/[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 426 by ph10, Wed Aug 26 15:38:32 2009 UTC revision 498 by ph10, Wed Mar 3 19:29:38 2010 UTC
# Line 7  and semantics are as close as possible t Line 7  and semantics are as close as possible t
7  below for why this module is different).  below for why this module is different).
8    
9                         Written by Philip Hazel                         Written by Philip Hazel
10             Copyright (c) 1997-2009 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 45  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 81  never stored, so we push them well clear Line 109  never stored, so we push them well clear
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 is 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                                    */
163      0, 0,                          /* BRAZERO, BRAMINZERO                    */
164      0, 0, 0, 0,                    /* PRUNE, SKIP, THEN, COMMIT              */
165      0, 0, 0, 0                     /* FAIL, ACCEPT, CLOSE, SKIPZERO          */
166    };
167    
168    /* This table identifies those opcodes that inspect a character. It is used to
169    remember the fact that a character could have been inspected when the end of
170    the subject is reached. ***NOTE*** If the start of this table is modified, the
171    two tables that follow must also be modified. */
172    
173    static const uschar poptable[] = {
174      0,                             /* End                                    */
175      0, 0, 0, 1, 1,                 /* \A, \G, \K, \B, \b                     */
176      1, 1, 1, 1, 1, 1,              /* \D, \d, \S, \s, \W, \w                 */
177      1, 1, 1,                       /* Any, AllAny, Anybyte                   */
178      1, 1,                          /* \P, \p                                 */
179      1, 1, 1, 1, 1,                 /* \R, \H, \h, \V, \v                     */
180      1,                             /* \X                                     */
181      0, 0, 0, 0, 0,                 /* \Z, \z, Opt, ^, $                      */
182      1,                             /* Char                                   */
183      1,                             /* Charnc                                 */
184      1,                             /* not                                    */
185      /* Positive single-char repeats                                          */
186      1, 1, 1, 1, 1, 1,              /* *, *?, +, +?, ?, ??                    */
187      1, 1, 1,                       /* upto, minupto, exact                   */
188      1, 1, 1, 1,                    /* *+, ++, ?+, upto+                      */
189      /* Negative single-char repeats - only for chars < 256                   */
190      1, 1, 1, 1, 1, 1,              /* NOT *, *?, +, +?, ?, ??                */
191      1, 1, 1,                       /* NOT upto, minupto, exact               */
192      1, 1, 1, 1,                    /* NOT *+, ++, ?+, upto+                  */
193      /* Positive type repeats                                                 */
194      1, 1, 1, 1, 1, 1,              /* Type *, *?, +, +?, ?, ??               */
195      1, 1, 1,                       /* Type upto, minupto, exact              */
196      1, 1, 1, 1,                    /* Type *+, ++, ?+, upto+                 */
197      /* Character class & ref repeats                                         */
198      1, 1, 1, 1, 1, 1,              /* *, *?, +, +?, ?, ??                    */
199      1, 1,                          /* CRRANGE, CRMINRANGE                    */
200      1,                             /* CLASS                                  */
201      1,                             /* NCLASS                                 */
202      1,                             /* XCLASS - variable length               */
203      0,                             /* REF                                    */
204      0,                             /* RECURSE                                */
205      0,                             /* CALLOUT                                */
206      0,                             /* Alt                                    */
207      0,                             /* Ket                                    */
208      0,                             /* KetRmax                                */
209      0,                             /* KetRmin                                */
210      0,                             /* Assert                                 */
211      0,                             /* Assert not                             */
212      0,                             /* Assert behind                          */
213      0,                             /* Assert behind not                      */
214      0,                             /* Reverse                                */
215      0, 0, 0, 0,                    /* ONCE, BRA, CBRA, COND                  */
216      0, 0, 0,                       /* SBRA, SCBRA, SCOND                     */
217      0, 0,                          /* CREF, NCREF                            */
218      0, 0,                          /* RREF, NRREF                            */
219    0,                             /* DEF                                    */    0,                             /* DEF                                    */
220    0, 0,                          /* BRAZERO, BRAMINZERO                    */    0, 0,                          /* BRAZERO, BRAMINZERO                    */
221    0, 0, 0, 0,                    /* PRUNE, SKIP, THEN, COMMIT              */    0, 0, 0, 0,                    /* PRUNE, SKIP, THEN, COMMIT              */
222    0, 0, 0                        /* FAIL, ACCEPT, SKIPZERO                 */    0, 0, 0, 0                     /* FAIL, ACCEPT, CLOSE, SKIPZERO          */
223  };  };
224    
225  /* 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 257  typedef struct stateblock {
257  #define INTS_PER_STATEBLOCK  (sizeof(stateblock)/sizeof(int))  #define INTS_PER_STATEBLOCK  (sizeof(stateblock)/sizeof(int))
258    
259    
260  #ifdef DEBUG  #ifdef PCRE_DEBUG
261  /*************************************************  /*************************************************
262  *             Print character string             *  *             Print character string             *
263  *************************************************/  *************************************************/
# Line 390  if (*first_op == OP_REVERSE) Line 477  if (*first_op == OP_REVERSE)
477      current_subject -= gone_back;      current_subject -= gone_back;
478      }      }
479    
480      /* Save the earliest consulted character */
481    
482      if (current_subject < md->start_used_ptr)
483        md->start_used_ptr = current_subject;
484    
485    /* Now we can process the individual branches. */    /* Now we can process the individual branches. */
486    
487    end_code = this_start_code;    end_code = this_start_code;
# Line 454  for (;;) Line 546  for (;;)
546    int i, j;    int i, j;
547    int clen, dlen;    int clen, dlen;
548    unsigned int c, d;    unsigned int c, d;
549      int forced_fail = 0;
550      BOOL could_continue = FALSE;
551    
552    /* 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
553    new state list. */    new state list. */
# Line 467  for (;;) Line 561  for (;;)
561    workspace[0] ^= 1;              /* Remember for the restarting feature */    workspace[0] ^= 1;              /* Remember for the restarting feature */
562    workspace[1] = active_count;    workspace[1] = active_count;
563    
564  #ifdef DEBUG  #ifdef PCRE_DEBUG
565    printf("%.*sNext character: rest of subject = \"", rlevel*2-2, SP);    printf("%.*sNext character: rest of subject = \"", rlevel*2-2, SP);
566    pchars((uschar *)ptr, strlen((char *)ptr), stdout);    pchars((uschar *)ptr, strlen((char *)ptr), stdout);
567    printf("\"\n");    printf("\"\n");
# Line 513  for (;;) Line 607  for (;;)
607      int state_offset = current_state->offset;      int state_offset = current_state->offset;
608      int count, codevalue, rrc;      int count, codevalue, rrc;
609    
610  #ifdef DEBUG  #ifdef PCRE_DEBUG
611      printf ("%.*sProcessing state %d c=", rlevel*2-2, SP, state_offset);      printf ("%.*sProcessing state %d c=", rlevel*2-2, SP, state_offset);
612      if (clen == 0) printf("EOL\n");      if (clen == 0) printf("EOL\n");
613        else if (c > 32 && c < 127) printf("'%c'\n", c);        else if (c > 32 && c < 127) printf("'%c'\n", c);
# Line 543  for (;;) Line 637  for (;;)
637          }          }
638        }        }
639    
640      /* 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.
641        See the note at the head of this module about the possibility of improving
642        performance here. */
643    
644      for (j = 0; j < i; j++)      for (j = 0; j < i; j++)
645        {        {
# Line 560  for (;;) Line 656  for (;;)
656      code = start_code + state_offset;      code = start_code + state_offset;
657      codevalue = *code;      codevalue = *code;
658    
659        /* If this opcode inspects a character, but we are at the end of the
660        subject, remember the fact for use when testing for a partial match. */
661    
662        if (clen == 0 && poptable[codevalue] != 0)
663          could_continue = TRUE;
664    
665      /* 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
666      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
667      is wrong, because sometimes zero repetitions of the subject are      is wrong, because sometimes zero repetitions of the subject are
# Line 606  for (;;) Line 708  for (;;)
708    
709      switch (codevalue)      switch (codevalue)
710        {        {
711    /* ========================================================================== */
712          /* These cases are never obeyed. This is a fudge that causes a compile-
713          time error if the vectors coptable or poptable, which are indexed by
714          opcode, are not the correct length. It seems to be the only way to do
715          such a check at compile time, as the sizeof() operator does not work
716          in the C preprocessor. */
717    
718          case OP_TABLE_LENGTH:
719          case OP_TABLE_LENGTH +
720            ((sizeof(coptable) == OP_TABLE_LENGTH) &&
721             (sizeof(poptable) == OP_TABLE_LENGTH)):
722          break;
723    
724  /* ========================================================================== */  /* ========================================================================== */
725        /* 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
726        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
727        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
728          start of the subject, save the match data, shifting up all previous
729        matches so we always have the longest first. */        matches so we always have the longest first. */
730    
731        case OP_KET:        case OP_KET:
# Line 624  for (;;) Line 739  for (;;)
739            ADD_ACTIVE(state_offset - GET(code, 1), 0);            ADD_ACTIVE(state_offset - GET(code, 1), 0);
740            }            }
741          }          }
742        else if (ptr > current_subject || (md->moptions & PCRE_NOTEMPTY) == 0)        else
743          {          {
744          if (match_count < 0) match_count = (offsetcount >= 2)? 1 : 0;          if (ptr > current_subject ||
745            else if (match_count > 0 && ++match_count * 2 >= offsetcount)              ((md->moptions & PCRE_NOTEMPTY) == 0 &&
746              match_count = 0;                ((md->moptions & PCRE_NOTEMPTY_ATSTART) == 0 ||
747          count = ((match_count == 0)? offsetcount : match_count * 2) - 2;                  current_subject > start_subject + md->start_offset)))
748          if (count > 0) memmove(offsets + 2, offsets, count * sizeof(int));            {
749          if (offsetcount >= 2)            if (match_count < 0) match_count = (offsetcount >= 2)? 1 : 0;
750            {              else if (match_count > 0 && ++match_count * 2 >= offsetcount)
751            offsets[0] = current_subject - start_subject;                match_count = 0;
752            offsets[1] = ptr - start_subject;            count = ((match_count == 0)? offsetcount : match_count * 2) - 2;
753            DPRINTF(("%.*sSet matched string = \"%.*s\"\n", rlevel*2-2, SP,            if (count > 0) memmove(offsets + 2, offsets, count * sizeof(int));
754              offsets[1] - offsets[0], current_subject));            if (offsetcount >= 2)
755            }              {
756          if ((md->moptions & PCRE_DFA_SHORTEST) != 0)              offsets[0] = current_subject - start_subject;
757            {              offsets[1] = ptr - start_subject;
758            DPRINTF(("%.*sEnd of internal_dfa_exec %d: returning %d\n"              DPRINTF(("%.*sSet matched string = \"%.*s\"\n", rlevel*2-2, SP,
759              "%.*s---------------------\n\n", rlevel*2-2, SP, rlevel,                offsets[1] - offsets[0], current_subject));
760              match_count, rlevel*2-2, SP));              }
761            return match_count;            if ((md->moptions & PCRE_DFA_SHORTEST) != 0)
762                {
763                DPRINTF(("%.*sEnd of internal_dfa_exec %d: returning %d\n"
764                  "%.*s---------------------\n\n", rlevel*2-2, SP, rlevel,
765                  match_count, rlevel*2-2, SP));
766                return match_count;
767                }
768            }            }
769          }          }
770        break;        break;
# Line 794  for (;;) Line 915  for (;;)
915          if (ptr > start_subject)          if (ptr > start_subject)
916            {            {
917            const uschar *temp = ptr - 1;            const uschar *temp = ptr - 1;
918              if (temp < md->start_used_ptr) md->start_used_ptr = temp;
919  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
920            if (utf8) BACKCHAR(temp);            if (utf8) BACKCHAR(temp);
921  #endif  #endif
# Line 802  for (;;) Line 924  for (;;)
924            }            }
925          else left_word = 0;          else left_word = 0;
926    
927          if (clen > 0) right_word = c < 256 && (ctypes[c] & ctype_word) != 0;          if (clen > 0)
928            else right_word = 0;            right_word = c < 256 && (ctypes[c] & ctype_word) != 0;
929            else right_word = 0;
930    
931          if ((left_word == right_word) == (codevalue == OP_NOT_WORD_BOUNDARY))          if ((left_word == right_word) == (codevalue == OP_NOT_WORD_BOUNDARY))
932            { ADD_ACTIVE(state_offset + 1, 0); }            { ADD_ACTIVE(state_offset + 1, 0); }
# Line 2162  for (;;) Line 2285  for (;;)
2285        though the other "backtracking verbs" are not supported. */        though the other "backtracking verbs" are not supported. */
2286    
2287        case OP_FAIL:        case OP_FAIL:
2288          forced_fail++;    /* Count FAILs for multiple states */
2289        break;        break;
2290    
2291        case OP_ASSERT:        case OP_ASSERT:
# Line 2189  for (;;) Line 2313  for (;;)
2313            rlevel,                               /* function recursion level */            rlevel,                               /* function recursion level */
2314            recursing);                           /* pass on regex recursion */            recursing);                           /* pass on regex recursion */
2315    
2316            if (rc == PCRE_ERROR_DFA_UITEM) return rc;
2317          if ((rc >= 0) == (codevalue == OP_ASSERT || codevalue == OP_ASSERTBACK))          if ((rc >= 0) == (codevalue == OP_ASSERT || codevalue == OP_ASSERTBACK))
2318              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }
2319          }          }
# Line 2235  for (;;) Line 2360  for (;;)
2360    
2361          /* Back reference conditions are not supported */          /* Back reference conditions are not supported */
2362    
2363          if (condcode == OP_CREF) return PCRE_ERROR_DFA_UCOND;          if (condcode == OP_CREF || condcode == OP_NCREF)
2364              return PCRE_ERROR_DFA_UCOND;
2365    
2366          /* The DEFINE condition is always false */          /* The DEFINE condition is always false */
2367    
# Line 2246  for (;;) Line 2372  for (;;)
2372          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
2373          recursed groups. */          recursed groups. */
2374    
2375          else if (condcode == OP_RREF)          else if (condcode == OP_RREF || condcode == OP_NRREF)
2376            {            {
2377            int value = GET2(code, LINK_SIZE+2);            int value = GET2(code, LINK_SIZE+2);
2378            if (value != RREF_ANY) return PCRE_ERROR_DFA_UCOND;            if (value != RREF_ANY) return PCRE_ERROR_DFA_UCOND;
# Line 2278  for (;;) Line 2404  for (;;)
2404              rlevel,                               /* function recursion level */              rlevel,                               /* function recursion level */
2405              recursing);                           /* pass on regex recursion */              recursing);                           /* pass on regex recursion */
2406    
2407              if (rc == PCRE_ERROR_DFA_UITEM) return rc;
2408            if ((rc >= 0) ==            if ((rc >= 0) ==
2409                  (condcode == OP_ASSERT || condcode == OP_ASSERTBACK))                  (condcode == OP_ASSERT || condcode == OP_ASSERTBACK))
2410              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }
# Line 2469  for (;;) Line 2596  for (;;)
2596    /* We have finished the processing at the current subject character. If no    /* We have finished the processing at the current subject character. If no
2597    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
2598    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
2599    matching has been requested, check for appropriate conditions. */    matching has been requested, check for appropriate conditions.
2600    
2601      The "forced_ fail" variable counts the number of (*F) encountered for the
2602      character. If it is equal to the original active_count (saved in
2603      workspace[1]) it means that (*F) was found on every active state. In this
2604      case we don't want to give a partial match.
2605    
2606      The "could_continue" variable is true if a state could have continued but
2607      for the fact that the end of the subject was reached. */
2608    
2609    if (new_count <= 0)    if (new_count <= 0)
2610      {      {
2611      if (match_count < 0 &&                     /* No matches found */      if (rlevel == 1 &&                               /* Top level, and */
2612          rlevel == 1 &&                         /* Top level match function */          could_continue &&                            /* Some could go on */
2613          (md->moptions & PCRE_PARTIAL) != 0 &&  /* Want partial matching */          forced_fail != workspace[1] &&               /* Not all forced fail & */
2614          ptr >= end_subject &&                  /* Reached end of subject */          (                                            /* either... */
2615          ptr > current_subject)                 /* Matched non-empty string */          (md->moptions & PCRE_PARTIAL_HARD) != 0      /* Hard partial */
2616            ||                                           /* or... */
2617            ((md->moptions & PCRE_PARTIAL_SOFT) != 0 &&  /* Soft partial and */
2618             match_count < 0)                            /* no matches */
2619            ) &&                                         /* And... */
2620            ptr >= end_subject &&                     /* Reached end of subject */
2621            ptr > current_subject)                    /* Matched non-empty string */
2622        {        {
2623        if (offsetcount >= 2)        if (offsetcount >= 2)
2624          {          {
2625          offsets[0] = current_subject - start_subject;          offsets[0] = md->start_used_ptr - start_subject;
2626          offsets[1] = end_subject - start_subject;          offsets[1] = end_subject - start_subject;
2627          }          }
2628        match_count = PCRE_ERROR_PARTIAL;        match_count = PCRE_ERROR_PARTIAL;
# Line 2623  md->start_code = (const uschar *)argumen Line 2764  md->start_code = (const uschar *)argumen
2764      re->name_table_offset + re->name_count * re->name_entry_size;      re->name_table_offset + re->name_count * re->name_entry_size;
2765  md->start_subject = (const unsigned char *)subject;  md->start_subject = (const unsigned char *)subject;
2766  md->end_subject = end_subject;  md->end_subject = end_subject;
2767    md->start_offset = start_offset;
2768  md->moptions = options;  md->moptions = options;
2769  md->poptions = re->options;  md->poptions = re->options;
2770    
# Line 2727  if (!anchored) Line 2869  if (!anchored)
2869      }      }
2870    else    else
2871      {      {
2872      if (startline && study != NULL &&      if (!startline && study != NULL &&
2873           (study->options & PCRE_STUDY_MAPPED) != 0)           (study->flags & PCRE_STUDY_MAPPED) != 0)
2874        start_bits = study->start_bits;        start_bits = study->start_bits;
2875      }      }
2876    }    }
# Line 2779  for (;;) Line 2921  for (;;)
2921        }        }
2922    
2923      /* There are some optimizations that avoid running the match if a known      /* There are some optimizations that avoid running the match if a known
2924      starting point is not found, or if a known later character is not present.      starting point is not found. However, there is an option that disables
2925      However, there is an option that disables these, for testing and for      these, for testing and for ensuring that all callouts do actually occur. */
     ensuring that all callouts do actually occur. */  
2926    
2927      if ((options & PCRE_NO_START_OPTIMIZE) == 0)      if ((options & PCRE_NO_START_OPTIMIZE) == 0)
2928        {        {
   
2929        /* Advance to a known first byte. */        /* Advance to a known first byte. */
2930    
2931        if (first_byte >= 0)        if (first_byte >= 0)
# Line 2851  for (;;) Line 2991  for (;;)
2991      /* Restore fudged end_subject */      /* Restore fudged end_subject */
2992    
2993      end_subject = save_end_subject;      end_subject = save_end_subject;
     }  
2994    
2995    /* If req_byte is set, we know that that character must appear in the subject      /* The following two optimizations are disabled for partial matching or if
2996    for the match to succeed. If the first character is set, req_byte must be      disabling is explicitly requested (and of course, by the test above, this
2997    later in the subject; otherwise the test starts at the match point. This      code is not obeyed when restarting after a partial match). */
   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, and can  
   also be explicitly deactivated. */  
   
   if ((options & PCRE_NO_START_OPTIMIZE) == 0 &&  
       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);  
   
     /* We don't need to repeat the search if we haven't yet reached the  
     place we found it at last time. */  
2998    
2999      if (p > req_byte_ptr)      if ((options & PCRE_NO_START_OPTIMIZE) == 0 &&
3000            (options & (PCRE_PARTIAL_HARD|PCRE_PARTIAL_SOFT)) == 0)
3001        {        {
3002        if (req_byte_caseless)        /* If the pattern was studied, a minimum subject length may be set. This
3003          {        is a lower bound; no actual string of that length may actually match the
3004          while (p < end_subject)        pattern. Although the value is, strictly, in characters, we treat it as
3005            {        bytes to avoid spending too much time in this optimization. */
3006            register int pp = *p++;  
3007            if (pp == req_byte || pp == req_byte2) { p--; break; }        if (study != NULL && (study->flags & PCRE_STUDY_MINLEN) != 0 &&
3008            }            (pcre_uint32)(end_subject - current_subject) < study->minlength)
3009          }          return PCRE_ERROR_NOMATCH;
3010        else  
3011          /* If req_byte is set, we know that that character must appear in the
3012          subject for the match to succeed. If the first character is set, req_byte
3013          must be later in the subject; otherwise the test starts at the match
3014          point. This optimization can save a huge amount of work in patterns with
3015          nested unlimited repeats that aren't going to match. Writing separate
3016          code for cased/caseless versions makes it go faster, as does using an
3017          autoincrement and backing off on a match.
3018    
3019          HOWEVER: when the subject string is very, very long, searching to its end
3020          can take a long time, and give bad performance on quite ordinary
3021          patterns. This showed up when somebody was matching /^C/ on a 32-megabyte
3022          string... so we don't do this when the string is sufficiently long. */
3023    
3024          if (req_byte >= 0 && end_subject - current_subject < REQ_BYTE_MAX)
3025          {          {
3026          while (p < end_subject)          register const uschar *p = current_subject + ((first_byte >= 0)? 1 : 0);
3027    
3028            /* We don't need to repeat the search if we haven't yet reached the
3029            place we found it at last time. */
3030    
3031            if (p > req_byte_ptr)
3032            {            {
3033            if (*p++ == req_byte) { p--; break; }            if (req_byte_caseless)
3034            }              {
3035          }              while (p < end_subject)
3036                  {
3037                  register int pp = *p++;
3038                  if (pp == req_byte || pp == req_byte2) { p--; break; }
3039                  }
3040                }
3041              else
3042                {
3043                while (p < end_subject)
3044                  {
3045                  if (*p++ == req_byte) { p--; break; }
3046                  }
3047                }
3048    
3049        /* If we can't find the required character, break the matching loop,            /* If we can't find the required character, break the matching loop,
3050        which will cause a return or PCRE_ERROR_NOMATCH. */            which will cause a return or PCRE_ERROR_NOMATCH. */
3051    
3052        if (p >= end_subject) break;            if (p >= end_subject) break;
3053    
3054        /* If we have found the required character, save the point where we            /* If we have found the required character, save the point where we
3055        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
3056        the start hasn't passed this character yet. */            the start hasn't passed this character yet. */
3057    
3058        req_byte_ptr = p;            req_byte_ptr = p;
3059              }
3060            }
3061        }        }
3062      }      }   /* End of optimizations that are done when not restarting */
3063    
3064    /* OK, now we can do the business */    /* OK, now we can do the business */
3065    
3066      md->start_used_ptr = current_subject;
3067    
3068    rc = internal_dfa_exec(    rc = internal_dfa_exec(
3069      md,                                /* fixed match data */      md,                                /* fixed match data */
3070      md->start_code,                    /* this subexpression's code */      md->start_code,                    /* this subexpression's code */

Legend:
Removed from v.426  
changed lines
  Added in v.498

  ViewVC Help
Powered by ViewVC 1.1.5