/[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 530 by ph10, Tue Jun 1 13:42:06 2010 UTC revision 532 by ph10, Tue Jun 1 16:21:42 2010 UTC
# Line 2392  for (++c; c <= d; c++) Line 2392  for (++c; c <= d; c++)
2392    
2393  return TRUE;  return TRUE;
2394  }  }
2395    
2396    
2397    
2398    /*************************************************
2399    *        Check a character and a property        *
2400    *************************************************/
2401    
2402    /* This function is called by check_auto_possessive() when a property item
2403    is adjacent to a fixed character.
2404    
2405    Arguments:
2406      c            the character
2407      ptype        the property type
2408      pdata        the data for the type
2409      negated      TRUE if it's a negated property (\P or \p{^)
2410    
2411    Returns:       TRUE if auto-possessifying is OK
2412    */
2413    
2414    static BOOL
2415    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2416    {
2417    const ucd_record *prop = GET_UCD(c);
2418    switch(ptype)
2419      {
2420      case PT_LAMP:
2421      return (prop->chartype == ucp_Lu ||
2422              prop->chartype == ucp_Ll ||
2423              prop->chartype == ucp_Lt) == negated;
2424    
2425      case PT_GC:
2426      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2427    
2428      case PT_PC:
2429      return (pdata == prop->chartype) == negated;
2430    
2431      case PT_SC:
2432      return (pdata == prop->script) == negated;
2433    
2434      /* These are specials */
2435    
2436      case PT_ALNUM:
2437      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2438              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2439    
2440      case PT_SPACE:    /* Perl space */
2441      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2442              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2443              == negated;
2444    
2445      case PT_PXSPACE:  /* POSIX space */
2446      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2447              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2448              c == CHAR_FF || c == CHAR_CR)
2449              == negated;
2450    
2451      case PT_WORD:
2452      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2453              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2454              c == CHAR_UNDERSCORE) == negated;
2455      }
2456    return FALSE;
2457    }
2458  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2459    
2460    
# Line 2405  whether the next thing could possibly ma Line 2468  whether the next thing could possibly ma
2468  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2469    
2470  Arguments:  Arguments:
2471    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2472    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2473    ptr           next character in pattern    ptr           next character in pattern
2474    options       options bits    options       options bits
2475    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2417  Returns:        TRUE if possessifying is Line 2478  Returns:        TRUE if possessifying is
2478  */  */
2479    
2480  static BOOL  static BOOL
2481  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2482    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2483  {  {
2484  int next;  int c, next;
2485    int op_code = *previous++;
2486    
2487  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2488    
# Line 2481  if (*ptr == CHAR_ASTERISK || *ptr == CHA Line 2543  if (*ptr == CHAR_ASTERISK || *ptr == CHA
2543    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2544      return FALSE;      return FALSE;
2545    
2546  /* Now compare the next item with the previous opcode. If the previous is a  /* Now compare the next item with the previous opcode. First, handle cases when
2547  positive single character match, "item" either contains the character or, if  the next item is a character. */
 "item" is greater than 127 in utf8 mode, the character's bytes are in  
 utf8_char. */  
   
   
 /* Handle cases when the next item is a character. */  
2548    
2549  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2550    {    {
2551    case OP_CHAR:    case OP_CHAR:
2552  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2553    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2554  #else  #else
2555    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2556  #endif  #endif
2557    return item != next;    return c != next;
2558    
2559    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARNC (caseless character) we must check the other case. If we have
2560    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
2561    high-valued characters. */    high-valued characters. */
2562    
2563    case OP_CHARNC:    case OP_CHARNC:
2564  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2565    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2566  #endif  #else
2567    if (item == next) return FALSE;    c = *previous;
2568    #endif
2569      if (c == next) return FALSE;
2570  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2571    if (utf8)    if (utf8)
2572      {      {
# Line 2518  if (next >= 0) switch(op_code) Line 2577  if (next >= 0) switch(op_code)
2577  #else  #else
2578      othercase = NOTACHAR;      othercase = NOTACHAR;
2579  #endif  #endif
2580      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2581      }      }
2582    else    else
2583  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2584    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2585    
2586    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, its data is always a single-byte character. */
2587    
2588    case OP_NOT:    case OP_NOT:
2589    if (item == next) return TRUE;    if ((c = *previous) == next) return TRUE;
2590    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2591  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2592    if (utf8)    if (utf8)
# Line 2539  if (next >= 0) switch(op_code) Line 2598  if (next >= 0) switch(op_code)
2598  #else  #else
2599      othercase = NOTACHAR;      othercase = NOTACHAR;
2600  #endif  #endif
2601      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2602      }      }
2603    else    else
2604  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2605    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2606    
2607    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2608    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
# Line 2611  if (next >= 0) switch(op_code) Line 2670  if (next >= 0) switch(op_code)
2670      return op_code != OP_NOT_VSPACE;      return op_code != OP_NOT_VSPACE;
2671      }      }
2672    
2673    #ifdef SUPPORT_UCP
2674      case OP_PROP:
2675      return check_char_prop(next, previous[0], previous[1], FALSE);
2676    
2677      case OP_NOTPROP:
2678      return check_char_prop(next, previous[0], previous[1], TRUE);
2679    #endif
2680    
2681    default:    default:
2682    return FALSE;    return FALSE;
2683    }    }
# Line 2619  if (next >= 0) switch(op_code) Line 2686  if (next >= 0) switch(op_code)
2686  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2687  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2688  generated only when PCRE_UCP is *not* set, that is, when only ASCII  generated only when PCRE_UCP is *not* set, that is, when only ASCII
2689  characteristics are recognized. */  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2690    replaced by OP_PROP codes when PCRE_UCP is set. */
2691    
2692  switch(op_code)  switch(op_code)
2693    {    {
2694    case OP_CHAR:    case OP_CHAR:
2695    case OP_CHARNC:    case OP_CHARNC:
2696  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2697    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2698  #endif  #else
2699      c = *previous;
2700    #endif
2701    switch(-next)    switch(-next)
2702      {      {
2703      case ESC_d:      case ESC_d:
2704      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2705    
2706      case ESC_D:      case ESC_D:
2707      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2708    
2709      case ESC_s:      case ESC_s:
2710      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2711    
2712      case ESC_S:      case ESC_S:
2713      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2714    
2715      case ESC_w:      case ESC_w:
2716      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2717    
2718      case ESC_W:      case ESC_W:
2719      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2720    
2721      case ESC_h:      case ESC_h:
2722      case ESC_H:      case ESC_H:
2723      switch(item)      switch(c)
2724        {        {
2725        case 0x09:        case 0x09:
2726        case 0x20:        case 0x20:
# Line 2678  switch(op_code) Line 2748  switch(op_code)
2748    
2749      case ESC_v:      case ESC_v:
2750      case ESC_V:      case ESC_V:
2751      switch(item)      switch(c)
2752        {        {
2753        case 0x0a:        case 0x0a:
2754        case 0x0b:        case 0x0b:
# Line 2691  switch(op_code) Line 2761  switch(op_code)
2761        default:        default:
2762        return -next == ESC_v;        return -next == ESC_v;
2763        }        }
2764    
2765        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2766        their substitutions and process them. The result will always be either
2767        -ESC_p or -ESC_P. Then fall through to process those values. */
2768    
2769    #ifdef SUPPORT_UCP
2770        case ESC_du:
2771        case ESC_DU:
2772        case ESC_wu:
2773        case ESC_WU:
2774        case ESC_su:
2775        case ESC_SU:
2776          {
2777          int temperrorcode = 0;
2778          ptr = substitutes[-next - ESC_DU];
2779          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2780          if (temperrorcode != 0) return FALSE;
2781          ptr++;    /* For compatibility */
2782          }
2783        /* Fall through */
2784    
2785        case ESC_p:
2786        case ESC_P:
2787          {
2788          int ptype, pdata, errorcodeptr;
2789          BOOL negated;
2790    
2791          ptr--;      /* Make ptr point at the p or P */
2792          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2793          if (ptype < 0) return FALSE;
2794          ptr++;      /* Point past the final curly ket */
2795    
2796          /* If the property item is optional, we have to give up. (When generated
2797          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2798          to the original \d etc. At this point, ptr will point to a zero byte. */
2799    
2800          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2801            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2802              return FALSE;
2803    
2804          /* Do the property check. */
2805    
2806          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2807          }
2808    #endif
2809    
2810      default:      default:
2811      return FALSE;      return FALSE;
2812      }      }
2813    
2814      /* In principle, support for Unicode properties should be integrated here as
2815      well. It means re-organizing the above code so as to get hold of the property
2816      values before switching on the op-code. However, I wonder how many patterns
2817      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2818      these op-codes are never generated.) */
2819    
2820    case OP_DIGIT:    case OP_DIGIT:
2821    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
# Line 3998  for (;; ptr++) Line 4119  for (;; ptr++)
4119    
4120        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4121            repeat_max < 0 &&            repeat_max < 0 &&
4122            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4123          {          {
4124          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4125          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4020  for (;; ptr++) Line 4140  for (;; ptr++)
4140        c = previous[1];        c = previous[1];
4141        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4142            repeat_max < 0 &&            repeat_max < 0 &&
4143            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4144          {          {
4145          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4146          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4044  for (;; ptr++) Line 4164  for (;; ptr++)
4164    
4165        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4166            repeat_max < 0 &&            repeat_max < 0 &&
4167            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4168          {          {
4169          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4170          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;

Legend:
Removed from v.530  
changed lines
  Added in v.532

  ViewVC Help
Powered by ViewVC 1.1.5