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

Diff of /code/trunk/pcre_compile.c

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

revision 213 by ph10, Wed Aug 15 11:34:14 2007 UTC revision 227 by ph10, Tue Aug 21 15:00:15 2007 UTC
# Line 283  static const char *error_texts[] = { Line 283  static const char *error_texts[] = {
283    "(*VERB) with an argument is not supported",    "(*VERB) with an argument is not supported",
284    /* 60 */    /* 60 */
285    "(*VERB) not recognized",    "(*VERB) not recognized",
286    "number is too big"    "number is too big"
287  };  };
288    
289    
# Line 524  else Line 524  else
524      c = 0;      c = 0;
525      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
526        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - '0';
527    
528      if (c < 0)      if (c < 0)
529        {        {
530        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
531        break;        break;
532        }        }
533    
534      if (c == 0 || (braced && *(++ptr) != '}'))      if (c == 0 || (braced && *(++ptr) != '}'))
535        {        {
# Line 574  else Line 574  else
574        if (c < 0)        if (c < 0)
575          {          {
576          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
577          break;          break;
578          }          }
579        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
580          {          {
581          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 950  for (; *ptr != 0; ptr++) Line 950  for (; *ptr != 0; ptr++)
950      {      {
951      while (*(++ptr) != ']')      while (*(++ptr) != ']')
952        {        {
953          if (*ptr == 0) return -1;
954        if (*ptr == '\\')        if (*ptr == '\\')
955          {          {
956          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) return -1;
# Line 1105  for (;;) Line 1106  for (;;)
1106    {    {
1107    int d;    int d;
1108    register int op = *cc;    register int op = *cc;
   
1109    switch (op)    switch (op)
1110      {      {
1111      case OP_CBRA:      case OP_CBRA:
# Line 1194  for (;;) Line 1194  for (;;)
1194    
1195      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1196      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1197        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1198      cc += 4;      cc += 4;
1199      break;      break;
1200    
# Line 1302  for (;;) Line 1303  for (;;)
1303      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1304      }      }
1305    
1306    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* Otherwise, we can get the item's length from the table, except that for
1307    a multi-byte character. The length in the table is a minimum, so we have to    repeated character types, we have to test for \p and \P, which have an extra
1308    arrange to skip the extra bytes. */    two bytes of parameters. */
1309    
1310    else    else
1311      {      {
1312        switch(c)
1313          {
1314          case OP_TYPESTAR:
1315          case OP_TYPEMINSTAR:
1316          case OP_TYPEPLUS:
1317          case OP_TYPEMINPLUS:
1318          case OP_TYPEQUERY:
1319          case OP_TYPEMINQUERY:
1320          case OP_TYPEPOSSTAR:
1321          case OP_TYPEPOSPLUS:
1322          case OP_TYPEPOSQUERY:
1323          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1324          break;
1325    
1326          case OP_TYPEUPTO:
1327          case OP_TYPEMINUPTO:
1328          case OP_TYPEEXACT:
1329          case OP_TYPEPOSUPTO:
1330          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1331          break;
1332          }
1333    
1334        /* Add in the fixed length from the table */
1335    
1336      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1337    
1338      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1339      a multi-byte character. The length in the table is a minimum, so we have to
1340      arrange to skip the extra bytes. */
1341    
1342  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1343      if (utf8) switch(c)      if (utf8) switch(c)
1344        {        {
# Line 1366  for (;;) Line 1396  for (;;)
1396    
1397    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1398    
1399    /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes    /* Otherwise, we can get the item's length from the table, except that for
1400    that are followed by a character may be followed by a multi-byte character.    repeated character types, we have to test for \p and \P, which have an extra
1401    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters. */
   bytes. */  
1402    
1403    else    else
1404      {      {
1405        switch(c)
1406          {
1407          case OP_TYPESTAR:
1408          case OP_TYPEMINSTAR:
1409          case OP_TYPEPLUS:
1410          case OP_TYPEMINPLUS:
1411          case OP_TYPEQUERY:
1412          case OP_TYPEMINQUERY:
1413          case OP_TYPEPOSSTAR:
1414          case OP_TYPEPOSPLUS:
1415          case OP_TYPEPOSQUERY:
1416          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1417          break;
1418    
1419          case OP_TYPEPOSUPTO:
1420          case OP_TYPEUPTO:
1421          case OP_TYPEMINUPTO:
1422          case OP_TYPEEXACT:
1423          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1424          break;
1425          }
1426    
1427        /* Add in the fixed length from the table */
1428    
1429      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1430    
1431        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1432        by a multi-byte character. The length in the table is a minimum, so we have
1433        to arrange to skip the extra bytes. */
1434    
1435  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1436      if (utf8) switch(c)      if (utf8) switch(c)
1437        {        {
# Line 1469  for (code = first_significant_code(code Line 1527  for (code = first_significant_code(code
1527    
1528    switch (c)    switch (c)
1529      {      {
1530      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1531        cannot be represented just by a bit map. This includes negated single
1532        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1533        actual length is stored in the compiled code, so we must update "code"
1534        here. */
1535    
1536  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1537      case OP_XCLASS:      case OP_XCLASS:
1538      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1539      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1540  #endif  #endif
1541    
# Line 1535  for (code = first_significant_code(code Line 1597  for (code = first_significant_code(code
1597      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1598      return FALSE;      return FALSE;
1599    
1600        /* These are going to continue, as they may be empty, but we have to
1601        fudge the length for the \p and \P cases. */
1602    
1603        case OP_TYPESTAR:
1604        case OP_TYPEMINSTAR:
1605        case OP_TYPEPOSSTAR:
1606        case OP_TYPEQUERY:
1607        case OP_TYPEMINQUERY:
1608        case OP_TYPEPOSQUERY:
1609        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1610        break;
1611    
1612        /* Same for these */
1613    
1614        case OP_TYPEUPTO:
1615        case OP_TYPEMINUPTO:
1616        case OP_TYPEPOSUPTO:
1617        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1618        break;
1619    
1620      /* End of branch */      /* End of branch */
1621    
1622      case OP_KET:      case OP_KET:
# Line 1697  adjust_recurse(uschar *group, int adjust Line 1779  adjust_recurse(uschar *group, int adjust
1779    uschar *save_hwm)    uschar *save_hwm)
1780  {  {
1781  uschar *ptr = group;  uschar *ptr = group;
1782    
1783  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1784    {    {
1785    int offset;    int offset;
# Line 2666  for (;; ptr++) Line 2749  for (;; ptr++)
2749            else inescq = TRUE;            else inescq = TRUE;
2750            continue;            continue;
2751            }            }
2752            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2753    
2754          if (c < 0)          if (c < 0)
2755            {            {
# Line 3112  for (;; ptr++) Line 3196  for (;; ptr++)
3196        goto FAILED;        goto FAILED;
3197        }        }
3198    
3199        /* Remember whether \r or \n are in this class */
3200    
3201        if (negate_class)
3202          {
3203          if ((classbits[1] & 0x24) != 0x24) cd->external_options |= PCRE_HASCRORLF;
3204          }
3205        else
3206          {
3207          if ((classbits[1] & 0x24) != 0) cd->external_options |= PCRE_HASCRORLF;
3208          }
3209    
3210      /* If class_charcount is 1, we saw precisely one character whose value is      /* If class_charcount is 1, we saw precisely one character whose value is
3211      less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we      less than 256. As long as there were no characters >= 128 and there was no
3212      can optimize the negative case only if there were no characters >= 128      use of \p or \P, in other words, no use of any XCLASS features, we can
3213      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3214      single-bytes only. This is an historical hangover. Maybe one day we can  
3215      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3216        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3217        operate on single-bytes only. This is an historical hangover. Maybe one day
3218        we can tidy these opcodes to handle multi-byte characters.
3219    
3220      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
3221      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 3127  for (;; ptr++) Line 3225  for (;; ptr++)
3225      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3226    
3227  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3228      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3229            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3230  #else  #else
3231      if (class_charcount == 1)      if (class_charcount == 1)
3232  #endif  #endif
# Line 4740  for (;; ptr++) Line 4836  for (;; ptr++)
4836        goto FAILED;        goto FAILED;
4837        }        }
4838    
4839      /* In the pre-compile phase, update the length by the length of the nested      /* In the pre-compile phase, update the length by the length of the group,
4840      group, less the brackets at either end. Then reduce the compiled code to      less the brackets at either end. Then reduce the compiled code to just a
4841      just the brackets so that it doesn't use much memory if it is duplicated by      set of non-capturing brackets so that it doesn't use much memory if it is
4842      a quantifier. */      duplicated by a quantifier.*/
4843    
4844      if (lengthptr != NULL)      if (lengthptr != NULL)
4845        {        {
# Line 4753  for (;; ptr++) Line 4849  for (;; ptr++)
4849          goto FAILED;          goto FAILED;
4850          }          }
4851        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
4852        code++;        *code++ = OP_BRA;
4853        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
4854        *code++ = OP_KET;        *code++ = OP_KET;
4855        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
4856          break;    /* No need to waste time with special character handling */
4857        }        }
4858    
4859      /* Otherwise update the main code pointer to the end of the group. */      /* Otherwise update the main code pointer to the end of the group. */
4860    
4861      else code = tempcode;      code = tempcode;
4862    
4863      /* For a DEFINE group, required and first character settings are not      /* For a DEFINE group, required and first character settings are not
4864      relevant. */      relevant. */
# Line 4965  for (;; ptr++) Line 5062  for (;; ptr++)
5062      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;
5063      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
5064    
5065        /* Remember if \r or \n were seen */
5066    
5067        if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')
5068          cd->external_options |= PCRE_HASCRORLF;
5069    
5070      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
5071      byte, set it from this character, but revert to none on a zero repeat.      byte, set it from this character, but revert to none on a zero repeat.
5072      Otherwise, leave the firstbyte value alone, and don't change it on a zero      Otherwise, leave the firstbyte value alone, and don't change it on a zero
# Line 5564  real_pcre *re; Line 5666  real_pcre *re;
5666  int length = 1;  /* For final END opcode */  int length = 1;  /* For final END opcode */
5667  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
5668  int errorcode = 0;  int errorcode = 0;
5669    int skipatstart = 0;
5670  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
5671  BOOL utf8;  BOOL utf8;
5672  #endif  #endif
# Line 5642  cd->fcc = tables + fcc_offset; Line 5745  cd->fcc = tables + fcc_offset;
5745  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
5746  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
5747    
5748    /* Check for newline settings at the start of the pattern, and remember the
5749    offset for later. */
5750    
5751    if (ptr[0] == '(' && ptr[1] == '*')
5752      {
5753      int newnl = 0;
5754      if (strncmp((char *)(ptr+2), "CR)", 3) == 0)
5755        { skipatstart = 5; newnl = PCRE_NEWLINE_CR; }
5756      else if (strncmp((char *)(ptr+2), "LF)", 3)  == 0)
5757        { skipatstart = 5; newnl = PCRE_NEWLINE_LF; }
5758      else if (strncmp((char *)(ptr+2), "CRLF)", 5)  == 0)
5759        { skipatstart = 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
5760      else if (strncmp((char *)(ptr+2), "ANY)", 4) == 0)
5761        { skipatstart = 6; newnl = PCRE_NEWLINE_ANY; }
5762      else if (strncmp((char *)(ptr+2), "ANYCRLF)", 8)  == 0)
5763        { skipatstart = 10; newnl = PCRE_NEWLINE_ANYCRLF; }
5764      if (skipatstart > 0)
5765        options = (options & ~PCRE_NEWLINE_BITS) | newnl;
5766      }
5767    
5768  /* Handle different types of newline. The three bits give seven cases. The  /* Handle different types of newline. The three bits give seven cases. The
5769  current code allows for fixed one- or two-byte sequences, plus "any" and  current code allows for fixed one- or two-byte sequences, plus "any" and
5770  "anycrlf". */  "anycrlf". */
5771    
5772  switch (options & (PCRE_NEWLINE_CRLF | PCRE_NEWLINE_ANY))  switch (options & PCRE_NEWLINE_BITS)
5773    {    {
5774    case 0: newline = NEWLINE; break;   /* Compile-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
5775    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = '\r'; break;
5776    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = '\n'; break;
5777    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
# Line 5720  been put into the cd block so that they Line 5843  been put into the cd block so that they
5843  found within the regex right at the beginning. Bringing initial option settings  found within the regex right at the beginning. Bringing initial option settings
5844  outside can help speed up starting point checks. */  outside can help speed up starting point checks. */
5845    
5846    ptr += skipatstart;
5847  code = cworkspace;  code = cworkspace;
5848  *code = OP_BRA;  *code = OP_BRA;
5849  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,
# Line 5789  cd->had_accept = FALSE; Line 5913  cd->had_accept = FALSE;
5913  error, errorcode will be set non-zero, so we don't need to look at the result  error, errorcode will be set non-zero, so we don't need to look at the result
5914  of the function here. */  of the function here. */
5915    
5916  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern + skipatstart;
5917  code = (uschar *)codestart;  code = (uschar *)codestart;
5918  *code = OP_BRA;  *code = OP_BRA;
5919  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,
# Line 5897  case when building a production library. Line 6021  case when building a production library.
6021  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
6022    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
6023    
6024  if (re->options != 0)  printf("Options=%08x\n", re->options);
   {  
   printf("%s%s%s%s%s%s%s%s%s\n",  
     ((re->options & PCRE_NOPARTIAL) != 0)? "nopartial " : "",  
     ((re->options & PCRE_ANCHORED) != 0)? "anchored " : "",  
     ((re->options & PCRE_CASELESS) != 0)? "caseless " : "",  
     ((re->options & PCRE_EXTENDED) != 0)? "extended " : "",  
     ((re->options & PCRE_MULTILINE) != 0)? "multiline " : "",  
     ((re->options & PCRE_DOTALL) != 0)? "dotall " : "",  
     ((re->options & PCRE_DOLLAR_ENDONLY) != 0)? "endonly " : "",  
     ((re->options & PCRE_EXTRA) != 0)? "extra " : "",  
     ((re->options & PCRE_UNGREEDY) != 0)? "ungreedy " : "");  
   }  
6025    
6026  if ((re->options & PCRE_FIRSTSET) != 0)  if ((re->options & PCRE_FIRSTSET) != 0)
6027    {    {

Legend:
Removed from v.213  
changed lines
  Added in v.227

  ViewVC Help
Powered by ViewVC 1.1.5