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

Diff of /code/trunk/pcre_study.c

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

revision 605 by ph10, Fri Jun 3 18:18:30 2011 UTC revision 754 by ph10, Sat Nov 19 18:32:18 2011 UTC
# Line 66  string of that length that matches. In U Line 66  string of that length that matches. In U
66  rather than bytes.  rather than bytes.
67    
68  Arguments:  Arguments:
69    code       pointer to start of group (the bracket)    code            pointer to start of group (the bracket)
70    startcode  pointer to start of the whole pattern    startcode       pointer to start of the whole pattern
71    options    the compiling options    options         the compiling options
72      int             RECURSE depth
73    
74  Returns:   the minimum length  Returns:   the minimum length
75             -1 if \C was encountered             -1 if \C in UTF-8 mode or (*ACCEPT) was encountered
76             -2 internal error (missing capturing bracket)             -2 internal error (missing capturing bracket)
77             -3 internal error (opcode not listed)             -3 internal error (opcode not listed)
78  */  */
79    
80  static int  static int
81  find_minlength(const uschar *code, const uschar *startcode, int options)  find_minlength(const uschar *code, const uschar *startcode, int options,
82      int recurse_depth)
83  {  {
84  int length = -1;  int length = -1;
85  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 125  for (;;) Line 127  for (;;)
127      case OP_BRAPOS:      case OP_BRAPOS:
128      case OP_SBRAPOS:      case OP_SBRAPOS:
129      case OP_ONCE:      case OP_ONCE:
130      d = find_minlength(cc, startcode, options);      case OP_ONCE_NC:
131        d = find_minlength(cc, startcode, options, recurse_depth);
132      if (d < 0) return d;      if (d < 0) return d;
133      branchlength += d;      branchlength += d;
134      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
135      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
136      break;      break;
137    
138        /* ACCEPT makes things far too complicated; we have to give up. */
139    
140        case OP_ACCEPT:
141        case OP_ASSERT_ACCEPT:
142        return -1;
143    
144      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested
145      call. If it's ALT it is an alternation in a nested call. If it is      call. If it's ALT it is an alternation in a nested call. If it is END it's
146      END it's the end of the outer call. All can be handled by the same code. */      the end of the outer call. All can be handled by the same code. If an
147        ACCEPT was previously encountered, use the length that was in force at that
148        time, and pass back the shortest ACCEPT length. */
149    
150      case OP_ALT:      case OP_ALT:
151      case OP_KET:      case OP_KET:
# Line 144  for (;;) Line 155  for (;;)
155      case OP_END:      case OP_END:
156      if (length < 0 || (!had_recurse && branchlength < length))      if (length < 0 || (!had_recurse && branchlength < length))
157        length = branchlength;        length = branchlength;
158      if (*cc != OP_ALT) return length;      if (op != OP_ALT) return length;
159      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
160      branchlength = 0;      branchlength = 0;
161      had_recurse = FALSE;      had_recurse = FALSE;
# Line 267  for (;;) Line 278  for (;;)
278      cc++;      cc++;
279      break;      break;
280    
281      /* "Any newline" might match two characters */      /* "Any newline" might match two characters, but it also might match just
282        one. */
283    
284      case OP_ANYNL:      case OP_ANYNL:
285      branchlength += 2;      branchlength += 1;
286      cc++;      cc++;
287      break;      break;
288    
289      /* The single-byte matcher means we can't proceed in UTF-8 mode */      /* The single-byte matcher means we can't proceed in UTF-8 mode. (In
290        non-UTF-8 mode \C will actually be turned into OP_ALLANY, so won't ever
291        appear, but leave the code, just in case.) */
292    
293      case OP_ANYBYTE:      case OP_ANYBYTE:
294  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 366  for (;;) Line 380  for (;;)
380          d = 0;          d = 0;
381          had_recurse = TRUE;          had_recurse = TRUE;
382          }          }
383        else d = find_minlength(cs, startcode, options);        else
384            {
385            d = find_minlength(cs, startcode, options, recurse_depth);
386            }
387        }        }
388      else d = 0;      else d = 0;
389      cc += 3;      cc += 3;
# Line 403  for (;;) Line 420  for (;;)
420      branchlength += min * d;      branchlength += min * d;
421      break;      break;
422    
423        /* We can easily detect direct recursion, but not mutual recursion. This is
424        caught by a recursion depth count. */
425    
426      case OP_RECURSE:      case OP_RECURSE:
427      cs = ce = (uschar *)startcode + GET(cc, 1);      cs = ce = (uschar *)startcode + GET(cc, 1);
     if (cs == NULL) return -2;  
428      do ce += GET(ce, 1); while (*ce == OP_ALT);      do ce += GET(ce, 1); while (*ce == OP_ALT);
429      if (cc > cs && cc < ce)      if ((cc > cs && cc < ce) || recurse_depth > 10)
430        had_recurse = TRUE;        had_recurse = TRUE;
431      else      else
432        branchlength += find_minlength(cs, startcode, options);        {
433          branchlength += find_minlength(cs, startcode, options, recurse_depth + 1);
434          }
435      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
436      break;      break;
437    
# Line 472  for (;;) Line 493  for (;;)
493      case OP_MARK:      case OP_MARK:
494      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
495      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     cc += _pcre_OP_lengths[op] + cc[1];  
     break;  
   
496      case OP_THEN_ARG:      case OP_THEN_ARG:
497      cc += _pcre_OP_lengths[op] + cc[1+LINK_SIZE];      cc += _pcre_OP_lengths[op] + cc[1];
498      break;      break;
499    
500      /* The remaining opcodes are just skipped over. */      /* The remaining opcodes are just skipped over. */
501    
     case OP_ACCEPT:  
502      case OP_CLOSE:      case OP_CLOSE:
503      case OP_COMMIT:      case OP_COMMIT:
504      case OP_FAIL:      case OP_FAIL:
# Line 687  do Line 704  do
704    while (try_next)    /* Loop for items in this branch */    while (try_next)    /* Loop for items in this branch */
705      {      {
706      int rc;      int rc;
707    
708      switch(*tcode)      switch(*tcode)
709        {        {
710        /* If we reach something we don't understand, it means a new opcode has        /* If we reach something we don't understand, it means a new opcode has
# Line 699  do Line 717  do
717        /* Fail for a valid opcode that implies no starting bits. */        /* Fail for a valid opcode that implies no starting bits. */
718    
719        case OP_ACCEPT:        case OP_ACCEPT:
720          case OP_ASSERT_ACCEPT:
721        case OP_ALLANY:        case OP_ALLANY:
722        case OP_ANY:        case OP_ANY:
723        case OP_ANYBYTE:        case OP_ANYBYTE:
724        case OP_CIRC:        case OP_CIRC:
725        case OP_CIRCM:        case OP_CIRCM:
726        case OP_CLOSE:        case OP_CLOSE:
727        case OP_COMMIT:        case OP_COMMIT:
728        case OP_COND:        case OP_COND:
729        case OP_CREF:        case OP_CREF:
730        case OP_DEF:        case OP_DEF:
731        case OP_DOLL:        case OP_DOLL:
732        case OP_DOLLM:        case OP_DOLLM:
733        case OP_END:        case OP_END:
734        case OP_EOD:        case OP_EOD:
735        case OP_EODN:        case OP_EODN:
736        case OP_EXTUNI:        case OP_EXTUNI:
737        case OP_FAIL:        case OP_FAIL:
738        case OP_MARK:        case OP_MARK:
# Line 721  do Line 740  do
740        case OP_NOT:        case OP_NOT:
741        case OP_NOTEXACT:        case OP_NOTEXACT:
742        case OP_NOTEXACTI:        case OP_NOTEXACTI:
743        case OP_NOTI:        case OP_NOTI:
744        case OP_NOTMINPLUS:        case OP_NOTMINPLUS:
745        case OP_NOTMINPLUSI:        case OP_NOTMINPLUSI:
746        case OP_NOTMINQUERY:        case OP_NOTMINQUERY:
# Line 749  do Line 768  do
768        case OP_NOTUPTOI:        case OP_NOTUPTOI:
769        case OP_NOT_HSPACE:        case OP_NOT_HSPACE:
770        case OP_NOT_VSPACE:        case OP_NOT_VSPACE:
       case OP_NOT_WORD_BOUNDARY:  
771        case OP_NRREF:        case OP_NRREF:
772        case OP_PROP:        case OP_PROP:
773        case OP_PRUNE:        case OP_PRUNE:
# Line 759  do Line 777  do
777        case OP_REFI:        case OP_REFI:
778        case OP_REVERSE:        case OP_REVERSE:
779        case OP_RREF:        case OP_RREF:
780        case OP_SCOND:        case OP_SCOND:
781        case OP_SET_SOM:        case OP_SET_SOM:
782        case OP_SKIP:        case OP_SKIP:
783        case OP_SKIP_ARG:        case OP_SKIP_ARG:
# Line 767  do Line 785  do
785        case OP_SOM:        case OP_SOM:
786        case OP_THEN:        case OP_THEN:
787        case OP_THEN_ARG:        case OP_THEN_ARG:
       case OP_WORD_BOUNDARY:  
788        case OP_XCLASS:        case OP_XCLASS:
789        return SSB_FAIL;        return SSB_FAIL;
790    
791          /* We can ignore word boundary tests. */
792    
793          case OP_WORD_BOUNDARY:
794          case OP_NOT_WORD_BOUNDARY:
795          tcode++;
796          break;
797    
798        /* If we hit a bracket or a positive lookahead assertion, recurse to set        /* If we hit a bracket or a positive lookahead assertion, recurse to set
799        bits from within the subpattern. If it can't find anything, we have to        bits from within the subpattern. If it can't find anything, we have to
800        give up. If it finds some mandatory character(s), we are done for this        give up. If it finds some mandatory character(s), we are done for this
# Line 785  do Line 809  do
809        case OP_CBRAPOS:        case OP_CBRAPOS:
810        case OP_SCBRAPOS:        case OP_SCBRAPOS:
811        case OP_ONCE:        case OP_ONCE:
812          case OP_ONCE_NC:
813        case OP_ASSERT:        case OP_ASSERT:
814        rc = set_start_bits(tcode, start_bits, utf8, cd);        rc = set_start_bits(tcode, start_bits, utf8, cd);
815        if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;        if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
# Line 1136  do Line 1161  do
1161            for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];            for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];
1162            }            }
1163    
1164          /* Advance past the bit map, and act on what follows. For a zero          /* Advance past the bit map, and act on what follows. For a zero
1165          minimum repeat, continue; otherwise stop processing. */          minimum repeat, continue; otherwise stop processing. */
1166    
1167          tcode += 32;          tcode += 32;
# Line 1154  do Line 1179  do
1179            if (((tcode[1] << 8) + tcode[2]) == 0) tcode += 5;            if (((tcode[1] << 8) + tcode[2]) == 0) tcode += 5;
1180              else try_next = FALSE;              else try_next = FALSE;
1181            break;            break;
1182    
1183            default:            default:
1184            try_next = FALSE;            try_next = FALSE;
1185            break;            break;
# Line 1200  pcre_study(const pcre *external_re, int Line 1225  pcre_study(const pcre *external_re, int
1225  int min;  int min;
1226  BOOL bits_set = FALSE;  BOOL bits_set = FALSE;
1227  uschar start_bits[32];  uschar start_bits[32];
1228  pcre_extra *extra;  pcre_extra *extra = NULL;
1229  pcre_study_data *study;  pcre_study_data *study;
1230  const uschar *tables;  const uschar *tables;
1231  uschar *code;  uschar *code;
# Line 1251  if ((re->options & PCRE_ANCHORED) == 0 & Line 1276  if ((re->options & PCRE_ANCHORED) == 0 &
1276    rc = set_start_bits(code, start_bits, (re->options & PCRE_UTF8) != 0,    rc = set_start_bits(code, start_bits, (re->options & PCRE_UTF8) != 0,
1277      &compile_block);      &compile_block);
1278    bits_set = rc == SSB_DONE;    bits_set = rc == SSB_DONE;
1279    if (rc == SSB_UNKNOWN) *errorptr = "internal error: opcode not recognized";    if (rc == SSB_UNKNOWN)
1280        {
1281        *errorptr = "internal error: opcode not recognized";
1282        return NULL;
1283        }
1284    }    }
1285    
1286  /* Find the minimum length of subject string. */  /* Find the minimum length of subject string. */
1287    
1288  switch(min = find_minlength(code, code, re->options))  switch(min = find_minlength(code, code, re->options, 0))
1289    {    {
1290    case -2: *errorptr = "internal error: missing capturing bracket"; break;    case -2: *errorptr = "internal error: missing capturing bracket"; return NULL;
1291    case -3: *errorptr = "internal error: opcode not recognized"; break;    case -3: *errorptr = "internal error: opcode not recognized"; return NULL;
1292    default: break;    default: break;
1293    }    }
1294    
1295  /* Return NULL if there's been an error or if no optimization is possible. */  /* If a set of starting bytes has been identified, or if the minimum length is
1296    greater than zero, or if JIT optimization has been requested, get a pcre_extra
1297    block and a pcre_study_data block. The study data is put in the latter, which
1298    is pointed to by the former, which may also get additional data set later by
1299    the calling program. At the moment, the size of pcre_study_data is fixed. We
1300    nevertheless save it in a field for returning via the pcre_fullinfo() function
1301    so that if it becomes variable in the future, we don't have to change that
1302    code. */
1303    
1304    if (bits_set || min > 0
1305    #ifdef SUPPORT_JIT
1306        || (options & PCRE_STUDY_JIT_COMPILE) != 0
1307    #endif
1308      )
1309      {
1310      extra = (pcre_extra *)(pcre_malloc)
1311        (sizeof(pcre_extra) + sizeof(pcre_study_data));
1312      if (extra == NULL)
1313        {
1314        *errorptr = "failed to get memory";
1315        return NULL;
1316        }
1317    
1318  if (*errorptr != NULL || (!bits_set && min < 0)) return NULL;    study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));
1319      extra->flags = PCRE_EXTRA_STUDY_DATA;
1320      extra->study_data = study;
1321    
1322  /* Get a pcre_extra block and a pcre_study_data block. The study data is put in    study->size = sizeof(pcre_study_data);
1323  the latter, which is pointed to by the former, which may also get additional    study->flags = 0;
 data set later by the calling program. At the moment, the size of  
 pcre_study_data is fixed. We nevertheless save it in a field for returning via  
 the pcre_fullinfo() function so that if it becomes variable in the future, we  
 don't have to change that code. */  
1324    
1325  extra = (pcre_extra *)(pcre_malloc)    if (bits_set)
1326    (sizeof(pcre_extra) + sizeof(pcre_study_data));      {
1327        study->flags |= PCRE_STUDY_MAPPED;
1328        memcpy(study->start_bits, start_bits, sizeof(start_bits));
1329        }
1330    
1331  if (extra == NULL)    /* Always set the minlength value in the block, because the JIT compiler
1332    {    makes use of it. However, don't set the bit unless the length is greater than
1333    *errorptr = "failed to get memory";    zero - the interpretive pcre_exec() and pcre_dfa_exec() needn't waste time
1334    return NULL;    checking the zero case. */
1335    
1336      if (min > 0)
1337        {
1338        study->flags |= PCRE_STUDY_MINLEN;
1339        study->minlength = min;
1340        }
1341      else study->minlength = 0;
1342    
1343      /* If JIT support was compiled and requested, attempt the JIT compilation.
1344      If no starting bytes were found, and the minimum length is zero, and JIT
1345      compilation fails, abandon the extra block and return NULL. */
1346    
1347    #ifdef SUPPORT_JIT
1348      extra->executable_jit = NULL;
1349      if ((options & PCRE_STUDY_JIT_COMPILE) != 0) _pcre_jit_compile(re, extra);
1350      if (study->flags == 0 && (extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) == 0)
1351        {
1352        pcre_free_study(extra);
1353        extra = NULL;
1354        }
1355    #endif
1356    }    }
1357    
1358  study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));  return extra;
1359  extra->flags = PCRE_EXTRA_STUDY_DATA;  }
 extra->study_data = study;  
1360    
 study->size = sizeof(pcre_study_data);  
 study->flags = 0;  
1361    
1362  if (bits_set)  /*************************************************
1363    {  *          Free the study data                   *
1364    study->flags |= PCRE_STUDY_MAPPED;  *************************************************/
   memcpy(study->start_bits, start_bits, sizeof(start_bits));  
   }  
1365    
1366  if (min >= 0)  /* This function frees the memory that was obtained by pcre_study().
   {  
   study->flags |= PCRE_STUDY_MINLEN;  
   study->minlength = min;  
   }  
1367    
1368  return extra;  Argument:   a pointer to the pcre_extra block
1369    Returns:    nothing
1370    */
1371    
1372    PCRE_EXP_DEFN void
1373    pcre_free_study(pcre_extra *extra)
1374    {
1375    #ifdef SUPPORT_JIT
1376    if ((extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0 &&
1377         extra->executable_jit != NULL)
1378      _pcre_jit_free(extra->executable_jit);
1379    #endif
1380    pcre_free(extra);
1381  }  }
1382    
1383  /* End of pcre_study.c */  /* End of pcre_study.c */

Legend:
Removed from v.605  
changed lines
  Added in v.754

  ViewVC Help
Powered by ViewVC 1.1.5