/[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 657 by ph10, Mon Aug 15 17:39:09 2011 UTC revision 787 by ph10, Tue Dec 6 15:37:24 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    had_accept  pointer to flag for (*ACCEPT) encountered    int             RECURSE depth
   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    BOOL *had_accept_ptr, int recurse_depth)    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 128  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, had_accept_ptr, recurse_depth);      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;
     if (*had_accept_ptr) return branchlength;  
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      /* Reached end of a branch; if it's a ket it is the end of a nested      /* ACCEPT makes things far too complicated; we have to give up. */
     call. If it's ALT it is an alternation in a nested call. If it is END it's  
     the end of the outer call. All can be handled by the same code. If it is  
     ACCEPT, it is essentially the same as END, but we set a flag so that  
     counting stops. */  
139    
140      case OP_ACCEPT:      case OP_ACCEPT:
141      case OP_ASSERT_ACCEPT:      case OP_ASSERT_ACCEPT:
142      *had_accept_ptr = TRUE;      return -1;
143      /* Fall through */  
144        /* 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 END it's
146        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:
152      case OP_KETRMAX:      case OP_KETRMAX:
# Line 285  for (;;) Line 286  for (;;)
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 379  for (;;) Line 382  for (;;)
382          }          }
383        else        else
384          {          {
385          d = find_minlength(cs, startcode, options, had_accept_ptr,          d = find_minlength(cs, startcode, options, recurse_depth);
           recurse_depth);  
         *had_accept_ptr = FALSE;  
386          }          }
387        }        }
388      else d = 0;      else d = 0;
# Line 424  for (;;) Line 425  for (;;)
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) || recurse_depth > 10)      if ((cc > cs && cc < ce) || recurse_depth > 10)
430        had_recurse = TRUE;        had_recurse = TRUE;
431      else      else
432        {        {
433        branchlength += find_minlength(cs, startcode, options, had_accept_ptr,        branchlength += find_minlength(cs, startcode, options, recurse_depth + 1);
         recurse_depth + 1);  
       *had_accept_ptr = FALSE;  
434        }        }
435      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
436      break;      break;
# Line 495  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. */
# Line 794  do Line 789  do
789        return SSB_FAIL;        return SSB_FAIL;
790    
791        /* We can ignore word boundary tests. */        /* We can ignore word boundary tests. */
792    
793        case OP_WORD_BOUNDARY:        case OP_WORD_BOUNDARY:
794        case OP_NOT_WORD_BOUNDARY:        case OP_NOT_WORD_BOUNDARY:
795        tcode++;        tcode++;
796        break;        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
# Line 814  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 1228  pcre_study(const pcre *external_re, int Line 1224  pcre_study(const pcre *external_re, int
1224  {  {
1225  int min;  int min;
1226  BOOL bits_set = FALSE;  BOOL bits_set = FALSE;
 BOOL had_accept = 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 1281  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, &had_accept, 0))  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      study->size = sizeof(pcre_study_data);
1323      study->flags = 0;
1324    
1325      /* Set the start bits always, to avoid unset memory errors if the
1326      study data is written to a file, but set the flag only if any of the bits
1327      are set, to save time looking when none are. */
1328    
1329  /* Get a pcre_extra block and a pcre_study_data block. The study data is put in    if (bits_set)
1330  the latter, which is pointed to by the former, which may also get additional      {
1331  data set later by the calling program. At the moment, the size of      study->flags |= PCRE_STUDY_MAPPED;
1332  pcre_study_data is fixed. We nevertheless save it in a field for returning via      memcpy(study->start_bits, start_bits, sizeof(start_bits));
1333  the pcre_fullinfo() function so that if it becomes variable in the future, we      }
1334  don't have to change that code. */    else memset(study->start_bits, 0, 32 * sizeof(uschar));
1335    
1336  extra = (pcre_extra *)(pcre_malloc)    /* Always set the minlength value in the block, because the JIT compiler
1337    (sizeof(pcre_extra) + sizeof(pcre_study_data));    makes use of it. However, don't set the bit unless the length is greater than
1338      zero - the interpretive pcre_exec() and pcre_dfa_exec() needn't waste time
1339      checking the zero case. */
1340    
1341  if (extra == NULL)    if (min > 0)
1342    {      {
1343    *errorptr = "failed to get memory";      study->flags |= PCRE_STUDY_MINLEN;
1344    return NULL;      study->minlength = min;
1345        }
1346      else study->minlength = 0;
1347    
1348      /* If JIT support was compiled and requested, attempt the JIT compilation.
1349      If no starting bytes were found, and the minimum length is zero, and JIT
1350      compilation fails, abandon the extra block and return NULL. */
1351    
1352    #ifdef SUPPORT_JIT
1353      extra->executable_jit = NULL;
1354      if ((options & PCRE_STUDY_JIT_COMPILE) != 0) _pcre_jit_compile(re, extra);
1355      if (study->flags == 0 && (extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) == 0)
1356        {
1357        pcre_free_study(extra);
1358        extra = NULL;
1359        }
1360    #endif
1361    }    }
1362    
1363  study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));  return extra;
1364  extra->flags = PCRE_EXTRA_STUDY_DATA;  }
 extra->study_data = study;  
1365    
 study->size = sizeof(pcre_study_data);  
 study->flags = 0;  
1366    
1367  if (bits_set)  /*************************************************
1368    {  *          Free the study data                   *
1369    study->flags |= PCRE_STUDY_MAPPED;  *************************************************/
   memcpy(study->start_bits, start_bits, sizeof(start_bits));  
   }  
1370    
1371  if (min >= 0)  /* This function frees the memory that was obtained by pcre_study().
   {  
   study->flags |= PCRE_STUDY_MINLEN;  
   study->minlength = min;  
   }  
1372    
1373  return extra;  Argument:   a pointer to the pcre_extra block
1374    Returns:    nothing
1375    */
1376    
1377    PCRE_EXP_DEFN void
1378    pcre_free_study(pcre_extra *extra)
1379    {
1380    #ifdef SUPPORT_JIT
1381    if ((extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0 &&
1382         extra->executable_jit != NULL)
1383      _pcre_jit_free(extra->executable_jit);
1384    #endif
1385    pcre_free(extra);
1386  }  }
1387    
1388  /* End of pcre_study.c */  /* End of pcre_study.c */
Legend:
Removed from v.657  
changed lines
  Added in v.787
webmaster@exim.org
 ViewVC Help
Powered by ViewVC 1.1.12