/[pcre]/code/trunk/pcre_dfa_exec.c
ViewVC logotype

Diff of /code/trunk/pcre_dfa_exec.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 365 by ph10, Fri Jul 11 17:06:55 2008 UTC revision 510 by ph10, Sat Mar 27 17:45:29 2010 UTC
# Line 3  Line 3 
3  *************************************************/  *************************************************/
4    
5  /* PCRE is a library of functions to support regular expressions whose syntax  /* PCRE is a library of functions to support regular expressions whose syntax
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language (but see
7    below for why this module is different).
8    
9                         Written by Philip Hazel                         Written by Philip Hazel
10             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2010 University of Cambridge
11    
12  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
13  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 44  FSM). This is NOT Perl- compatible, but Line 45  FSM). This is NOT Perl- compatible, but
45  applications. */  applications. */
46    
47    
48    /* NOTE ABOUT PERFORMANCE: A user of this function sent some code that improved
49    the performance of his patterns greatly. I could not use it as it stood, as it
50    was not thread safe, and made assumptions about pattern sizes. Also, it caused
51    test 7 to loop, and test 9 to crash with a segfault.
52    
53    The issue is the check for duplicate states, which is done by a simple linear
54    search up the state list. (Grep for "duplicate" below to find the code.) For
55    many patterns, there will never be many states active at one time, so a simple
56    linear search is fine. In patterns that have many active states, it might be a
57    bottleneck. The suggested code used an indexing scheme to remember which states
58    had previously been used for each character, and avoided the linear search when
59    it knew there was no chance of a duplicate. This was implemented when adding
60    states to the state lists.
61    
62    I wrote some thread-safe, not-limited code to try something similar at the time
63    of checking for duplicates (instead of when adding states), using index vectors
64    on the stack. It did give a 13% improvement with one specially constructed
65    pattern for certain subject strings, but on other strings and on many of the
66    simpler patterns in the test suite it did worse. The major problem, I think,
67    was the extra time to initialize the index. This had to be done for each call
68    of internal_dfa_exec(). (The supplied patch used a static vector, initialized
69    only once - I suspect this was the cause of the problems with the tests.)
70    
71    Overall, I concluded that the gains in some cases did not outweigh the losses
72    in others, so I abandoned this code. */
73    
74    
75    
76  #ifdef HAVE_CONFIG_H  #ifdef HAVE_CONFIG_H
77  #include "config.h"  #include "config.h"
78  #endif  #endif
# Line 60  applications. */ Line 89  applications. */
89  #define SP "                   "  #define SP "                   "
90    
91    
   
92  /*************************************************  /*************************************************
93  *      Code parameters and static tables         *  *      Code parameters and static tables         *
94  *************************************************/  *************************************************/
# Line 78  never stored, so we push them well clear Line 106  never stored, so we push them well clear
106    
107    
108  /* This table identifies those opcodes that are followed immediately by a  /* This table identifies those opcodes that are followed immediately by a
109  character that is to be tested in some way. This makes is possible to  character that is to be tested in some way. This makes it possible to
110  centralize the loading of these characters. In the case of Type * etc, the  centralize the loading of these characters. In the case of Type * etc, the
111  "character" is the opcode for \D, \d, \S, \s, \W, or \w, which will always be a  "character" is the opcode for \D, \d, \S, \s, \W, or \w, which will always be a
112  small value. ***NOTE*** If the start of this table is modified, the two tables  small value. Non-zero values in the table are the offsets from the opcode where
113  that follow must also be modified. */  the character is to be found. ***NOTE*** If the start of this table is
114    modified, the three tables that follow must also be modified. */
115    
116  static const uschar coptable[] = {  static const uschar coptable[] = {
117    0,                             /* End                                    */    0,                             /* End                                    */
118    0, 0, 0, 0, 0,                 /* \A, \G, \K, \B, \b                     */    0, 0, 0, 0, 0,                 /* \A, \G, \K, \B, \b                     */
119    0, 0, 0, 0, 0, 0,              /* \D, \d, \S, \s, \W, \w                 */    0, 0, 0, 0, 0, 0,              /* \D, \d, \S, \s, \W, \w                 */
120    0, 0, 0,                       /* Any, AllAny, Anybyte                   */    0, 0, 0,                       /* Any, AllAny, Anybyte                   */
121    0, 0, 0,                       /* NOTPROP, PROP, EXTUNI                  */    0, 0,                          /* \P, \p                                 */
122    0, 0, 0, 0, 0,                 /* \R, \H, \h, \V, \v                     */    0, 0, 0, 0, 0,                 /* \R, \H, \h, \V, \v                     */
123      0,                             /* \X                                     */
124    0, 0, 0, 0, 0,                 /* \Z, \z, Opt, ^, $                      */    0, 0, 0, 0, 0,                 /* \Z, \z, Opt, ^, $                      */
125    1,                             /* Char                                   */    1,                             /* Char                                   */
126    1,                             /* Charnc                                 */    1,                             /* Charnc                                 */
# Line 127  static const uschar coptable[] = { Line 157  static const uschar coptable[] = {
157    0,                             /* Reverse                                */    0,                             /* Reverse                                */
158    0, 0, 0, 0,                    /* ONCE, BRA, CBRA, COND                  */    0, 0, 0, 0,                    /* ONCE, BRA, CBRA, COND                  */
159    0, 0, 0,                       /* SBRA, SCBRA, SCOND                     */    0, 0, 0,                       /* SBRA, SCBRA, SCOND                     */
160    0,                             /* CREF                                   */    0, 0,                          /* CREF, NCREF                            */
161    0,                             /* RREF                                   */    0, 0,                          /* RREF, NRREF                            */
162    0,                             /* DEF                                    */    0,                             /* DEF                                    */
163    0, 0,                          /* BRAZERO, BRAMINZERO                    */    0, 0,                          /* BRAZERO, BRAMINZERO                    */
164    0, 0, 0, 0,                    /* PRUNE, SKIP, THEN, COMMIT              */    0, 0, 0,                       /* MARK, PRUNE, PRUNE_ARG,                */
165    0, 0, 0                        /* FAIL, ACCEPT, SKIPZERO                 */    0, 0, 0, 0,                    /* SKIP, SKIP_ARG, THEN, THEN_ARG,        */
166      0, 0, 0, 0, 0                  /* COMMIT, FAIL, ACCEPT, CLOSE, SKIPZERO  */
167    };
168    
169    /* This table identifies those opcodes that inspect a character. It is used to
170    remember the fact that a character could have been inspected when the end of
171    the subject is reached. ***NOTE*** If the start of this table is modified, the
172    two tables that follow must also be modified. */
173    
174    static const uschar poptable[] = {
175      0,                             /* End                                    */
176      0, 0, 0, 1, 1,                 /* \A, \G, \K, \B, \b                     */
177      1, 1, 1, 1, 1, 1,              /* \D, \d, \S, \s, \W, \w                 */
178      1, 1, 1,                       /* Any, AllAny, Anybyte                   */
179      1, 1,                          /* \P, \p                                 */
180      1, 1, 1, 1, 1,                 /* \R, \H, \h, \V, \v                     */
181      1,                             /* \X                                     */
182      0, 0, 0, 0, 0,                 /* \Z, \z, Opt, ^, $                      */
183      1,                             /* Char                                   */
184      1,                             /* Charnc                                 */
185      1,                             /* not                                    */
186      /* Positive single-char repeats                                          */
187      1, 1, 1, 1, 1, 1,              /* *, *?, +, +?, ?, ??                    */
188      1, 1, 1,                       /* upto, minupto, exact                   */
189      1, 1, 1, 1,                    /* *+, ++, ?+, upto+                      */
190      /* Negative single-char repeats - only for chars < 256                   */
191      1, 1, 1, 1, 1, 1,              /* NOT *, *?, +, +?, ?, ??                */
192      1, 1, 1,                       /* NOT upto, minupto, exact               */
193      1, 1, 1, 1,                    /* NOT *+, ++, ?+, upto+                  */
194      /* Positive type repeats                                                 */
195      1, 1, 1, 1, 1, 1,              /* Type *, *?, +, +?, ?, ??               */
196      1, 1, 1,                       /* Type upto, minupto, exact              */
197      1, 1, 1, 1,                    /* Type *+, ++, ?+, upto+                 */
198      /* Character class & ref repeats                                         */
199      1, 1, 1, 1, 1, 1,              /* *, *?, +, +?, ?, ??                    */
200      1, 1,                          /* CRRANGE, CRMINRANGE                    */
201      1,                             /* CLASS                                  */
202      1,                             /* NCLASS                                 */
203      1,                             /* XCLASS - variable length               */
204      0,                             /* REF                                    */
205      0,                             /* RECURSE                                */
206      0,                             /* CALLOUT                                */
207      0,                             /* Alt                                    */
208      0,                             /* Ket                                    */
209      0,                             /* KetRmax                                */
210      0,                             /* KetRmin                                */
211      0,                             /* Assert                                 */
212      0,                             /* Assert not                             */
213      0,                             /* Assert behind                          */
214      0,                             /* Assert behind not                      */
215      0,                             /* Reverse                                */
216      0, 0, 0, 0,                    /* ONCE, BRA, CBRA, COND                  */
217      0, 0, 0,                       /* SBRA, SCBRA, SCOND                     */
218      0, 0,                          /* CREF, NCREF                            */
219      0, 0,                          /* RREF, NRREF                            */
220      0,                             /* DEF                                    */
221      0, 0,                          /* BRAZERO, BRAMINZERO                    */
222      0, 0, 0,                       /* MARK, PRUNE, PRUNE_ARG,                */
223      0, 0, 0, 0,                    /* SKIP, SKIP_ARG, THEN, THEN_ARG,        */
224      0, 0, 0, 0, 0                  /* COMMIT, FAIL, ACCEPT, CLOSE, SKIPZERO  */
225  };  };
226    
227  /* These 2 tables allow for compact code for testing for \D, \d, \S, \s, \W,  /* These 2 tables allow for compact code for testing for \D, \d, \S, \s, \W,
# Line 170  typedef struct stateblock { Line 259  typedef struct stateblock {
259  #define INTS_PER_STATEBLOCK  (sizeof(stateblock)/sizeof(int))  #define INTS_PER_STATEBLOCK  (sizeof(stateblock)/sizeof(int))
260    
261    
262  #ifdef DEBUG  #ifdef PCRE_DEBUG
263  /*************************************************  /*************************************************
264  *             Print character string             *  *             Print character string             *
265  *************************************************/  *************************************************/
# Line 390  if (*first_op == OP_REVERSE) Line 479  if (*first_op == OP_REVERSE)
479      current_subject -= gone_back;      current_subject -= gone_back;
480      }      }
481    
482      /* Save the earliest consulted character */
483    
484      if (current_subject < md->start_used_ptr)
485        md->start_used_ptr = current_subject;
486    
487    /* Now we can process the individual branches. */    /* Now we can process the individual branches. */
488    
489    end_code = this_start_code;    end_code = this_start_code;
# Line 454  for (;;) Line 548  for (;;)
548    int i, j;    int i, j;
549    int clen, dlen;    int clen, dlen;
550    unsigned int c, d;    unsigned int c, d;
551      int forced_fail = 0;
552      BOOL could_continue = FALSE;
553    
554    /* Make the new state list into the active state list and empty the    /* Make the new state list into the active state list and empty the
555    new state list. */    new state list. */
# Line 467  for (;;) Line 563  for (;;)
563    workspace[0] ^= 1;              /* Remember for the restarting feature */    workspace[0] ^= 1;              /* Remember for the restarting feature */
564    workspace[1] = active_count;    workspace[1] = active_count;
565    
566  #ifdef DEBUG  #ifdef PCRE_DEBUG
567    printf("%.*sNext character: rest of subject = \"", rlevel*2-2, SP);    printf("%.*sNext character: rest of subject = \"", rlevel*2-2, SP);
568    pchars((uschar *)ptr, strlen((char *)ptr), stdout);    pchars((uschar *)ptr, strlen((char *)ptr), stdout);
569    printf("\"\n");    printf("\"\n");
# Line 511  for (;;) Line 607  for (;;)
607      stateblock *current_state = active_states + i;      stateblock *current_state = active_states + i;
608      const uschar *code;      const uschar *code;
609      int state_offset = current_state->offset;      int state_offset = current_state->offset;
610      int count, codevalue;      int count, codevalue, rrc;
611    
612  #ifdef DEBUG  #ifdef PCRE_DEBUG
613      printf ("%.*sProcessing state %d c=", rlevel*2-2, SP, state_offset);      printf ("%.*sProcessing state %d c=", rlevel*2-2, SP, state_offset);
614      if (clen == 0) printf("EOL\n");      if (clen == 0) printf("EOL\n");
615        else if (c > 32 && c < 127) printf("'%c'\n", c);        else if (c > 32 && c < 127) printf("'%c'\n", c);
# Line 543  for (;;) Line 639  for (;;)
639          }          }
640        }        }
641    
642      /* Check for a duplicate state with the same count, and skip if found. */      /* Check for a duplicate state with the same count, and skip if found.
643        See the note at the head of this module about the possibility of improving
644        performance here. */
645    
646      for (j = 0; j < i; j++)      for (j = 0; j < i; j++)
647        {        {
# Line 560  for (;;) Line 658  for (;;)
658      code = start_code + state_offset;      code = start_code + state_offset;
659      codevalue = *code;      codevalue = *code;
660    
661        /* If this opcode inspects a character, but we are at the end of the
662        subject, remember the fact for use when testing for a partial match. */
663    
664        if (clen == 0 && poptable[codevalue] != 0)
665          could_continue = TRUE;
666    
667      /* If this opcode is followed by an inline character, load it. It is      /* If this opcode is followed by an inline character, load it. It is
668      tempting to test for the presence of a subject character here, but that      tempting to test for the presence of a subject character here, but that
669      is wrong, because sometimes zero repetitions of the subject are      is wrong, because sometimes zero repetitions of the subject are
# Line 606  for (;;) Line 710  for (;;)
710    
711      switch (codevalue)      switch (codevalue)
712        {        {
713    /* ========================================================================== */
714          /* These cases are never obeyed. This is a fudge that causes a compile-
715          time error if the vectors coptable or poptable, which are indexed by
716          opcode, are not the correct length. It seems to be the only way to do
717          such a check at compile time, as the sizeof() operator does not work
718          in the C preprocessor. */
719    
720          case OP_TABLE_LENGTH:
721          case OP_TABLE_LENGTH +
722            ((sizeof(coptable) == OP_TABLE_LENGTH) &&
723             (sizeof(poptable) == OP_TABLE_LENGTH)):
724          break;
725    
726  /* ========================================================================== */  /* ========================================================================== */
727        /* Reached a closing bracket. If not at the end of the pattern, carry        /* Reached a closing bracket. If not at the end of the pattern, carry
728        on with the next opcode. Otherwise, unless we have an empty string and        on with the next opcode. Otherwise, unless we have an empty string and
729        PCRE_NOTEMPTY is set, save the match data, shifting up all previous        PCRE_NOTEMPTY is set, or PCRE_NOTEMPTY_ATSTART is set and we are at the
730          start of the subject, save the match data, shifting up all previous
731        matches so we always have the longest first. */        matches so we always have the longest first. */
732    
733        case OP_KET:        case OP_KET:
# Line 624  for (;;) Line 741  for (;;)
741            ADD_ACTIVE(state_offset - GET(code, 1), 0);            ADD_ACTIVE(state_offset - GET(code, 1), 0);
742            }            }
743          }          }
744        else if (ptr > current_subject || (md->moptions & PCRE_NOTEMPTY) == 0)        else
745          {          {
746          if (match_count < 0) match_count = (offsetcount >= 2)? 1 : 0;          if (ptr > current_subject ||
747            else if (match_count > 0 && ++match_count * 2 >= offsetcount)              ((md->moptions & PCRE_NOTEMPTY) == 0 &&
748              match_count = 0;                ((md->moptions & PCRE_NOTEMPTY_ATSTART) == 0 ||
749          count = ((match_count == 0)? offsetcount : match_count * 2) - 2;                  current_subject > start_subject + md->start_offset)))
750          if (count > 0) memmove(offsets + 2, offsets, count * sizeof(int));            {
751          if (offsetcount >= 2)            if (match_count < 0) match_count = (offsetcount >= 2)? 1 : 0;
752            {              else if (match_count > 0 && ++match_count * 2 >= offsetcount)
753            offsets[0] = current_subject - start_subject;                match_count = 0;
754            offsets[1] = ptr - start_subject;            count = ((match_count == 0)? offsetcount : match_count * 2) - 2;
755            DPRINTF(("%.*sSet matched string = \"%.*s\"\n", rlevel*2-2, SP,            if (count > 0) memmove(offsets + 2, offsets, count * sizeof(int));
756              offsets[1] - offsets[0], current_subject));            if (offsetcount >= 2)
757            }              {
758          if ((md->moptions & PCRE_DFA_SHORTEST) != 0)              offsets[0] = current_subject - start_subject;
759            {              offsets[1] = ptr - start_subject;
760            DPRINTF(("%.*sEnd of internal_dfa_exec %d: returning %d\n"              DPRINTF(("%.*sSet matched string = \"%.*s\"\n", rlevel*2-2, SP,
761              "%.*s---------------------\n\n", rlevel*2-2, SP, rlevel,                offsets[1] - offsets[0], current_subject));
762              match_count, rlevel*2-2, SP));              }
763            return match_count;            if ((md->moptions & PCRE_DFA_SHORTEST) != 0)
764                {
765                DPRINTF(("%.*sEnd of internal_dfa_exec %d: returning %d\n"
766                  "%.*s---------------------\n\n", rlevel*2-2, SP, rlevel,
767                  match_count, rlevel*2-2, SP));
768                return match_count;
769                }
770            }            }
771          }          }
772        break;        break;
# Line 757  for (;;) Line 880  for (;;)
880        if ((md->moptions & PCRE_NOTEOL) == 0)        if ((md->moptions & PCRE_NOTEOL) == 0)
881          {          {
882          if (clen == 0 ||          if (clen == 0 ||
883              (IS_NEWLINE(ptr) &&              ((md->poptions & PCRE_DOLLAR_ENDONLY) == 0 && IS_NEWLINE(ptr) &&
884                 ((ims & PCRE_MULTILINE) != 0 || ptr == end_subject - md->nllen)                 ((ims & PCRE_MULTILINE) != 0 || ptr == end_subject - md->nllen)
885              ))              ))
886            { ADD_ACTIVE(state_offset + 1, 0); }            { ADD_ACTIVE(state_offset + 1, 0); }
# Line 794  for (;;) Line 917  for (;;)
917          if (ptr > start_subject)          if (ptr > start_subject)
918            {            {
919            const uschar *temp = ptr - 1;            const uschar *temp = ptr - 1;
920              if (temp < md->start_used_ptr) md->start_used_ptr = temp;
921  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
922            if (utf8) BACKCHAR(temp);            if (utf8) BACKCHAR(temp);
923  #endif  #endif
# Line 802  for (;;) Line 926  for (;;)
926            }            }
927          else left_word = 0;          else left_word = 0;
928    
929          if (clen > 0) right_word = c < 256 && (ctypes[c] & ctype_word) != 0;          if (clen > 0)
930            else right_word = 0;            right_word = c < 256 && (ctypes[c] & ctype_word) != 0;
931            else right_word = 0;
932    
933          if ((left_word == right_word) == (codevalue == OP_NOT_WORD_BOUNDARY))          if ((left_word == right_word) == (codevalue == OP_NOT_WORD_BOUNDARY))
934            { ADD_ACTIVE(state_offset + 1, 0); }            { ADD_ACTIVE(state_offset + 1, 0); }
# Line 2157  for (;;) Line 2282  for (;;)
2282    
2283  /* ========================================================================== */  /* ========================================================================== */
2284        /* These are the opcodes for fancy brackets of various kinds. We have        /* These are the opcodes for fancy brackets of various kinds. We have
2285        to use recursion in order to handle them. The "always failing" assersion        to use recursion in order to handle them. The "always failing" assertion
2286        (?!) is optimised when compiling to OP_FAIL, so we have to support that,        (?!) is optimised to OP_FAIL when compiling, so we have to support that,
2287        though the other "backtracking verbs" are not supported. */        though the other "backtracking verbs" are not supported. */
2288    
2289        case OP_FAIL:        case OP_FAIL:
2290          forced_fail++;    /* Count FAILs for multiple states */
2291        break;        break;
2292    
2293        case OP_ASSERT:        case OP_ASSERT:
# Line 2189  for (;;) Line 2315  for (;;)
2315            rlevel,                               /* function recursion level */            rlevel,                               /* function recursion level */
2316            recursing);                           /* pass on regex recursion */            recursing);                           /* pass on regex recursion */
2317    
2318            if (rc == PCRE_ERROR_DFA_UITEM) return rc;
2319          if ((rc >= 0) == (codevalue == OP_ASSERT || codevalue == OP_ASSERTBACK))          if ((rc >= 0) == (codevalue == OP_ASSERT || codevalue == OP_ASSERTBACK))
2320              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }
2321          }          }
# Line 2200  for (;;) Line 2327  for (;;)
2327          {          {
2328          int local_offsets[1000];          int local_offsets[1000];
2329          int local_workspace[1000];          int local_workspace[1000];
2330          int condcode = code[LINK_SIZE+1];          int codelink = GET(code, 1);
2331            int condcode;
2332    
2333            /* Because of the way auto-callout works during compile, a callout item
2334            is inserted between OP_COND and an assertion condition. This does not
2335            happen for the other conditions. */
2336    
2337            if (code[LINK_SIZE+1] == OP_CALLOUT)
2338              {
2339              rrc = 0;
2340              if (pcre_callout != NULL)
2341                {
2342                pcre_callout_block cb;
2343                cb.version          = 1;   /* Version 1 of the callout block */
2344                cb.callout_number   = code[LINK_SIZE+2];
2345                cb.offset_vector    = offsets;
2346                cb.subject          = (PCRE_SPTR)start_subject;
2347                cb.subject_length   = end_subject - start_subject;
2348                cb.start_match      = current_subject - start_subject;
2349                cb.current_position = ptr - start_subject;
2350                cb.pattern_position = GET(code, LINK_SIZE + 3);
2351                cb.next_item_length = GET(code, 3 + 2*LINK_SIZE);
2352                cb.capture_top      = 1;
2353                cb.capture_last     = -1;
2354                cb.callout_data     = md->callout_data;
2355                if ((rrc = (*pcre_callout)(&cb)) < 0) return rrc;   /* Abandon */
2356                }
2357              if (rrc > 0) break;                      /* Fail this thread */
2358              code += _pcre_OP_lengths[OP_CALLOUT];    /* Skip callout data */
2359              }
2360    
2361            condcode = code[LINK_SIZE+1];
2362    
2363          /* Back reference conditions are not supported */          /* Back reference conditions are not supported */
2364    
2365          if (condcode == OP_CREF) return PCRE_ERROR_DFA_UCOND;          if (condcode == OP_CREF || condcode == OP_NCREF)
2366              return PCRE_ERROR_DFA_UCOND;
2367    
2368          /* The DEFINE condition is always false */          /* The DEFINE condition is always false */
2369    
2370          if (condcode == OP_DEF)          if (condcode == OP_DEF)
2371            {            { ADD_ACTIVE(state_offset + codelink + LINK_SIZE + 1, 0); }
           ADD_ACTIVE(state_offset + GET(code, 1) + LINK_SIZE + 1, 0);  
           }  
2372    
2373          /* The only supported version of OP_RREF is for the value RREF_ANY,          /* The only supported version of OP_RREF is for the value RREF_ANY,
2374          which means "test if in any recursion". We can't test for specifically          which means "test if in any recursion". We can't test for specifically
2375          recursed groups. */          recursed groups. */
2376    
2377          else if (condcode == OP_RREF)          else if (condcode == OP_RREF || condcode == OP_NRREF)
2378            {            {
2379            int value = GET2(code, LINK_SIZE+2);            int value = GET2(code, LINK_SIZE+2);
2380            if (value != RREF_ANY) return PCRE_ERROR_DFA_UCOND;            if (value != RREF_ANY) return PCRE_ERROR_DFA_UCOND;
2381            if (recursing > 0) { ADD_ACTIVE(state_offset + LINK_SIZE + 4, 0); }            if (recursing > 0)
2382              else { ADD_ACTIVE(state_offset + GET(code, 1) + LINK_SIZE + 1, 0); }              { ADD_ACTIVE(state_offset + LINK_SIZE + 4, 0); }
2383              else { ADD_ACTIVE(state_offset + codelink + LINK_SIZE + 1, 0); }
2384            }            }
2385    
2386          /* Otherwise, the condition is an assertion */          /* Otherwise, the condition is an assertion */
# Line 2248  for (;;) Line 2406  for (;;)
2406              rlevel,                               /* function recursion level */              rlevel,                               /* function recursion level */
2407              recursing);                           /* pass on regex recursion */              recursing);                           /* pass on regex recursion */
2408    
2409              if (rc == PCRE_ERROR_DFA_UITEM) return rc;
2410            if ((rc >= 0) ==            if ((rc >= 0) ==
2411                  (condcode == OP_ASSERT || condcode == OP_ASSERTBACK))                  (condcode == OP_ASSERT || condcode == OP_ASSERTBACK))
2412              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }
2413            else            else
2414              { ADD_ACTIVE(state_offset + GET(code, 1) + LINK_SIZE + 1, 0); }              { ADD_ACTIVE(state_offset + codelink + LINK_SIZE + 1, 0); }
2415            }            }
2416          }          }
2417        break;        break;
# Line 2404  for (;;) Line 2563  for (;;)
2563        /* Handle callouts */        /* Handle callouts */
2564    
2565        case OP_CALLOUT:        case OP_CALLOUT:
2566          rrc = 0;
2567        if (pcre_callout != NULL)        if (pcre_callout != NULL)
2568          {          {
         int rrc;  
2569          pcre_callout_block cb;          pcre_callout_block cb;
2570          cb.version          = 1;   /* Version 1 of the callout block */          cb.version          = 1;   /* Version 1 of the callout block */
2571          cb.callout_number   = code[1];          cb.callout_number   = code[1];
# Line 2421  for (;;) Line 2580  for (;;)
2580          cb.capture_last     = -1;          cb.capture_last     = -1;
2581          cb.callout_data     = md->callout_data;          cb.callout_data     = md->callout_data;
2582          if ((rrc = (*pcre_callout)(&cb)) < 0) return rrc;   /* Abandon */          if ((rrc = (*pcre_callout)(&cb)) < 0) return rrc;   /* Abandon */
         if (rrc == 0) { ADD_ACTIVE(state_offset + 2 + 2*LINK_SIZE, 0); }  
2583          }          }
2584          if (rrc == 0)
2585            { ADD_ACTIVE(state_offset + _pcre_OP_lengths[OP_CALLOUT], 0); }
2586        break;        break;
2587    
2588    
# Line 2438  for (;;) Line 2598  for (;;)
2598    /* We have finished the processing at the current subject character. If no    /* We have finished the processing at the current subject character. If no
2599    new states have been set for the next character, we have found all the    new states have been set for the next character, we have found all the
2600    matches that we are going to find. If we are at the top level and partial    matches that we are going to find. If we are at the top level and partial
2601    matching has been requested, check for appropriate conditions. */    matching has been requested, check for appropriate conditions.
2602    
2603      The "forced_ fail" variable counts the number of (*F) encountered for the
2604      character. If it is equal to the original active_count (saved in
2605      workspace[1]) it means that (*F) was found on every active state. In this
2606      case we don't want to give a partial match.
2607    
2608      The "could_continue" variable is true if a state could have continued but
2609      for the fact that the end of the subject was reached. */
2610    
2611    if (new_count <= 0)    if (new_count <= 0)
2612      {      {
2613      if (match_count < 0 &&                     /* No matches found */      if (rlevel == 1 &&                               /* Top level, and */
2614          rlevel == 1 &&                         /* Top level match function */          could_continue &&                            /* Some could go on */
2615          (md->moptions & PCRE_PARTIAL) != 0 &&  /* Want partial matching */          forced_fail != workspace[1] &&               /* Not all forced fail & */
2616          ptr >= end_subject &&                  /* Reached end of subject */          (                                            /* either... */
2617          ptr > current_subject)                 /* Matched non-empty string */          (md->moptions & PCRE_PARTIAL_HARD) != 0      /* Hard partial */
2618            ||                                           /* or... */
2619            ((md->moptions & PCRE_PARTIAL_SOFT) != 0 &&  /* Soft partial and */
2620             match_count < 0)                            /* no matches */
2621            ) &&                                         /* And... */
2622            ptr >= end_subject &&                     /* Reached end of subject */
2623            ptr > current_subject)                    /* Matched non-empty string */
2624        {        {
2625        if (offsetcount >= 2)        if (offsetcount >= 2)
2626          {          {
2627          offsets[0] = current_subject - start_subject;          offsets[0] = md->start_used_ptr - start_subject;
2628          offsets[1] = end_subject - start_subject;          offsets[1] = end_subject - start_subject;
2629          }          }
2630        match_count = PCRE_ERROR_PARTIAL;        match_count = PCRE_ERROR_PARTIAL;
# Line 2592  md->start_code = (const uschar *)argumen Line 2766  md->start_code = (const uschar *)argumen
2766      re->name_table_offset + re->name_count * re->name_entry_size;      re->name_table_offset + re->name_count * re->name_entry_size;
2767  md->start_subject = (const unsigned char *)subject;  md->start_subject = (const unsigned char *)subject;
2768  md->end_subject = end_subject;  md->end_subject = end_subject;
2769    md->start_offset = start_offset;
2770  md->moptions = options;  md->moptions = options;
2771  md->poptions = re->options;  md->poptions = re->options;
2772    
# Line 2614  switch ((((options & PCRE_NEWLINE_BITS) Line 2789  switch ((((options & PCRE_NEWLINE_BITS)
2789           PCRE_NEWLINE_BITS)           PCRE_NEWLINE_BITS)
2790    {    {
2791    case 0: newline = NEWLINE; break;   /* Compile-time default */    case 0: newline = NEWLINE; break;   /* Compile-time default */
2792    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
2793    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
2794    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
2795         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;         PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
2796    case PCRE_NEWLINE_ANY: newline = -1; break;    case PCRE_NEWLINE_ANY: newline = -1; break;
2797    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
2798    default: return PCRE_ERROR_BADNEWLINE;    default: return PCRE_ERROR_BADNEWLINE;
# Line 2696  if (!anchored) Line 2871  if (!anchored)
2871      }      }
2872    else    else
2873      {      {
2874      if (startline && study != NULL &&      if (!startline && study != NULL &&
2875           (study->options & PCRE_STUDY_MAPPED) != 0)           (study->flags & PCRE_STUDY_MAPPED) != 0)
2876        start_bits = study->start_bits;        start_bits = study->start_bits;
2877      }      }
2878    }    }
# Line 2713  if ((re->flags & PCRE_REQCHSET) != 0) Line 2888  if ((re->flags & PCRE_REQCHSET) != 0)
2888    }    }
2889    
2890  /* Call the main matching function, looping for a non-anchored regex after a  /* Call the main matching function, looping for a non-anchored regex after a
2891  failed match. Unless restarting, optimize by moving to the first match  failed match. If not restarting, perform certain optimizations at the start of
2892  character if possible, when not anchored. Then unless wanting a partial match,  a match. */
 check for a required later character. */  
2893    
2894  for (;;)  for (;;)
2895    {    {
# Line 2725  for (;;) Line 2899  for (;;)
2899      {      {
2900      const uschar *save_end_subject = end_subject;      const uschar *save_end_subject = end_subject;
2901    
2902      /* Advance to a unique first char if possible. If firstline is TRUE, the      /* If firstline is TRUE, the start of the match is constrained to the first
2903      start of the match is constrained to the first line of a multiline string.      line of a multiline string. Implement this by temporarily adjusting
2904      Implement this by temporarily adjusting end_subject so that we stop      end_subject so that we stop scanning at a newline. If the match fails at
2905      scanning at a newline. If the match fails at the newline, later code breaks      the newline, later code breaks this loop. */
     this loop. */  
2906    
2907      if (firstline)      if (firstline)
2908        {        {
2909        USPTR t = current_subject;        USPTR t = current_subject;
2910  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2911        if (utf8)        if (utf8)
2912          {          {
2913          while (t < md->end_subject && !IS_NEWLINE(t))          while (t < md->end_subject && !IS_NEWLINE(t))
2914            {            {
2915            t++;            t++;
2916            while (t < end_subject && (*t & 0xc0) == 0x80) t++;            while (t < end_subject && (*t & 0xc0) == 0x80) t++;
2917            }            }
2918          }          }
2919        else        else
2920  #endif  #endif
2921        while (t < md->end_subject && !IS_NEWLINE(t)) t++;        while (t < md->end_subject && !IS_NEWLINE(t)) t++;
2922        end_subject = t;        end_subject = t;
2923        }        }
2924    
2925      if (first_byte >= 0)      /* There are some optimizations that avoid running the match if a known
2926        starting point is not found. However, there is an option that disables
2927        these, for testing and for ensuring that all callouts do actually occur. */
2928    
2929        if ((options & PCRE_NO_START_OPTIMIZE) == 0)
2930        {        {
2931        if (first_byte_caseless)        /* Advance to a known first byte. */
2932          while (current_subject < end_subject &&  
2933                 lcc[*current_subject] != first_byte)        if (first_byte >= 0)
2934            current_subject++;          {
2935        else          if (first_byte_caseless)
2936          while (current_subject < end_subject && *current_subject != first_byte)            while (current_subject < end_subject &&
2937            current_subject++;                   lcc[*current_subject] != first_byte)
2938        }              current_subject++;
2939            else
2940              while (current_subject < end_subject &&
2941                     *current_subject != first_byte)
2942                current_subject++;
2943            }
2944    
2945      /* Or to just after a linebreak for a multiline match if possible */        /* Or to just after a linebreak for a multiline match if possible */
2946    
2947      else if (startline)        else if (startline)
       {  
       if (current_subject > md->start_subject + start_offset)  
2948          {          {
2949  #ifdef SUPPORT_UTF8          if (current_subject > md->start_subject + start_offset)
         if (utf8)  
2950            {            {
2951            while (current_subject < end_subject && !WAS_NEWLINE(current_subject))  #ifdef SUPPORT_UTF8
2952              if (utf8)
2953              {              {
2954              current_subject++;              while (current_subject < end_subject &&
2955              while(current_subject < end_subject &&                     !WAS_NEWLINE(current_subject))
2956                    (*current_subject & 0xc0) == 0x80)                {
2957                current_subject++;                current_subject++;
2958              }                while(current_subject < end_subject &&
2959                        (*current_subject & 0xc0) == 0x80)
2960                    current_subject++;
2961                  }
2962                }
2963              else
2964    #endif
2965              while (current_subject < end_subject && !WAS_NEWLINE(current_subject))
2966                current_subject++;
2967    
2968              /* If we have just passed a CR and the newline option is ANY or
2969              ANYCRLF, and we are now at a LF, advance the match position by one
2970              more character. */
2971    
2972              if (current_subject[-1] == CHAR_CR &&
2973                   (md->nltype == NLTYPE_ANY || md->nltype == NLTYPE_ANYCRLF) &&
2974                   current_subject < end_subject &&
2975                   *current_subject == CHAR_NL)
2976                current_subject++;
2977            }            }
         else  
 #endif  
         while (current_subject < end_subject && !WAS_NEWLINE(current_subject))  
           current_subject++;  
   
         /* If we have just passed a CR and the newline option is ANY or  
         ANYCRLF, and we are now at a LF, advance the match position by one more  
         character. */  
   
         if (current_subject[-1] == '\r' &&  
              (md->nltype == NLTYPE_ANY || md->nltype == NLTYPE_ANYCRLF) &&  
              current_subject < end_subject &&  
              *current_subject == '\n')  
           current_subject++;  
2978          }          }
       }  
2979    
2980      /* Or to a non-unique first char after study */        /* Or to a non-unique first char after study */
2981    
2982      else if (start_bits != NULL)        else if (start_bits != NULL)
       {  
       while (current_subject < end_subject)  
2983          {          {
2984          register unsigned int c = *current_subject;          while (current_subject < end_subject)
2985          if ((start_bits[c/8] & (1 << (c&7))) == 0) current_subject++;            {
2986            else break;            register unsigned int c = *current_subject;
2987              if ((start_bits[c/8] & (1 << (c&7))) == 0) current_subject++;
2988                else break;
2989              }
2990          }          }
2991        }        }
2992    
2993      /* Restore fudged end_subject */      /* Restore fudged end_subject */
2994    
2995      end_subject = save_end_subject;      end_subject = save_end_subject;
     }  
   
   /* If req_byte is set, we know that that character must appear in the subject  
   for the match to succeed. If the first character is set, req_byte must be  
   later in the subject; otherwise the test starts at the match point. This  
   optimization can save a huge amount of work in patterns with nested unlimited  
   repeats that aren't going to match. Writing separate code for cased/caseless  
   versions makes it go faster, as does using an autoincrement and backing off  
   on a match.  
   
   HOWEVER: when the subject string is very, very long, searching to its end can  
   take a long time, and give bad performance on quite ordinary patterns. This  
   showed up when somebody was matching /^C/ on a 32-megabyte string... so we  
   don't do this when the string is sufficiently long.  
   
   ALSO: this processing is disabled when partial matching is requested.  
   */  
   
   if (req_byte >= 0 &&  
       end_subject - current_subject < REQ_BYTE_MAX &&  
       (options & PCRE_PARTIAL) == 0)  
     {  
     register const uschar *p = current_subject + ((first_byte >= 0)? 1 : 0);  
2996    
2997      /* We don't need to repeat the search if we haven't yet reached the      /* The following two optimizations are disabled for partial matching or if
2998      place we found it at last time. */      disabling is explicitly requested (and of course, by the test above, this
2999        code is not obeyed when restarting after a partial match). */
3000    
3001      if (p > req_byte_ptr)      if ((options & PCRE_NO_START_OPTIMIZE) == 0 &&
3002            (options & (PCRE_PARTIAL_HARD|PCRE_PARTIAL_SOFT)) == 0)
3003        {        {
3004        if (req_byte_caseless)        /* If the pattern was studied, a minimum subject length may be set. This
3005          {        is a lower bound; no actual string of that length may actually match the
3006          while (p < end_subject)        pattern. Although the value is, strictly, in characters, we treat it as
3007            {        bytes to avoid spending too much time in this optimization. */
3008            register int pp = *p++;  
3009            if (pp == req_byte || pp == req_byte2) { p--; break; }        if (study != NULL && (study->flags & PCRE_STUDY_MINLEN) != 0 &&
3010            }            (pcre_uint32)(end_subject - current_subject) < study->minlength)
3011          }          return PCRE_ERROR_NOMATCH;
3012        else  
3013          /* If req_byte is set, we know that that character must appear in the
3014          subject for the match to succeed. If the first character is set, req_byte
3015          must be later in the subject; otherwise the test starts at the match
3016          point. This optimization can save a huge amount of work in patterns with
3017          nested unlimited repeats that aren't going to match. Writing separate
3018          code for cased/caseless versions makes it go faster, as does using an
3019          autoincrement and backing off on a match.
3020    
3021          HOWEVER: when the subject string is very, very long, searching to its end
3022          can take a long time, and give bad performance on quite ordinary
3023          patterns. This showed up when somebody was matching /^C/ on a 32-megabyte
3024          string... so we don't do this when the string is sufficiently long. */
3025    
3026          if (req_byte >= 0 && end_subject - current_subject < REQ_BYTE_MAX)
3027          {          {
3028          while (p < end_subject)          register const uschar *p = current_subject + ((first_byte >= 0)? 1 : 0);
3029    
3030            /* We don't need to repeat the search if we haven't yet reached the
3031            place we found it at last time. */
3032    
3033            if (p > req_byte_ptr)
3034            {            {
3035            if (*p++ == req_byte) { p--; break; }            if (req_byte_caseless)
3036            }              {
3037          }              while (p < end_subject)
3038                  {
3039                  register int pp = *p++;
3040                  if (pp == req_byte || pp == req_byte2) { p--; break; }
3041                  }
3042                }
3043              else
3044                {
3045                while (p < end_subject)
3046                  {
3047                  if (*p++ == req_byte) { p--; break; }
3048                  }
3049                }
3050    
3051        /* If we can't find the required character, break the matching loop,            /* If we can't find the required character, break the matching loop,
3052        which will cause a return or PCRE_ERROR_NOMATCH. */            which will cause a return or PCRE_ERROR_NOMATCH. */
3053    
3054        if (p >= end_subject) break;            if (p >= end_subject) break;
3055    
3056        /* If we have found the required character, save the point where we            /* If we have found the required character, save the point where we
3057        found it, so that we don't search again next time round the loop if            found it, so that we don't search again next time round the loop if
3058        the start hasn't passed this character yet. */            the start hasn't passed this character yet. */
3059    
3060        req_byte_ptr = p;            req_byte_ptr = p;
3061              }
3062            }
3063        }        }
3064      }      }   /* End of optimizations that are done when not restarting */
3065    
3066    /* OK, now we can do the business */    /* OK, now we can do the business */
3067    
3068      md->start_used_ptr = current_subject;
3069    
3070    rc = internal_dfa_exec(    rc = internal_dfa_exec(
3071      md,                                /* fixed match data */      md,                                /* fixed match data */
3072      md->start_code,                    /* this subexpression's code */      md->start_code,                    /* this subexpression's code */
# Line 2903  for (;;) Line 3101  for (;;)
3101    not contain any explicit matches for \r or \n, and the newline option is CRLF    not contain any explicit matches for \r or \n, and the newline option is CRLF
3102    or ANY or ANYCRLF, advance the match position by one more character. */    or ANY or ANYCRLF, advance the match position by one more character. */
3103    
3104    if (current_subject[-1] == '\r' &&    if (current_subject[-1] == CHAR_CR &&
3105        current_subject < end_subject &&        current_subject < end_subject &&
3106        *current_subject == '\n' &&        *current_subject == CHAR_NL &&
3107        (re->flags & PCRE_HASCRORLF) == 0 &&        (re->flags & PCRE_HASCRORLF) == 0 &&
3108          (md->nltype == NLTYPE_ANY ||          (md->nltype == NLTYPE_ANY ||
3109           md->nltype == NLTYPE_ANYCRLF ||           md->nltype == NLTYPE_ANYCRLF ||

Legend:
Removed from v.365  
changed lines
  Added in v.510

  ViewVC Help
Powered by ViewVC 1.1.5