitute structure
265 * 1 : replacement = replacement part of s/// operator
267 * 2 : trivialflag = Flag that causes backreferences to be
269 * 3 : capturecount = Number of capturing subpatterns in
270 * the pattern. Needed for $+ handling.
271 * 4 : errptr = pointer to an integer in which error
272 * conditions can be returned.
274 * Returns : pcrs_substitute data structure, or NULL if an
275 * error is encountered. In that case, *errptr has
278 *********************************************************************/
279 static pcrs_substitute *pcrs_compile_replacement(const char *replacement, int trivialflag, int capturecount, int *errptr)
286 i = k = l = quoted = 0;
291 if (NULL == replacement)
299 if (NULL == (r = (pcrs_substitute *)malloc(sizeof(pcrs_substitute))))
301 *errptr = PCRS_ERR_NOMEM;
304 memset(r, '\0', sizeof(pcrs_substitute));
306 length = strlen(replacement);
308 if (NULL == (text = (char *)malloc(length + 1)))
311 *errptr = PCRS_ERR_NOMEM;
314 memset(text, '\0', length + 1);
318 * In trivial mode, just copy the substitute text
322 text = strncpy(text, replacement, length + 1);
327 * Else, parse, cut out and record all backreferences
331 while (i < (int)length)
334 if (replacement[i] == '\\')
338 text[k++] = replacement[i++];
343 if (replacement[i+1] && strchr("tnrfae0", replacement[i+1]))
345 switch (replacement[++i])
381 if (replacement[i] == '$' && !quoted && i < (int)(length - 1))
383 char *symbol, symbols[] = "'`+&";
384 r->block_length[l] = k - r->block_offset[l];
386 /* Numerical backreferences */
387 if (isdigit((int)replacement[i + 1]))
389 while (i < (int)length && isdigit((int)replacement[++i]))
391 r->backref[l] = r->backref[l] * 10 + replacement[i] - 48;
393 if (r->backref[l] > capturecount)
395 *errptr = PCRS_WARN_BADREF;
399 /* Symbolic backreferences: */
400 else if (NULL != (symbol = strchr(symbols, replacement[i + 1])))
403 if (symbol - symbols == 2) /* $+ */
405 r->backref[l] = capturecount;
407 else if (symbol - symbols == 3) /* $& */
413 r->backref[l] = PCRS_MAX_SUBMATCHES + 1 - (symbol - symbols);
418 /* Invalid backref -> plain '$' */
424 /* Valid and in range? -> record */
425 if (r->backref[l] < PCRS_MAX_SUBMATCHES + 2)
427 r->backref_count[r->backref[l]] += 1;
428 r->block_offset[++l] = k;
432 *errptr = PCRS_WARN_BADREF;
438 /* Plain chars are copied */
439 text[k++] = replacement[i++];
442 } /* -END- if (!trivialflag) */
449 r->block_length[l] = k - r->block_offset[l];
456 /*********************************************************************
458 * Function : pcrs_free_job
460 * Description : Frees the memory used by a pcrs_job struct and its
461 * dependant structures.
464 * 1 : job = pointer to the pcrs_job structure to be freed
466 * Returns : a pointer to the next job, if there was any, or
469 *********************************************************************/
470 pcrs_job *pcrs_free_job(pcrs_job *job)
481 if (job->pattern != NULL) free(job->pattern);
482 if (job->hints != NULL) free(job->hints);
483 if (job->substitute != NULL)
485 if (job->substitute->text != NULL) free(job->substitute->text);
486 free(job->substitute);
495 /*********************************************************************
497 * Function : pcrs_free_joblist
499 * Description : Iterates through a chained list of pcrs_job's and
500 * frees them using pcrs_free_job.
503 * 1 : joblist = pointer to the first pcrs_job structure to
508 *********************************************************************/
509 void pcrs_free_joblist(pcrs_job *joblist)
511 while ( NULL != (joblist = pcrs_free_job(joblist)) ) {};
518 /*********************************************************************
520 * Function : pcrs_compile_command
522 * Description : Parses a string with a Perl-style s/// command,
523 * calls pcrs_compile, and returns a corresponding
524 * pcrs_job, or NULL if parsing or compiling the job
528 * 1 : command = string with perl-style s/// command
529 * 2 : errptr = pointer to an integer in which error
530 * conditions can be returned.
532 * Returns : a corresponding pcrs_job data structure, or NULL
533 * if an error was encountered. In that case, *errptr
536 *********************************************************************/
537 pcrs_job *pcrs_compile_command(const char *command, int *errptr)
539 int i, k, l, quoted = FALSE;
548 * Tokenize the perl command
550 limit = strlen(command);
553 *errptr = PCRS_ERR_CMDSYNTAX;
558 delimiter = command[1];
561 tokens[l] = (char *) malloc(limit + 1);
563 for (i = 0; i <= (int)limit; i++)
566 if (command[i] == delimiter && !quoted)
573 tokens[0][k++] = '\0';
574 tokens[++l] = tokens[0] + k;
578 else if (command[i] == '\\' && !quoted)
581 if (command[i+1] == delimiter) continue;
587 tokens[0][k++] = command[i];
595 *errptr = PCRS_ERR_CMDSYNTAX;
600 newjob = pcrs_compile(tokens[1], tokens[2], tokens[3], errptr);
607 /*********************************************************************
609 * Function : pcrs_compile
611 * Description : Takes the three arguments to a perl s/// command
612 * and compiles a pcrs_job structure from them.
615 * 1 : pattern = string with perl-style pattern
616 * 2 : substitute = string with perl-style substitute
617 * 3 : options = string with perl-style options
618 * 4 : errptr = pointer to an integer in which error
619 * conditions can be returned.
621 * Returns : a corresponding pcrs_job data structure, or NULL
622 * if an error was encountered. In that case, *errptr
625 *********************************************************************/
626 pcrs_job *pcrs_compile(const char *pattern, const char *substitute, const char *options, int *errptr)
636 * Handle NULL arguments
638 if (pattern == NULL) pattern = "";
639 if (substitute == NULL) substitute = "";
643 * Get and init memory
645 if (NULL == (newjob = (pcrs_job *)malloc(sizeof(pcrs_job))))
647 *errptr = PCRS_ERR_NOMEM;
650 memset(newjob, '\0', sizeof(pcrs_job));
654 * Evaluate the options
656 newjob->options = pcrs_parse_perl_options(options, &flags);
657 newjob->flags = flags;
661 * Compile the pattern
663 newjob->pattern = pcre_compile(pattern, newjob->options, &error, errptr, NULL);
664 if (newjob->pattern == NULL)
666 pcrs_free_job(newjob);
672 * Generate hints. This has little overhead, since the
673 * hints will be NULL for a boring pattern anyway.
675 newjob->hints = pcre_study(newjob->pattern, 0, &error);
678 *errptr = PCRS_ERR_STUDY;
679 pcrs_free_job(newjob);
685 * Determine the number of capturing subpatterns.
686 * This is needed for handling $+ in the substitute.
688 if (0 > (*errptr = pcre_fullinfo(newjob->pattern, newjob->hints, PCRE_INFO_CAPTURECOUNT, &capturecount)))
690 pcrs_free_job(newjob);
696 * Compile the substitute
698 if (NULL == (newjob->substitute = pcrs_compile_replacement(substitute, newjob->flags & PCRS_TRIVIAL, capturecount, errptr)))
700 pcrs_free_job(newjob);
709 /*********************************************************************
711 * Function : pcrs_execute_list
713 * Description : This is a multiple job wrapper for pcrs_execute().
714 * Apply the regular substitutions defined by the jobs in
715 * the joblist to the subject.
716 * The subject itself is left untouched, memory for the result
717 * is malloc()ed and it is the caller's responsibility to free
718 * the result when it's no longer needed.
721 * 1 : joblist = the chained list of pcrs_jobs to be executed
722 * 2 : subject = the subject string
723 * 3 : subject_length = the subject's length
724 * INCLUDING the terminating zero, if string!
725 * 4 : result = char** for returning the result
726 * 5 : result_length = size_t* for returning the result's length
728 * Returns : On success, the number of substitutions that were made.
729 * May be > 1 if job->flags contained PCRS_GLOBAL
730 * On failure, the (negative) pcre error code describing the
731 * failure, which may be translated to text using pcrs_strerror().
733 *********************************************************************/
734 int pcrs_execute_list(pcrs_job *joblist, char *subject, size_t subject_length, char **result, size_t *result_length)
738 int hits, total_hits;
741 *result_length = subject_length;
742 hits = total_hits = 0;
744 for (job = joblist; job != NULL; job = job->next)
746 hits = pcrs_execute(job, old, *result_length, &new, result_length);
748 if (old != subject) free(old);
767 /*********************************************************************
769 * Function : pcrs_execute
771 * Description : Apply the regular substitution defined by the job to the
773 * The subject itself is left untouched, memory for the result
774 * is malloc()ed and it is the caller's responsibility to free
775 * the result when it's no longer needed.
778 * 1 : job = the pcrs_job to be executed
779 * 2 : subject = the subject (== original) string
780 * 3 : subject_length = the subject's length
781 * INCLUDING the terminating zero, if string!
782 * 4 : result = char** for returning the result
783 * 5 : result_length = size_t* for returning the result's length
785 * Returns : On success, the number of substitutions that were made.
786 * May be > 1 if job->flags contained PCRS_GLOBAL
787 * On failure, the (negative) pcre error code describing the
788 * failure, which may be translated to text using pcrs_strerror().
790 *********************************************************************/
791 int pcrs_execute(pcrs_job *job, char *subject, size_t subject_length, char **result, size_t *result_length)
793 int offsets[3 * PCRS_MAX_SUBMATCHES],
798 max_matches = PCRS_MAX_MATCH_INIT;
800 pcrs_match *matches, *dummy;
806 * Sanity check & memory allocation
808 if (job == NULL || job->pattern == NULL || job->substitute == NULL)
811 return(PCRS_ERR_BADJOB);
814 if (NULL == (matches = (pcrs_match *)malloc(max_matches * sizeof(pcrs_match))))
817 return(PCRS_ERR_NOMEM);
819 memset(matches, '\0', max_matches * sizeof(pcrs_match));
823 * Find the pattern and calculate the space
824 * requirements for the result
826 newsize = subject_length;
828 while ((submatches = pcre_exec(job->pattern, job->hints, subject, (int)subject_length, offset, 0, offsets, 3 * PCRS_MAX_SUBMATCHES)) > 0)
830 job->flags |= PCRS_SUCCESS;
831 matches[i].submatches = submatches;
833 for (k = 0; k < submatches; k++)
835 matches[i].submatch_offset[k] = offsets[2 * k];
837 /* Note: Non-found optional submatches have length -1-(-1)==0 */
838 matches[i].submatch_length[k] = offsets[2 * k + 1] - offsets[2 * k];
840 /* reserve mem for each submatch as often as it is ref'd */
841 newsize += matches[i].submatch_length[k] * job->substitute->backref_count[k];
843 /* plus replacement text size minus match text size */
844 newsize += strlen(job->substitute->text) - matches[i].submatch_length[0];
846 /* chunk before match */
847 matches[i].submatch_offset[PCRS_MAX_SUBMATCHES] = 0;
848 matches[i].submatch_length[PCRS_MAX_SUBMATCHES] = offsets[0];
849 newsize += offsets[0] * job->substitute->backref_count[PCRS_MAX_SUBMATCHES];
851 /* chunk after match */
852 matches[i].submatch_offset[PCRS_MAX_SUBMATCHES + 1] = offsets[1];
853 matches[i].submatch_length[PCRS_MAX_SUBMATCHES + 1] = subject_length - offsets[1] - 1;
854 newsize += (subject_length - offsets[1]) * job->substitute->backref_count[PCRS_MAX_SUBMATCHES + 1];
856 /* Storage for matches exhausted? -> Extend! */
857 if (++i >= max_matches)
859 max_matches = (int)(max_matches * PCRS_MAX_MATCH_GROW);
860 if (NULL == (dummy = (pcrs_match *)realloc(matches, max_matches * sizeof(pcrs_match))))
864 return(PCRS_ERR_NOMEM);
869 /* Non-global search or limit reached? */
870 if (!(job->flags & PCRS_GLOBAL)) break;
872 /* Don't loop on empty matches */
873 if (offsets[1] == offset)
874 if ((size_t)offset < subject_length)
878 /* Go find the next one */
882 /* Pass pcre error through if (bad) failure */
883 if (submatches < PCRE_ERROR_NOMATCH)
892 * Get memory for the result
894 if ((*result = (char *)malloc(newsize)) == NULL) /* must be free()d by caller */
897 return PCRS_ERR_NOMEM;
905 result_offset = *result;
907 for (i = 0; i < matches_found; i++)
909 /* copy the chunk preceding the match */
910 memcpy(result_offset, subject + offset, (size_t)matches[i].submatch_offset[0] - offset);
911 result_offset += matches[i].submatch_offset[0] - offset;
913 /* For every segment of the substitute.. */
914 for (k = 0; k <= job->substitute->backrefs; k++)
916 /* ...copy its text.. */
917 memcpy(result_offset, job->substitute->text + job->substitute->block_offset[k], job->substitute->block_length[k]);
918 result_offset += job->substitute->block_length[k];
920 /* ..plus, if it's not the last chunk, i.e.: There *is* a backref.. */
921 if (k != job->substitute->backrefs
922 /* ..in legal range.. */
923 && job->substitute->backref[k] < PCRS_MAX_SUBMATCHES + 2
924 /* ..and referencing a nonempty match.. */
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;