1 /* $Cambridge: exim/src/src/pcre/pcre_study.c,v 1.2 2005/08/08 10:22:14 ph10 Exp $ */
3 /*************************************************
4 * Perl-Compatible Regular Expressions *
5 *************************************************/
7 /* PCRE is a library of functions to support regular expressions whose syntax
8 and semantics are as close as possible to those of the Perl 5 language.
10 Written by Philip Hazel
11 Copyright (c) 1997-2005 University of Cambridge
13 -----------------------------------------------------------------------------
14 Redistribution and use in source and binary forms, with or without
15 modification, are permitted provided that the following conditions are met:
17 * Redistributions of source code must retain the above copyright notice,
18 this list of conditions and the following disclaimer.
20 * Redistributions in binary form must reproduce the above copyright
21 notice, this list of conditions and the following disclaimer in the
22 documentation and/or other materials provided with the distribution.
24 * Neither the name of the University of Cambridge nor the names of its
25 contributors may be used to endorse or promote products derived from
26 this software without specific prior written permission.
28 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
29 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
32 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
33 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
34 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
35 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
36 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
37 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
38 POSSIBILITY OF SUCH DAMAGE.
39 -----------------------------------------------------------------------------
43 /* This module contains the external function pcre_study(), along with local
44 supporting functions. */
47 #include "pcre_internal.h"
50 /*************************************************
51 * Set a bit and maybe its alternate case *
52 *************************************************/
54 /* Given a character, set its bit in the table, and also the bit for the other
55 version of a letter if we are caseless.
58 start_bits points to the bit map
60 caseless the caseless flag
61 cd the block with char table pointers
67 set_bit(uschar *start_bits, unsigned int c, BOOL caseless, compile_data *cd)
69 start_bits[c/8] |= (1 << (c&7));
70 if (caseless && (cd->ctypes[c] & ctype_letter) != 0)
71 start_bits[cd->fcc[c]/8] |= (1 << (cd->fcc[c]&7));
76 /*************************************************
77 * Create bitmap of starting chars *
78 *************************************************/
80 /* This function scans a compiled unanchored expression and attempts to build a
81 bitmap of the set of initial characters. If it can't, it returns FALSE. As time
82 goes by, we may be able to get more clever at doing this.
85 code points to an expression
86 start_bits points to a 32-byte table, initialized to 0
87 caseless the current state of the caseless flag
88 utf8 TRUE if in UTF-8 mode
89 cd the block with char table pointers
91 Returns: TRUE if table built, FALSE otherwise
95 set_start_bits(const uschar *code, uschar *start_bits, BOOL caseless,
96 BOOL utf8, compile_data *cd)
100 /* This next statement and the later reference to dummy are here in order to
101 trick the optimizer of the IBM C compiler for OS/2 into generating correct
102 code. Apparently IBM isn't going to fix the problem, and we would rather not
103 disable optimization (in this module it actually makes a big difference, and
104 the pcre module can use all the optimization it can get). */
110 const uschar *tcode = code + 1 + LINK_SIZE;
111 BOOL try_next = TRUE;
115 /* If a branch starts with a bracket or a positive lookahead assertion,
116 recurse to set bits from within them. That's all for this branch. */
118 if ((int)*tcode >= OP_BRA || *tcode == OP_ASSERT)
120 if (!set_start_bits(tcode, start_bits, caseless, utf8, cd))
130 /* Skip over callout */
133 tcode += 2 + 2*LINK_SIZE;
136 /* Skip over extended extraction bracket number */
142 /* Skip over lookbehind and negative lookahead assertions */
146 case OP_ASSERTBACK_NOT:
147 do tcode += GET(tcode, 1); while (*tcode == OP_ALT);
148 tcode += 1+LINK_SIZE;
151 /* Skip over an option setting, changing the caseless flag */
154 caseless = (tcode[1] & PCRE_CASELESS) != 0;
158 /* BRAZERO does the bracket, but carries on. */
162 if (!set_start_bits(++tcode, start_bits, caseless, utf8, cd))
165 do tcode += GET(tcode,1); while (*tcode == OP_ALT);
166 tcode += 1+LINK_SIZE;
169 /* Single-char * or ? sets the bit and tries the next item */
175 set_bit(start_bits, tcode[1], caseless, cd);
178 if (utf8) while ((*tcode & 0xc0) == 0x80) tcode++;
182 /* Single-char upto sets the bit and tries the next */
186 set_bit(start_bits, tcode[3], caseless, cd);
189 if (utf8) while ((*tcode & 0xc0) == 0x80) tcode++;
193 /* At least one single char sets the bit and stops */
195 case OP_EXACT: /* Fall through */
202 set_bit(start_bits, tcode[1], caseless, cd);
206 /* Single character type sets the bits and stops */
209 for (c = 0; c < 32; c++)
210 start_bits[c] |= ~cd->cbits[c+cbit_digit];
215 for (c = 0; c < 32; c++)
216 start_bits[c] |= cd->cbits[c+cbit_digit];
220 case OP_NOT_WHITESPACE:
221 for (c = 0; c < 32; c++)
222 start_bits[c] |= ~cd->cbits[c+cbit_space];
227 for (c = 0; c < 32; c++)
228 start_bits[c] |= cd->cbits[c+cbit_space];
232 case OP_NOT_WORDCHAR:
233 for (c = 0; c < 32; c++)
234 start_bits[c] |= ~cd->cbits[c+cbit_word];
239 for (c = 0; c < 32; c++)
240 start_bits[c] |= cd->cbits[c+cbit_word];
244 /* One or more character type fudges the pointer and restarts, knowing
245 it will hit a single character type and stop there. */
256 /* Zero or more repeats of character types set the bits and then
261 tcode += 2; /* Fall through */
266 case OP_TYPEMINQUERY:
273 for (c = 0; c < 32; c++)
274 start_bits[c] |= ~cd->cbits[c+cbit_digit];
278 for (c = 0; c < 32; c++)
279 start_bits[c] |= cd->cbits[c+cbit_digit];
282 case OP_NOT_WHITESPACE:
283 for (c = 0; c < 32; c++)
284 start_bits[c] |= ~cd->cbits[c+cbit_space];
288 for (c = 0; c < 32; c++)
289 start_bits[c] |= cd->cbits[c+cbit_space];
292 case OP_NOT_WORDCHAR:
293 for (c = 0; c < 32; c++)
294 start_bits[c] |= ~cd->cbits[c+cbit_word];
298 for (c = 0; c < 32; c++)
299 start_bits[c] |= cd->cbits[c+cbit_word];
306 /* Character class where all the information is in a bit map: set the
307 bits and either carry on or not, according to the repeat count. If it was
308 a negative class, and we are operating with UTF-8 characters, any byte
309 with a value >= 0xc4 is a potentially valid starter because it starts a
310 character with a value > 255. */
315 start_bits[24] |= 0xf0; /* Bits for 0xc4 - 0xc8 */
316 memset(start_bits+25, 0xff, 7); /* Bits for 0xc9 - 0xff */
324 /* In UTF-8 mode, the bits in a bit map correspond to character
325 values, not to byte values. However, the bit map we are constructing is
326 for byte values. So we have to do a conversion for characters whose
327 value is > 127. In fact, there are only two possible starting bytes for
328 characters in the range 128 - 255. */
332 for (c = 0; c < 16; c++) start_bits[c] |= tcode[c];
333 for (c = 128; c < 256; c++)
335 if ((tcode[c/8] && (1 << (c&7))) != 0)
337 int d = (c >> 6) | 0xc0; /* Set bit for this starter */
338 start_bits[d/8] |= (1 << (d&7)); /* and then skip on to the */
339 c = (c & 0xc0) + 0x40 - 1; /* next relevant character. */
344 /* In non-UTF-8 mode, the two bit maps are completely compatible. */
348 for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];
351 /* Advance past the bit map, and act on what follows */
365 if (((tcode[1] << 8) + tcode[2]) == 0) tcode += 5;
366 else try_next = FALSE;
374 break; /* End of bitmap class handling */
376 } /* End of switch */
377 } /* End of try_next loop */
379 code += GET(code, 1); /* Advance to next branch */
381 while (*code == OP_ALT);
387 /*************************************************
388 * Study a compiled expression *
389 *************************************************/
391 /* This function is handed a compiled expression that it must study to produce
392 information that will speed up the matching. It returns a pcre_extra block
393 which then gets handed back to pcre_exec().
396 re points to the compiled expression
397 options contains option bits
398 errorptr points to where to place error messages;
399 set NULL unless error
401 Returns: pointer to a pcre_extra block, with study_data filled in and the
402 appropriate flag set;
403 NULL on error or if no optimization possible
407 pcre_study(const pcre *external_re, int options, const char **errorptr)
409 uschar start_bits[32];
411 pcre_study_data *study;
412 const uschar *tables;
413 const real_pcre *re = (const real_pcre *)external_re;
414 uschar *code = (uschar *)re + re->name_table_offset +
415 (re->name_count * re->name_entry_size);
416 compile_data compile_block;
420 if (re == NULL || re->magic_number != MAGIC_NUMBER)
422 *errorptr = "argument is not a compiled regular expression";
426 if ((options & ~PUBLIC_STUDY_OPTIONS) != 0)
428 *errorptr = "unknown or incorrect option bit(s) set";
432 /* For an anchored pattern, or an unanchored pattern that has a first char, or
433 a multiline pattern that matches only at "line starts", no further processing
436 if ((re->options & (PCRE_ANCHORED|PCRE_FIRSTSET|PCRE_STARTLINE)) != 0)
439 /* Set the character tables in the block that is passed around */
443 (void)pcre_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
446 compile_block.lcc = tables + lcc_offset;
447 compile_block.fcc = tables + fcc_offset;
448 compile_block.cbits = tables + cbits_offset;
449 compile_block.ctypes = tables + ctypes_offset;
451 /* See if we can find a fixed set of initial characters for the pattern. */
453 memset(start_bits, 0, 32 * sizeof(uschar));
454 if (!set_start_bits(code, start_bits, (re->options & PCRE_CASELESS) != 0,
455 (re->options & PCRE_UTF8) != 0, &compile_block)) return NULL;
457 /* Get a pcre_extra block and a pcre_study_data block. The study data is put in
458 the latter, which is pointed to by the former, which may also get additional
459 data set later by the calling program. At the moment, the size of
460 pcre_study_data is fixed. We nevertheless save it in a field for returning via
461 the pcre_fullinfo() function so that if it becomes variable in the future, we
462 don't have to change that code. */
464 extra = (pcre_extra *)(pcre_malloc)
465 (sizeof(pcre_extra) + sizeof(pcre_study_data));
469 *errorptr = "failed to get memory";
473 study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));
474 extra->flags = PCRE_EXTRA_STUDY_DATA;
475 extra->study_data = study;
477 study->size = sizeof(pcre_study_data);
478 study->options = PCRE_STUDY_MAPPED;
479 memcpy(study->start_bits, start_bits, sizeof(start_bits));
484 /* End of pcre_study.c */