/[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 455 by ph10, Sat Sep 26 19:12:32 2009 UTC revision 473 by ph10, Sat Jan 2 12:40:07 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  /* 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  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  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.  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  The issue is the check for duplicate states, which is done by a simple linear
# Line 68  was the extra time to initialize the ind Line 68  was the extra time to initialize the ind
68  of internal_dfa_exec(). (The supplied patch used a static vector, initialized  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.)  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  Overall, I concluded that the gains in some cases did not outweigh the losses
72  in others, so I abandoned this code. */  in others, so I abandoned this code. */
73    
74    
# Line 109  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                                    */
# Line 160  static const uschar coptable[] = { Line 161  static const uschar coptable[] = {
161    0,                             /* DEF                                    */    0,                             /* DEF                                    */
162    0, 0,                          /* BRAZERO, BRAMINZERO                    */    0, 0,                          /* BRAZERO, BRAMINZERO                    */
163    0, 0, 0, 0,                    /* PRUNE, SKIP, THEN, COMMIT              */    0, 0, 0, 0,                    /* PRUNE, SKIP, THEN, COMMIT              */
164    0, 0, 0                        /* FAIL, ACCEPT, SKIPZERO                 */    0, 0, 0, 0                     /* FAIL, ACCEPT, CLOSE, SKIPZERO          */
165    };
166    
167    /* This table identifies those opcodes that inspect a character. It is used to
168    remember the fact that a character could have been inspected when the end of
169    the subject is reached. ***NOTE*** If the start of this table is modified, the
170    two tables that follow must also be modified. */
171    
172    static const uschar poptable[] = {
173      0,                             /* End                                    */
174      0, 0, 0, 1, 1,                 /* \A, \G, \K, \B, \b                     */
175      1, 1, 1, 1, 1, 1,              /* \D, \d, \S, \s, \W, \w                 */
176      1, 1, 1,                       /* Any, AllAny, Anybyte                   */
177      1, 1, 1,                       /* NOTPROP, PROP, EXTUNI                  */
178      1, 1, 1, 1, 1,                 /* \R, \H, \h, \V, \v                     */
179      0, 0, 0, 0, 0,                 /* \Z, \z, Opt, ^, $                      */
180      1,                             /* Char                                   */
181      1,                             /* Charnc                                 */
182      1,                             /* not                                    */
183      /* Positive single-char repeats                                          */
184      1, 1, 1, 1, 1, 1,              /* *, *?, +, +?, ?, ??                    */
185      1, 1, 1,                       /* upto, minupto, exact                   */
186      1, 1, 1, 1,                    /* *+, ++, ?+, upto+                      */
187      /* Negative single-char repeats - only for chars < 256                   */
188      1, 1, 1, 1, 1, 1,              /* NOT *, *?, +, +?, ?, ??                */
189      1, 1, 1,                       /* NOT upto, minupto, exact               */
190      1, 1, 1, 1,                    /* NOT *+, ++, ?+, upto+                  */
191      /* Positive type repeats                                                 */
192      1, 1, 1, 1, 1, 1,              /* Type *, *?, +, +?, ?, ??               */
193      1, 1, 1,                       /* Type upto, minupto, exact              */
194      1, 1, 1, 1,                    /* Type *+, ++, ?+, upto+                 */
195      /* Character class & ref repeats                                         */
196      1, 1, 1, 1, 1, 1,              /* *, *?, +, +?, ?, ??                    */
197      1, 1,                          /* CRRANGE, CRMINRANGE                    */
198      1,                             /* CLASS                                  */
199      1,                             /* NCLASS                                 */
200      1,                             /* XCLASS - variable length               */
201      0,                             /* REF                                    */
202      0,                             /* RECURSE                                */
203      0,                             /* CALLOUT                                */
204      0,                             /* Alt                                    */
205      0,                             /* Ket                                    */
206      0,                             /* KetRmax                                */
207      0,                             /* KetRmin                                */
208      0,                             /* Assert                                 */
209      0,                             /* Assert not                             */
210      0,                             /* Assert behind                          */
211      0,                             /* Assert behind not                      */
212      0,                             /* Reverse                                */
213      0, 0, 0, 0,                    /* ONCE, BRA, CBRA, COND                  */
214      0, 0, 0,                       /* SBRA, SCBRA, SCOND                     */
215      0,                             /* CREF                                   */
216      0,                             /* RREF                                   */
217      0,                             /* DEF                                    */
218      0, 0,                          /* BRAZERO, BRAMINZERO                    */
219      0, 0, 0, 0,                    /* PRUNE, SKIP, THEN, COMMIT              */
220      0, 0, 0, 0                     /* FAIL, ACCEPT, CLOSE, SKIPZERO          */
221  };  };
222    
223  /* 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 417  if (*first_op == OP_REVERSE) Line 474  if (*first_op == OP_REVERSE)
474        current_subject - start_subject : max_back;        current_subject - start_subject : max_back;
475      current_subject -= gone_back;      current_subject -= gone_back;
476      }      }
477    
478    /* Save the earliest consulted character */    /* Save the earliest consulted character */
479    
480    if (current_subject < md->start_used_ptr)    if (current_subject < md->start_used_ptr)
481      md->start_used_ptr = current_subject;      md->start_used_ptr = current_subject;
482    
483    /* Now we can process the individual branches. */    /* Now we can process the individual branches. */
484    
# Line 488  for (;;) Line 545  for (;;)
545    int clen, dlen;    int clen, dlen;
546    unsigned int c, d;    unsigned int c, d;
547    int forced_fail = 0;    int forced_fail = 0;
548    int reached_end = 0;    BOOL could_continue = FALSE;
549    
550    /* 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
551    new state list. */    new state list. */
# Line 578  for (;;) Line 635  for (;;)
635          }          }
636        }        }
637    
638      /* 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.
639      See the note at the head of this module about the possibility of improving      See the note at the head of this module about the possibility of improving
640      performance here. */      performance here. */
641    
# Line 597  for (;;) Line 654  for (;;)
654      code = start_code + state_offset;      code = start_code + state_offset;
655      codevalue = *code;      codevalue = *code;
656    
657        /* If this opcode inspects a character, but we are at the end of the
658        subject, remember the fact for use when testing for a partial match. */
659    
660        if (clen == 0 && poptable[codevalue] != 0)
661          could_continue = TRUE;
662    
663      /* 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
664      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
665      is wrong, because sometimes zero repetitions of the subject are      is wrong, because sometimes zero repetitions of the subject are
# Line 647  for (;;) Line 710  for (;;)
710  /* ========================================================================== */  /* ========================================================================== */
711        /* 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
712        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
713        PCRE_NOTEMPTY is set, or PCRE_NOTEMPTY_ATSTART is set and we are at the        PCRE_NOTEMPTY is set, or PCRE_NOTEMPTY_ATSTART is set and we are at the
714        start of the subject, save the match data, shifting up all previous        start of the subject, save the match data, shifting up all previous
715        matches so we always have the longest first. */        matches so we always have the longest first. */
716    
# Line 662  for (;;) Line 725  for (;;)
725            ADD_ACTIVE(state_offset - GET(code, 1), 0);            ADD_ACTIVE(state_offset - GET(code, 1), 0);
726            }            }
727          }          }
728        else        else
729          {          {
730          reached_end++;    /* Count branches that reach the end */          if (ptr > current_subject ||
         if (ptr > current_subject ||  
731              ((md->moptions & PCRE_NOTEMPTY) == 0 &&              ((md->moptions & PCRE_NOTEMPTY) == 0 &&
732                ((md->moptions & PCRE_NOTEMPTY_ATSTART) == 0 ||                ((md->moptions & PCRE_NOTEMPTY_ATSTART) == 0 ||
733                  current_subject > start_subject + md->start_offset)))                  current_subject > start_subject + md->start_offset)))
# Line 689  for (;;) Line 751  for (;;)
751                match_count, rlevel*2-2, SP));                match_count, rlevel*2-2, SP));
752              return match_count;              return match_count;
753              }              }
754            }            }
755          }          }
756        break;        break;
757    
# Line 839  for (;;) Line 901  for (;;)
901          if (ptr > start_subject)          if (ptr > start_subject)
902            {            {
903            const uschar *temp = ptr - 1;            const uschar *temp = ptr - 1;
904            if (temp < md->start_used_ptr) md->start_used_ptr = temp;            if (temp < md->start_used_ptr) md->start_used_ptr = temp;
905  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
906            if (utf8) BACKCHAR(temp);            if (utf8) BACKCHAR(temp);
907  #endif  #endif
# Line 848  for (;;) Line 910  for (;;)
910            }            }
911          else left_word = 0;          else left_word = 0;
912    
913          if (clen > 0)          if (clen > 0)
914            right_word = c < 256 && (ctypes[c] & ctype_word) != 0;            right_word = c < 256 && (ctypes[c] & ctype_word) != 0;
915          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;  
           }  
916    
917          if ((left_word == right_word) == (codevalue == OP_NOT_WORD_BOUNDARY))          if ((left_word == right_word) == (codevalue == OP_NOT_WORD_BOUNDARY))
918            { ADD_ACTIVE(state_offset + 1, 0); }            { ADD_ACTIVE(state_offset + 1, 0); }
# Line 2240  for (;;) Line 2298  for (;;)
2298            ims,                                  /* the current ims flags */            ims,                                  /* the current ims flags */
2299            rlevel,                               /* function recursion level */            rlevel,                               /* function recursion level */
2300            recursing);                           /* pass on regex recursion */            recursing);                           /* pass on regex recursion */
2301    
2302            if (rc == PCRE_ERROR_DFA_UITEM) return rc;
2303          if ((rc >= 0) == (codevalue == OP_ASSERT || codevalue == OP_ASSERTBACK))          if ((rc >= 0) == (codevalue == OP_ASSERT || codevalue == OP_ASSERTBACK))
2304              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }
2305          }          }
# Line 2287  for (;;) Line 2346  for (;;)
2346    
2347          /* Back reference conditions are not supported */          /* Back reference conditions are not supported */
2348    
2349          if (condcode == OP_CREF) return PCRE_ERROR_DFA_UCOND;          if (condcode == OP_CREF || condcode == OP_NCREF)
2350              return PCRE_ERROR_DFA_UCOND;
2351    
2352          /* The DEFINE condition is always false */          /* The DEFINE condition is always false */
2353    
# Line 2298  for (;;) Line 2358  for (;;)
2358          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
2359          recursed groups. */          recursed groups. */
2360    
2361          else if (condcode == OP_RREF)          else if (condcode == OP_RREF || condcode == OP_NRREF)
2362            {            {
2363            int value = GET2(code, LINK_SIZE+2);            int value = GET2(code, LINK_SIZE+2);
2364            if (value != RREF_ANY) return PCRE_ERROR_DFA_UCOND;            if (value != RREF_ANY) return PCRE_ERROR_DFA_UCOND;
# Line 2330  for (;;) Line 2390  for (;;)
2390              rlevel,                               /* function recursion level */              rlevel,                               /* function recursion level */
2391              recursing);                           /* pass on regex recursion */              recursing);                           /* pass on regex recursion */
2392    
2393              if (rc == PCRE_ERROR_DFA_UITEM) return rc;
2394            if ((rc >= 0) ==            if ((rc >= 0) ==
2395                  (condcode == OP_ASSERT || condcode == OP_ASSERTBACK))                  (condcode == OP_ASSERT || condcode == OP_ASSERTBACK))
2396              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }
# Line 2521  for (;;) Line 2582  for (;;)
2582    /* We have finished the processing at the current subject character. If no    /* We have finished the processing at the current subject character. If no
2583    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
2584    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
2585    matching has been requested, check for appropriate conditions. The "forced_    matching has been requested, check for appropriate conditions.
2586    fail" variable counts the number of (*F) encountered for the character. If it  
2587    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
2588    (*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
2589    partial match. */    workspace[1]) it means that (*F) was found on every active state. In this
2590      case we don't want to give a partial match.
2591    
2592      The "could_continue" variable is true if a state could have continued but
2593      for the fact that the end of the subject was reached. */
2594    
2595    if (new_count <= 0)    if (new_count <= 0)
2596      {      {
2597      if (rlevel == 1 &&                               /* Top level, and */      if (rlevel == 1 &&                               /* Top level, and */
2598          reached_end != workspace[1] &&               /* Not all reached end */          could_continue &&                            /* Some could go on */
2599          forced_fail != workspace[1] &&               /* Not all forced fail & */          forced_fail != workspace[1] &&               /* Not all forced fail & */
2600          (                                            /* either... */          (                                            /* either... */
2601          (md->moptions & PCRE_PARTIAL_HARD) != 0      /* Hard partial */          (md->moptions & PCRE_PARTIAL_HARD) != 0      /* Hard partial */
# Line 2651  if (extra_data != NULL) Line 2716  if (extra_data != NULL)
2716    if ((flags & PCRE_EXTRA_TABLES) != 0)    if ((flags & PCRE_EXTRA_TABLES) != 0)
2717      md->tables = extra_data->tables;      md->tables = extra_data->tables;
2718    }    }
2719    
2720  /* Check that the first field in the block is the magic number. If it is not,  /* Check that the first field in the block is the magic number. If it is not,
2721  test for a regex that was compiled on a host of opposite endianness. If this is  test for a regex that was compiled on a host of opposite endianness. If this is
2722  the case, flipped values are put in internal_re and internal_study if there was  the case, flipped values are put in internal_re and internal_study if there was
# Line 2913  for (;;) Line 2978  for (;;)
2978    
2979      end_subject = save_end_subject;      end_subject = save_end_subject;
2980    
2981      /* The following two optimizations are disabled for partial matching or if      /* The following two optimizations are disabled for partial matching or if
2982      disabling is explicitly requested (and of course, by the test above, this      disabling is explicitly requested (and of course, by the test above, this
2983      code is not obeyed when restarting after a partial match). */      code is not obeyed when restarting after a partial match). */
2984    
2985      if ((options & PCRE_NO_START_OPTIMIZE) == 0 &&      if ((options & PCRE_NO_START_OPTIMIZE) == 0 &&
2986          (options & (PCRE_PARTIAL_HARD|PCRE_PARTIAL_SOFT)) == 0)          (options & (PCRE_PARTIAL_HARD|PCRE_PARTIAL_SOFT)) == 0)
2987        {        {
2988        /* If the pattern was studied, a minimum subject length may be set. This        /* If the pattern was studied, a minimum subject length may be set. This
2989        is a lower bound; no actual string of that length may actually match the        is a lower bound; no actual string of that length may actually match the
2990        pattern. Although the value is, strictly, in characters, we treat it as        pattern. Although the value is, strictly, in characters, we treat it as
# Line 2928  for (;;) Line 2993  for (;;)
2993        if (study != NULL && (study->flags & PCRE_STUDY_MINLEN) != 0 &&        if (study != NULL && (study->flags & PCRE_STUDY_MINLEN) != 0 &&
2994            end_subject - current_subject < study->minlength)            end_subject - current_subject < study->minlength)
2995          return PCRE_ERROR_NOMATCH;          return PCRE_ERROR_NOMATCH;
2996    
2997        /* If req_byte is set, we know that that character must appear in the        /* If req_byte is set, we know that that character must appear in the
2998        subject for the match to succeed. If the first character is set, req_byte        subject for the match to succeed. If the first character is set, req_byte
2999        must be later in the subject; otherwise the test starts at the match        must be later in the subject; otherwise the test starts at the match
# Line 2936  for (;;) Line 3001  for (;;)
3001        nested unlimited repeats that aren't going to match. Writing separate        nested unlimited repeats that aren't going to match. Writing separate
3002        code for cased/caseless versions makes it go faster, as does using an        code for cased/caseless versions makes it go faster, as does using an
3003        autoincrement and backing off on a match.        autoincrement and backing off on a match.
3004    
3005        HOWEVER: when the subject string is very, very long, searching to its end        HOWEVER: when the subject string is very, very long, searching to its end
3006        can take a long time, and give bad performance on quite ordinary        can take a long time, and give bad performance on quite ordinary
3007        patterns. This showed up when somebody was matching /^C/ on a 32-megabyte        patterns. This showed up when somebody was matching /^C/ on a 32-megabyte
3008        string... so we don't do this when the string is sufficiently long. */        string... so we don't do this when the string is sufficiently long. */
3009    
3010        if (req_byte >= 0 && end_subject - current_subject < REQ_BYTE_MAX)        if (req_byte >= 0 && end_subject - current_subject < REQ_BYTE_MAX)
3011          {          {
3012          register const uschar *p = current_subject + ((first_byte >= 0)? 1 : 0);          register const uschar *p = current_subject + ((first_byte >= 0)? 1 : 0);
3013    
3014          /* We don't need to repeat the search if we haven't yet reached the          /* We don't need to repeat the search if we haven't yet reached the
3015          place we found it at last time. */          place we found it at last time. */
3016    
3017          if (p > req_byte_ptr)          if (p > req_byte_ptr)
3018            {            {
3019            if (req_byte_caseless)            if (req_byte_caseless)
# Line 2966  for (;;) Line 3031  for (;;)
3031                if (*p++ == req_byte) { p--; break; }                if (*p++ == req_byte) { p--; break; }
3032                }                }
3033              }              }
3034    
3035            /* If we can't find the required character, break the matching loop,            /* If we can't find the required character, break the matching loop,
3036            which will cause a return or PCRE_ERROR_NOMATCH. */            which will cause a return or PCRE_ERROR_NOMATCH. */
3037    
3038            if (p >= end_subject) break;            if (p >= end_subject) break;
3039    
3040            /* If we have found the required character, save the point where we            /* If we have found the required character, save the point where we
3041            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
3042            the start hasn't passed this character yet. */            the start hasn't passed this character yet. */
3043    
3044            req_byte_ptr = p;            req_byte_ptr = p;
3045            }            }
3046          }          }
3047        }        }
3048      }   /* End of optimizations that are done when not restarting */      }   /* End of optimizations that are done when not restarting */
3049    
3050    /* OK, now we can do the business */    /* OK, now we can do the business */
3051    
3052    md->start_used_ptr = current_subject;    md->start_used_ptr = current_subject;
3053    
3054    rc = internal_dfa_exec(    rc = internal_dfa_exec(
3055      md,                                /* fixed match data */      md,                                /* fixed match data */
3056      md->start_code,                    /* this subexpression's code */      md->start_code,                    /* this subexpression's code */

Legend:
Removed from v.455  
changed lines
  Added in v.473

  ViewVC Help
Powered by ViewVC 1.1.5