/[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 428 by ph10, Mon Aug 31 17:10:26 2009 UTC revision 517 by ph10, Wed May 5 10:44:20 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 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 455  for (;;) Line 549  for (;;)
549    int clen, dlen;    int clen, dlen;
550    unsigned int c, d;    unsigned int c, d;
551    int forced_fail = 0;    int forced_fail = 0;
552    int reached_end = 0;    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 469  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 515  for (;;) Line 609  for (;;)
609      int state_offset = current_state->offset;      int state_offset = current_state->offset;
610      int count, codevalue, rrc;      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 545  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 562  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 608  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 626  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        else
745          {          {
746          reached_end++;    /* Count branches that reach the end */          if (ptr > current_subject ||
747          if (ptr > current_subject || (md->moptions & PCRE_NOTEMPTY) == 0)              ((md->moptions & PCRE_NOTEMPTY) == 0 &&
748                  ((md->moptions & PCRE_NOTEMPTY_ATSTART) == 0 ||
749                    current_subject > start_subject + md->start_offset)))
750            {            {
751            if (match_count < 0) match_count = (offsetcount >= 2)? 1 : 0;            if (match_count < 0) match_count = (offsetcount >= 2)? 1 : 0;
752              else if (match_count > 0 && ++match_count * 2 >= offsetcount)              else if (match_count > 0 && ++match_count * 2 >= offsetcount)
# Line 650  for (;;) Line 767  for (;;)
767                match_count, rlevel*2-2, SP));                match_count, rlevel*2-2, SP));
768              return match_count;              return match_count;
769              }              }
770            }            }
771          }          }
772        break;        break;
773    
# Line 800  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 808  for (;;) Line 926  for (;;)
926            }            }
927          else left_word = 0;          else left_word = 0;
928    
929          if (clen > 0)          if (clen > 0)
930            right_word = c < 256 && (ctypes[c] & ctype_word) != 0;            right_word = c < 256 && (ctypes[c] & ctype_word) != 0;
931          else              /* This is a fudge to ensure that if this is the */          else right_word = 0;
           {               /* last item in the pattern, we don't count it as */  
           reached_end--;  /* reached, thus disabling a partial match. */  
           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 841  for (;;) Line 955  for (;;)
955            break;            break;
956    
957            case PT_LAMP:            case PT_LAMP:
958            OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll || prop->chartype == ucp_Lt;            OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll ||
959                   prop->chartype == ucp_Lt;
960            break;            break;
961    
962            case PT_GC:            case PT_GC:
# Line 855  for (;;) Line 970  for (;;)
970            case PT_SC:            case PT_SC:
971            OK = prop->script == code[2];            OK = prop->script == code[2];
972            break;            break;
973    
974              /* These are specials for combination cases. */
975    
976              case PT_ALNUM:
977              OK = _pcre_ucp_gentype[prop->chartype] == ucp_L ||
978                   _pcre_ucp_gentype[prop->chartype] == ucp_N;
979              break;
980    
981              case PT_SPACE:    /* Perl space */
982              OK = _pcre_ucp_gentype[prop->chartype] == ucp_Z ||
983                   c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR;
984              break;
985    
986              case PT_PXSPACE:  /* POSIX space */
987              OK = _pcre_ucp_gentype[prop->chartype] == ucp_Z ||
988                   c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
989                   c == CHAR_FF || c == CHAR_CR;
990              break;
991    
992              case PT_WORD:
993              OK = _pcre_ucp_gentype[prop->chartype] == ucp_L ||
994                   _pcre_ucp_gentype[prop->chartype] == ucp_N ||
995                   c == CHAR_UNDERSCORE;
996              break;
997    
998            /* Should never occur, but keep compilers from grumbling. */            /* Should never occur, but keep compilers from grumbling. */
999    
# Line 1010  for (;;) Line 1149  for (;;)
1149            break;            break;
1150    
1151            case PT_LAMP:            case PT_LAMP:
1152            OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll || prop->chartype == ucp_Lt;            OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll ||
1153                prop->chartype == ucp_Lt;
1154            break;            break;
1155    
1156            case PT_GC:            case PT_GC:
# Line 1025  for (;;) Line 1165  for (;;)
1165            OK = prop->script == code[3];            OK = prop->script == code[3];
1166            break;            break;
1167    
1168              /* These are specials for combination cases. */
1169    
1170              case PT_ALNUM:
1171              OK = _pcre_ucp_gentype[prop->chartype] == ucp_L ||
1172                   _pcre_ucp_gentype[prop->chartype] == ucp_N;
1173              break;
1174    
1175              case PT_SPACE:    /* Perl space */
1176              OK = _pcre_ucp_gentype[prop->chartype] == ucp_Z ||
1177                   c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR;
1178              break;
1179    
1180              case PT_PXSPACE:  /* POSIX space */
1181              OK = _pcre_ucp_gentype[prop->chartype] == ucp_Z ||
1182                   c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
1183                   c == CHAR_FF || c == CHAR_CR;
1184              break;
1185    
1186              case PT_WORD:
1187              OK = _pcre_ucp_gentype[prop->chartype] == ucp_L ||
1188                   _pcre_ucp_gentype[prop->chartype] == ucp_N ||
1189                   c == CHAR_UNDERSCORE;
1190              break;
1191    
1192            /* Should never occur, but keep compilers from grumbling. */            /* Should never occur, but keep compilers from grumbling. */
1193    
1194            default:            default:
# Line 1232  for (;;) Line 1396  for (;;)
1396            break;            break;
1397    
1398            case PT_LAMP:            case PT_LAMP:
1399            OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll || prop->chartype == ucp_Lt;            OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll ||
1400                prop->chartype == ucp_Lt;
1401            break;            break;
1402    
1403            case PT_GC:            case PT_GC:
# Line 1246  for (;;) Line 1411  for (;;)
1411            case PT_SC:            case PT_SC:
1412            OK = prop->script == code[3];            OK = prop->script == code[3];
1413            break;            break;
1414    
1415              /* These are specials for combination cases. */
1416    
1417              case PT_ALNUM:
1418              OK = _pcre_ucp_gentype[prop->chartype] == ucp_L ||
1419                   _pcre_ucp_gentype[prop->chartype] == ucp_N;
1420              break;
1421    
1422              case PT_SPACE:    /* Perl space */
1423              OK = _pcre_ucp_gentype[prop->chartype] == ucp_Z ||
1424                   c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR;
1425              break;
1426    
1427              case PT_PXSPACE:  /* POSIX space */
1428              OK = _pcre_ucp_gentype[prop->chartype] == ucp_Z ||
1429                   c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
1430                   c == CHAR_FF || c == CHAR_CR;
1431              break;
1432    
1433              case PT_WORD:
1434              OK = _pcre_ucp_gentype[prop->chartype] == ucp_L ||
1435                   _pcre_ucp_gentype[prop->chartype] == ucp_N ||
1436                   c == CHAR_UNDERSCORE;
1437              break;
1438    
1439            /* Should never occur, but keep compilers from grumbling. */            /* Should never occur, but keep compilers from grumbling. */
1440    
# Line 1479  for (;;) Line 1668  for (;;)
1668            break;            break;
1669    
1670            case PT_LAMP:            case PT_LAMP:
1671            OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll || prop->chartype == ucp_Lt;            OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll ||
1672                prop->chartype == ucp_Lt;
1673            break;            break;
1674    
1675            case PT_GC:            case PT_GC:
# Line 1493  for (;;) Line 1683  for (;;)
1683            case PT_SC:            case PT_SC:
1684            OK = prop->script == code[5];            OK = prop->script == code[5];
1685            break;            break;
1686    
1687              /* These are specials for combination cases. */
1688    
1689              case PT_ALNUM:
1690              OK = _pcre_ucp_gentype[prop->chartype] == ucp_L ||
1691                   _pcre_ucp_gentype[prop->chartype] == ucp_N;
1692              break;
1693    
1694              case PT_SPACE:    /* Perl space */
1695              OK = _pcre_ucp_gentype[prop->chartype] == ucp_Z ||
1696                   c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR;
1697              break;
1698    
1699              case PT_PXSPACE:  /* POSIX space */
1700              OK = _pcre_ucp_gentype[prop->chartype] == ucp_Z ||
1701                   c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
1702                   c == CHAR_FF || c == CHAR_CR;
1703              break;
1704    
1705              case PT_WORD:
1706              OK = _pcre_ucp_gentype[prop->chartype] == ucp_L ||
1707                   _pcre_ucp_gentype[prop->chartype] == ucp_N ||
1708                   c == CHAR_UNDERSCORE;
1709              break;
1710    
1711            /* Should never occur, but keep compilers from grumbling. */            /* Should never occur, but keep compilers from grumbling. */
1712    
# Line 2201  for (;;) Line 2415  for (;;)
2415            rlevel,                               /* function recursion level */            rlevel,                               /* function recursion level */
2416            recursing);                           /* pass on regex recursion */            recursing);                           /* pass on regex recursion */
2417    
2418            if (rc == PCRE_ERROR_DFA_UITEM) return rc;
2419          if ((rc >= 0) == (codevalue == OP_ASSERT || codevalue == OP_ASSERTBACK))          if ((rc >= 0) == (codevalue == OP_ASSERT || codevalue == OP_ASSERTBACK))
2420              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }
2421          }          }
# Line 2247  for (;;) Line 2462  for (;;)
2462    
2463          /* Back reference conditions are not supported */          /* Back reference conditions are not supported */
2464    
2465          if (condcode == OP_CREF) return PCRE_ERROR_DFA_UCOND;          if (condcode == OP_CREF || condcode == OP_NCREF)
2466              return PCRE_ERROR_DFA_UCOND;
2467    
2468          /* The DEFINE condition is always false */          /* The DEFINE condition is always false */
2469    
# Line 2258  for (;;) Line 2474  for (;;)
2474          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
2475          recursed groups. */          recursed groups. */
2476    
2477          else if (condcode == OP_RREF)          else if (condcode == OP_RREF || condcode == OP_NRREF)
2478            {            {
2479            int value = GET2(code, LINK_SIZE+2);            int value = GET2(code, LINK_SIZE+2);
2480            if (value != RREF_ANY) return PCRE_ERROR_DFA_UCOND;            if (value != RREF_ANY) return PCRE_ERROR_DFA_UCOND;
# Line 2290  for (;;) Line 2506  for (;;)
2506              rlevel,                               /* function recursion level */              rlevel,                               /* function recursion level */
2507              recursing);                           /* pass on regex recursion */              recursing);                           /* pass on regex recursion */
2508    
2509              if (rc == PCRE_ERROR_DFA_UITEM) return rc;
2510            if ((rc >= 0) ==            if ((rc >= 0) ==
2511                  (condcode == OP_ASSERT || condcode == OP_ASSERTBACK))                  (condcode == OP_ASSERT || condcode == OP_ASSERTBACK))
2512              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }
# Line 2481  for (;;) Line 2698  for (;;)
2698    /* We have finished the processing at the current subject character. If no    /* We have finished the processing at the current subject character. If no
2699    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
2700    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
2701    matching has been requested, check for appropriate conditions. The "forced_    matching has been requested, check for appropriate conditions.
2702    fail" variable counts the number of (*F) encountered for the character. If it  
2703    is equal to the original active_count (saved in workspace[1]) it means that    The "forced_ fail" variable counts the number of (*F) encountered for the
2704    (*F) was found on every active state. In this case we don't want to give a    character. If it is equal to the original active_count (saved in
2705    partial match. */    workspace[1]) it means that (*F) was found on every active state. In this
2706      case we don't want to give a partial match.
2707    
2708      The "could_continue" variable is true if a state could have continued but
2709      for the fact that the end of the subject was reached. */
2710    
2711    if (new_count <= 0)    if (new_count <= 0)
2712      {      {
2713      if (rlevel == 1 &&                               /* Top level, and */      if (rlevel == 1 &&                               /* Top level, and */
2714          reached_end != workspace[1] &&               /* Not all reached end */          could_continue &&                            /* Some could go on */
2715          forced_fail != workspace[1] &&               /* Not all forced fail & */          forced_fail != workspace[1] &&               /* Not all forced fail & */
2716          (                                            /* either... */          (                                            /* either... */
2717          (md->moptions & PCRE_PARTIAL_HARD) != 0      /* Hard partial */          (md->moptions & PCRE_PARTIAL_HARD) != 0      /* Hard partial */
# Line 2503  for (;;) Line 2724  for (;;)
2724        {        {
2725        if (offsetcount >= 2)        if (offsetcount >= 2)
2726          {          {
2727          offsets[0] = current_subject - start_subject;          offsets[0] = md->start_used_ptr - start_subject;
2728          offsets[1] = end_subject - start_subject;          offsets[1] = end_subject - start_subject;
2729          }          }
2730        match_count = PCRE_ERROR_PARTIAL;        match_count = PCRE_ERROR_PARTIAL;
# Line 2645  md->start_code = (const uschar *)argumen Line 2866  md->start_code = (const uschar *)argumen
2866      re->name_table_offset + re->name_count * re->name_entry_size;      re->name_table_offset + re->name_count * re->name_entry_size;
2867  md->start_subject = (const unsigned char *)subject;  md->start_subject = (const unsigned char *)subject;
2868  md->end_subject = end_subject;  md->end_subject = end_subject;
2869    md->start_offset = start_offset;
2870  md->moptions = options;  md->moptions = options;
2871  md->poptions = re->options;  md->poptions = re->options;
2872    
# Line 2749  if (!anchored) Line 2971  if (!anchored)
2971      }      }
2972    else    else
2973      {      {
2974      if (startline && study != NULL &&      if (!startline && study != NULL &&
2975           (study->options & PCRE_STUDY_MAPPED) != 0)           (study->flags & PCRE_STUDY_MAPPED) != 0)
2976        start_bits = study->start_bits;        start_bits = study->start_bits;
2977      }      }
2978    }    }
# Line 2801  for (;;) Line 3023  for (;;)
3023        }        }
3024    
3025      /* There are some optimizations that avoid running the match if a known      /* There are some optimizations that avoid running the match if a known
3026      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
3027      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. */  
3028    
3029      if ((options & PCRE_NO_START_OPTIMIZE) == 0)      if ((options & PCRE_NO_START_OPTIMIZE) == 0)
3030        {        {
   
3031        /* Advance to a known first byte. */        /* Advance to a known first byte. */
3032    
3033        if (first_byte >= 0)        if (first_byte >= 0)
# Line 2873  for (;;) Line 3093  for (;;)
3093      /* Restore fudged end_subject */      /* Restore fudged end_subject */
3094    
3095      end_subject = save_end_subject;      end_subject = save_end_subject;
     }  
3096    
3097    /* 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
3098    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
3099    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. Furthermore, we have to disable when  
   restarting after a partial match, because the required character may have  
   already been matched. */  
   
   if ((options & PCRE_NO_START_OPTIMIZE) == 0 &&  
       req_byte >= 0 &&  
       end_subject - current_subject < REQ_BYTE_MAX &&  
       (options & (PCRE_PARTIAL_HARD|PCRE_PARTIAL_SOFT|PCRE_DFA_RESTART)) == 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. */  
3100    
3101      if (p > req_byte_ptr)      if ((options & PCRE_NO_START_OPTIMIZE) == 0 &&
3102            (options & (PCRE_PARTIAL_HARD|PCRE_PARTIAL_SOFT)) == 0)
3103        {        {
3104        if (req_byte_caseless)        /* If the pattern was studied, a minimum subject length may be set. This
3105          {        is a lower bound; no actual string of that length may actually match the
3106          while (p < end_subject)        pattern. Although the value is, strictly, in characters, we treat it as
3107            {        bytes to avoid spending too much time in this optimization. */
3108            register int pp = *p++;  
3109            if (pp == req_byte || pp == req_byte2) { p--; break; }        if (study != NULL && (study->flags & PCRE_STUDY_MINLEN) != 0 &&
3110            }            (pcre_uint32)(end_subject - current_subject) < study->minlength)
3111          }          return PCRE_ERROR_NOMATCH;
3112        else  
3113          /* If req_byte is set, we know that that character must appear in the
3114          subject for the match to succeed. If the first character is set, req_byte
3115          must be later in the subject; otherwise the test starts at the match
3116          point. This optimization can save a huge amount of work in patterns with
3117          nested unlimited repeats that aren't going to match. Writing separate
3118          code for cased/caseless versions makes it go faster, as does using an
3119          autoincrement and backing off on a match.
3120    
3121          HOWEVER: when the subject string is very, very long, searching to its end
3122          can take a long time, and give bad performance on quite ordinary
3123          patterns. This showed up when somebody was matching /^C/ on a 32-megabyte
3124          string... so we don't do this when the string is sufficiently long. */
3125    
3126          if (req_byte >= 0 && end_subject - current_subject < REQ_BYTE_MAX)
3127          {          {
3128          while (p < end_subject)          register const uschar *p = current_subject + ((first_byte >= 0)? 1 : 0);
3129    
3130            /* We don't need to repeat the search if we haven't yet reached the
3131            place we found it at last time. */
3132    
3133            if (p > req_byte_ptr)
3134            {            {
3135            if (*p++ == req_byte) { p--; break; }            if (req_byte_caseless)
3136            }              {
3137          }              while (p < end_subject)
3138                  {
3139                  register int pp = *p++;
3140                  if (pp == req_byte || pp == req_byte2) { p--; break; }
3141                  }
3142                }
3143              else
3144                {
3145                while (p < end_subject)
3146                  {
3147                  if (*p++ == req_byte) { p--; break; }
3148                  }
3149                }
3150    
3151        /* If we can't find the required character, break the matching loop,            /* If we can't find the required character, break the matching loop,
3152        which will cause a return or PCRE_ERROR_NOMATCH. */            which will cause a return or PCRE_ERROR_NOMATCH. */
3153    
3154        if (p >= end_subject) break;            if (p >= end_subject) break;
3155    
3156        /* If we have found the required character, save the point where we            /* If we have found the required character, save the point where we
3157        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
3158        the start hasn't passed this character yet. */            the start hasn't passed this character yet. */
3159    
3160        req_byte_ptr = p;            req_byte_ptr = p;
3161              }
3162            }
3163        }        }
3164      }      }   /* End of optimizations that are done when not restarting */
3165    
3166    /* OK, now we can do the business */    /* OK, now we can do the business */
3167    
3168      md->start_used_ptr = current_subject;
3169    
3170    rc = internal_dfa_exec(    rc = internal_dfa_exec(
3171      md,                                /* fixed match data */      md,                                /* fixed match data */
3172      md->start_code,                    /* this subexpression's code */      md->start_code,                    /* this subexpression's code */

Legend:
Removed from v.428  
changed lines
  Added in v.517

  ViewVC Help
Powered by ViewVC 1.1.5