/[pcre]/code/tags/pcre-4.0/study.c
ViewVC logotype

Diff of /code/tags/pcre-4.0/study.c

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

code/trunk/study.c revision 23 by nigel, Sat Feb 24 21:38:41 2007 UTC code/tags/pcre-4.0/study.c revision 64 by nigel, Sat Feb 24 21:40:05 2007 UTC
# Line 9  the file Tech.Notes for some information Line 9  the file Tech.Notes for some information
9    
10  Written by: Philip Hazel <ph10@cam.ac.uk>  Written by: Philip Hazel <ph10@cam.ac.uk>
11    
12             Copyright (c) 1998 University of Cambridge             Copyright (c) 1997-2002 University of Cambridge
13    
14  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
15  Permission is granted to anyone to use this software for any purpose on any  Permission is granted to anyone to use this software for any purpose on any
# Line 25  restrictions: Line 25  restrictions:
25    
26  3. Altered versions must be plainly marked as such, and must not be  3. Altered versions must be plainly marked as such, and must not be
27     misrepresented as being the original software.     misrepresented as being the original software.
28    
29    4. If PCRE is embedded in any software that is released under the GNU
30       General Purpose Licence (GPL), then the terms of that licence shall
31       supersede any condition above with which it is incompatible.
32  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
33  */  */
34    
# Line 47  Arguments: Line 51  Arguments:
51    start_bits    points to the bit map    start_bits    points to the bit map
52    c             is the character    c             is the character
53    caseless      the caseless flag    caseless      the caseless flag
54      cd            the block with char table pointers
55    
56  Returns:        nothing  Returns:        nothing
57  */  */
58    
59  static void  static void
60  set_bit(uschar *start_bits, int c, BOOL caseless)  set_bit(uschar *start_bits, int c, BOOL caseless, compile_data *cd)
61  {  {
62  start_bits[c/8] |= (1 << (c&7));  start_bits[c/8] |= (1 << (c&7));
63  if (caseless && (pcre_ctypes[c] & ctype_letter) != 0)  if (caseless && (cd->ctypes[c] & ctype_letter) != 0)
64    start_bits[pcre_fcc[c]/8] |= (1 << (pcre_fcc[c]&7));    start_bits[cd->fcc[c]/8] |= (1 << (cd->fcc[c]&7));
65  }  }
66    
67    
# Line 73  Arguments: Line 78  Arguments:
78    code         points to an expression    code         points to an expression
79    start_bits   points to a 32-byte table, initialized to 0    start_bits   points to a 32-byte table, initialized to 0
80    caseless     the current state of the caseless flag    caseless     the current state of the caseless flag
81      utf8         TRUE if in UTF-8 mode
82      cd           the block with char table pointers
83    
84  Returns:       TRUE if table built, FALSE otherwise  Returns:       TRUE if table built, FALSE otherwise
85  */  */
86    
87  static BOOL  static BOOL
88  set_start_bits(const uschar *code, uschar *start_bits, BOOL caseless)  set_start_bits(const uschar *code, uschar *start_bits, BOOL caseless,
89      BOOL utf8, compile_data *cd)
90  {  {
91  register int c;  register int c;
92    
93    /* This next statement and the later reference to dummy are here in order to
94    trick the optimizer of the IBM C compiler for OS/2 into generating correct
95    code. Apparently IBM isn't going to fix the problem, and we would rather not
96    disable optimization (in this module it actually makes a big difference, and
97    the pcre module can use all the optimization it can get). */
98    
99    volatile int dummy;
100    
101  do  do
102    {    {
103    const uschar *tcode = code + 3;    const uschar *tcode = code + 1 + LINK_SIZE;
104    BOOL try_next = TRUE;    BOOL try_next = TRUE;
105    
106    while (try_next)    while (try_next)
107      {      {
     try_next = FALSE;  
   
108      /* If a branch starts with a bracket or a positive lookahead assertion,      /* If a branch starts with a bracket or a positive lookahead assertion,
109      recurse to set bits from within them. That's all for this branch. */      recurse to set bits from within them. That's all for this branch. */
110    
111      if ((int)*tcode >= OP_BRA || *tcode == OP_ASSERT)      if ((int)*tcode >= OP_BRA || *tcode == OP_ASSERT)
112        {        {
113        if (!set_start_bits(tcode, start_bits, caseless)) return FALSE;        if (!set_start_bits(tcode, start_bits, caseless, utf8, cd))
114            return FALSE;
115          try_next = FALSE;
116        }        }
117    
118      else switch(*tcode)      else switch(*tcode)
# Line 104  do Line 120  do
120        default:        default:
121        return FALSE;        return FALSE;
122    
123          /* Skip over callout */
124    
125          case OP_CALLOUT:
126          tcode += 2;
127          break;
128    
129          /* Skip over extended extraction bracket number */
130    
131          case OP_BRANUMBER:
132          tcode += 3;
133          break;
134    
135        /* Skip over lookbehind and negative lookahead assertions */        /* Skip over lookbehind and negative lookahead assertions */
136    
137        case OP_ASSERT_NOT:        case OP_ASSERT_NOT:
138        case OP_ASSERTBACK:        case OP_ASSERTBACK:
139        case OP_ASSERTBACK_NOT:        case OP_ASSERTBACK_NOT:
140        try_next = TRUE;        do tcode += GET(tcode, 1); while (*tcode == OP_ALT);
141        do tcode += (tcode[1] << 8) + tcode[2]; while (*tcode == OP_ALT);        tcode += 1+LINK_SIZE;
       tcode += 3;  
142        break;        break;
143    
144        /* Skip over an option setting, changing the caseless flag */        /* Skip over an option setting, changing the caseless flag */
# Line 119  do Line 146  do
146        case OP_OPT:        case OP_OPT:
147        caseless = (tcode[1] & PCRE_CASELESS) != 0;        caseless = (tcode[1] & PCRE_CASELESS) != 0;
148        tcode += 2;        tcode += 2;
       try_next = TRUE;  
149        break;        break;
150    
151        /* BRAZERO does the bracket, but carries on. */        /* BRAZERO does the bracket, but carries on. */
152    
153        case OP_BRAZERO:        case OP_BRAZERO:
154        case OP_BRAMINZERO:        case OP_BRAMINZERO:
155        if (!set_start_bits(++tcode, start_bits, caseless)) return FALSE;        if (!set_start_bits(++tcode, start_bits, caseless, utf8, cd))
156        do tcode += (tcode[1] << 8) + tcode[2]; while (*tcode == OP_ALT);          return FALSE;
157        tcode += 3;        dummy = 1;
158        try_next = TRUE;        do tcode += GET(tcode,1); while (*tcode == OP_ALT);
159          tcode += 1+LINK_SIZE;
160        break;        break;
161    
162        /* Single-char * or ? sets the bit and tries the next item */        /* Single-char * or ? sets the bit and tries the next item */
# Line 138  do Line 165  do
165        case OP_MINSTAR:        case OP_MINSTAR:
166        case OP_QUERY:        case OP_QUERY:
167        case OP_MINQUERY:        case OP_MINQUERY:
168        set_bit(start_bits, tcode[1], caseless);        set_bit(start_bits, tcode[1], caseless, cd);
169        tcode += 2;        tcode += 2;
170        try_next = TRUE;  #ifdef SUPPORT_UTF8
171          if (utf8) while ((*tcode & 0xc0) == 0x80) tcode++;
172    #endif
173        break;        break;
174    
175        /* Single-char upto sets the bit and tries the next */        /* Single-char upto sets the bit and tries the next */
176    
177        case OP_UPTO:        case OP_UPTO:
178        case OP_MINUPTO:        case OP_MINUPTO:
179        set_bit(start_bits, tcode[3], caseless);        set_bit(start_bits, tcode[3], caseless, cd);
180        tcode += 4;        tcode += 4;
181        try_next = TRUE;  #ifdef SUPPORT_UTF8
182          if (utf8) while ((*tcode & 0xc0) == 0x80) tcode++;
183    #endif
184        break;        break;
185    
186        /* At least one single char sets the bit and stops */        /* At least one single char sets the bit and stops */
# Line 162  do Line 193  do
193    
194        case OP_PLUS:        case OP_PLUS:
195        case OP_MINPLUS:        case OP_MINPLUS:
196        set_bit(start_bits, tcode[1], caseless);        set_bit(start_bits, tcode[1], caseless, cd);
197          try_next = FALSE;
198        break;        break;
199    
200        /* Single character type sets the bits and stops */        /* Single character type sets the bits and stops */
201    
202        case OP_NOT_DIGIT:        case OP_NOT_DIGIT:
203        for (c = 0; c < 32; c++) start_bits[c] |= ~pcre_cbits[c+cbit_digit];        for (c = 0; c < 32; c++)
204            start_bits[c] |= ~cd->cbits[c+cbit_digit];
205          try_next = FALSE;
206        break;        break;
207    
208        case OP_DIGIT:        case OP_DIGIT:
209        for (c = 0; c < 32; c++) start_bits[c] |= pcre_cbits[c+cbit_digit];        for (c = 0; c < 32; c++)
210            start_bits[c] |= cd->cbits[c+cbit_digit];
211          try_next = FALSE;
212        break;        break;
213    
214        case OP_NOT_WHITESPACE:        case OP_NOT_WHITESPACE:
215        for (c = 0; c < 32; c++) start_bits[c] |= ~pcre_cbits[c+cbit_space];        for (c = 0; c < 32; c++)
216            start_bits[c] |= ~cd->cbits[c+cbit_space];
217          try_next = FALSE;
218        break;        break;
219    
220        case OP_WHITESPACE:        case OP_WHITESPACE:
221        for (c = 0; c < 32; c++) start_bits[c] |= pcre_cbits[c+cbit_space];        for (c = 0; c < 32; c++)
222            start_bits[c] |= cd->cbits[c+cbit_space];
223          try_next = FALSE;
224        break;        break;
225    
226        case OP_NOT_WORDCHAR:        case OP_NOT_WORDCHAR:
227        for (c = 0; c < 32; c++)        for (c = 0; c < 32; c++)
228          start_bits[c] |= ~(pcre_cbits[c] | pcre_cbits[c+cbit_word]);          start_bits[c] |= ~cd->cbits[c+cbit_word];
229          try_next = FALSE;
230        break;        break;
231    
232        case OP_WORDCHAR:        case OP_WORDCHAR:
233        for (c = 0; c < 32; c++)        for (c = 0; c < 32; c++)
234          start_bits[c] |= (pcre_cbits[c] | pcre_cbits[c+cbit_word]);          start_bits[c] |= cd->cbits[c+cbit_word];
235          try_next = FALSE;
236        break;        break;
237    
238        /* One or more character type fudges the pointer and restarts, knowing        /* One or more character type fudges the pointer and restarts, knowing
# Line 199  do Line 241  do
241        case OP_TYPEPLUS:        case OP_TYPEPLUS:
242        case OP_TYPEMINPLUS:        case OP_TYPEMINPLUS:
243        tcode++;        tcode++;
       try_next = TRUE;  
244        break;        break;
245    
246        case OP_TYPEEXACT:        case OP_TYPEEXACT:
247        tcode += 3;        tcode += 3;
       try_next = TRUE;  
248        break;        break;
249    
250        /* Zero or more repeats of character types set the bits and then        /* Zero or more repeats of character types set the bits and then
# Line 221  do Line 261  do
261        switch(tcode[1])        switch(tcode[1])
262          {          {
263          case OP_NOT_DIGIT:          case OP_NOT_DIGIT:
264          for (c = 0; c < 32; c++) start_bits[c] |= ~pcre_cbits[c+cbit_digit];          for (c = 0; c < 32; c++)
265              start_bits[c] |= ~cd->cbits[c+cbit_digit];
266          break;          break;
267    
268          case OP_DIGIT:          case OP_DIGIT:
269          for (c = 0; c < 32; c++) start_bits[c] |= pcre_cbits[c+cbit_digit];          for (c = 0; c < 32; c++)
270              start_bits[c] |= cd->cbits[c+cbit_digit];
271          break;          break;
272    
273          case OP_NOT_WHITESPACE:          case OP_NOT_WHITESPACE:
274          for (c = 0; c < 32; c++) start_bits[c] |= ~pcre_cbits[c+cbit_space];          for (c = 0; c < 32; c++)
275              start_bits[c] |= ~cd->cbits[c+cbit_space];
276          break;          break;
277    
278          case OP_WHITESPACE:          case OP_WHITESPACE:
279          for (c = 0; c < 32; c++) start_bits[c] |= pcre_cbits[c+cbit_space];          for (c = 0; c < 32; c++)
280              start_bits[c] |= cd->cbits[c+cbit_space];
281          break;          break;
282    
283          case OP_NOT_WORDCHAR:          case OP_NOT_WORDCHAR:
284          for (c = 0; c < 32; c++)          for (c = 0; c < 32; c++)
285            start_bits[c] |= ~(pcre_cbits[c] | pcre_cbits[c+cbit_word]);            start_bits[c] |= ~cd->cbits[c+cbit_word];
286          break;          break;
287    
288          case OP_WORDCHAR:          case OP_WORDCHAR:
289          for (c = 0; c < 32; c++)          for (c = 0; c < 32; c++)
290            start_bits[c] |= (pcre_cbits[c] | pcre_cbits[c+cbit_word]);            start_bits[c] |= cd->cbits[c+cbit_word];
291          break;          break;
292          }          }
293    
294        tcode += 2;        tcode += 2;
       try_next = TRUE;  
295        break;        break;
296    
297        /* Character class: set the bits and either carry on or not,        /* Character class where all the information is in a bit map: set the
298        according to the repeat count. */        bits and either carry on or not, according to the repeat count. If it was
299          a negative class, and we are operating with UTF-8 characters, any byte
300          with the top-bit set is a potentially valid starter because it may start
301          a character with a value > 255. (This is sub-optimal in that the
302          character may be in the range 128-255, and those characters might be
303          unwanted, but that's as far as we go for the moment.) */
304    
305          case OP_NCLASS:
306          if (utf8) memset(start_bits+16, 0xff, 16);
307          /* Fall through */
308    
309        case OP_CLASS:        case OP_CLASS:
310          {          {
# Line 266  do Line 318  do
318            case OP_CRQUERY:            case OP_CRQUERY:
319            case OP_CRMINQUERY:            case OP_CRMINQUERY:
320            tcode++;            tcode++;
           try_next = TRUE;  
321            break;            break;
322    
323            case OP_CRRANGE:            case OP_CRRANGE:
324            case OP_CRMINRANGE:            case OP_CRMINRANGE:
325            if (((tcode[1] << 8) + tcode[2]) == 0)            if (((tcode[1] << 8) + tcode[2]) == 0) tcode += 5;
326              {              else try_next = FALSE;
327              tcode += 5;            break;
328              try_next = TRUE;  
329              }            default:
330              try_next = FALSE;
331            break;            break;
332            }            }
333          }          }
334        break; /* End of class handling */        break; /* End of bitmap class handling */
335    
336        }      /* End of switch */        }      /* End of switch */
337      }        /* End of try_next loop */      }        /* End of try_next loop */
338    
339    code += (code[1] << 8) + code[2];   /* Advance to next branch */    code += GET(code, 1);   /* Advance to next branch */
340    }    }
341  while (*code == OP_ALT);  while (*code == OP_ALT);
342  return TRUE;  return TRUE;
# Line 306  Arguments: Line 358  Arguments:
358    errorptr  points to where to place error messages;    errorptr  points to where to place error messages;
359              set NULL unless error              set NULL unless error
360    
361  Returns:    pointer to a pcre_extra block,  Returns:    pointer to a pcre_extra block, with study_data filled in and the
362                  appropriate flag set;
363              NULL on error or if no optimization possible              NULL on error or if no optimization possible
364  */  */
365    
# Line 314  pcre_extra * Line 367  pcre_extra *
367  pcre_study(const pcre *external_re, int options, const char **errorptr)  pcre_study(const pcre *external_re, int options, const char **errorptr)
368  {  {
369  uschar start_bits[32];  uschar start_bits[32];
370  real_pcre_extra *extra;  pcre_extra *extra;
371    pcre_study_data *study;
372  const real_pcre *re = (const real_pcre *)external_re;  const real_pcre *re = (const real_pcre *)external_re;
373    uschar *code = (uschar *)re + sizeof(real_pcre) +
374      (re->name_count * re->name_entry_size);
375    compile_data compile_block;
376    
377  *errorptr = NULL;  *errorptr = NULL;
378    
# Line 331  if ((options & ~PUBLIC_STUDY_OPTIONS) != Line 388  if ((options & ~PUBLIC_STUDY_OPTIONS) !=
388    return NULL;    return NULL;
389    }    }
390    
391  /* For an anchored pattern, or an unchored pattern that has a first char, or a  /* For an anchored pattern, or an unanchored pattern that has a first char, or
392  multiline pattern that matches only at "line starts", no further processing at  a multiline pattern that matches only at "line starts", no further processing
393  present. */  at present. */
394    
395  if ((re->options & (PCRE_ANCHORED|PCRE_FIRSTSET|PCRE_STARTLINE)) != 0)  if ((re->options & (PCRE_ANCHORED|PCRE_FIRSTSET|PCRE_STARTLINE)) != 0)
396    return NULL;    return NULL;
397    
398    /* Set the character tables in the block which is passed around */
399    
400    compile_block.lcc = re->tables + lcc_offset;
401    compile_block.fcc = re->tables + fcc_offset;
402    compile_block.cbits = re->tables + cbits_offset;
403    compile_block.ctypes = re->tables + ctypes_offset;
404    
405  /* See if we can find a fixed set of initial characters for the pattern. */  /* See if we can find a fixed set of initial characters for the pattern. */
406    
407  memset(start_bits, 0, 32 * sizeof(uschar));  memset(start_bits, 0, 32 * sizeof(uschar));
408  if (!set_start_bits(re->code, start_bits, (re->options & PCRE_CASELESS) != 0))  if (!set_start_bits(code, start_bits, (re->options & PCRE_CASELESS) != 0,
409    return NULL;    (re->options & PCRE_UTF8) != 0, &compile_block)) return NULL;
410    
411  /* Get an "extra" block and put the information therein. */  /* Get a pcre_extra block and a pcre_study_data block. The study data is put in
412    the latter, which is pointed to by the former, which may also get additional
413    data set later by the calling program. At the moment, the size of
414    pcre_study_data is fixed. We nevertheless save it in a field for returning via
415    the pcre_fullinfo() function so that if it becomes variable in the future, we
416    don't have to change that code. */
417    
418  extra = (real_pcre_extra *)(pcre_malloc)(sizeof(real_pcre_extra));  extra = (pcre_extra *)(pcre_malloc)
419      (sizeof(pcre_extra) + sizeof(pcre_study_data));
420    
421  if (extra == NULL)  if (extra == NULL)
422    {    {
# Line 354  if (extra == NULL) Line 424  if (extra == NULL)
424    return NULL;    return NULL;
425    }    }
426    
427  extra->options = PCRE_STUDY_MAPPED;  study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));
428  memcpy(extra->start_bits, start_bits, sizeof(start_bits));  extra->flags = PCRE_EXTRA_STUDY_DATA;
429    extra->study_data = study;
430    
431    study->size = sizeof(pcre_study_data);
432    study->options = PCRE_STUDY_MAPPED;
433    memcpy(study->start_bits, start_bits, sizeof(start_bits));
434    
435  return (pcre_extra *)extra;  return extra;
436  }  }
437    
438  /* End of study.c */  /* End of study.c */

Legend:
Removed from v.23  
changed lines
  Added in v.64

  ViewVC Help
Powered by ViewVC 1.1.5