1 /* $Cambridge: exim/src/src/pcre/pcre_study.c,v 1.3 2006/11/07 16:50:36 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) 1997-2006 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)
101 /* ========================================================================= */
102 /* The following comment and code was inserted in January 1999. In May 2006,
103 when it was observed to cause compiler warnings about unused values, I took it
104 out again. If anybody is still using OS/2, they will have to put it back
107 /* This next statement and the later reference to dummy are here in order to
108 trick the optimizer of the IBM C compiler for OS/2 into generating correct
109 code. Apparently IBM isn't going to fix the problem, and we would rather not
110 disable optimization (in this module it actually makes a big difference, and
111 the pcre module can use all the optimization it can get). */
114 /* ========================================================================= */
119 const uschar *tcode = code + 1 + LINK_SIZE;
120 BOOL try_next = TRUE;
124 /* If a branch starts with a bracket or a positive lookahead assertion,
125 recurse to set bits from within them. That's all for this branch. */
127 if ((int)*tcode >= OP_BRA || *tcode == OP_ASSERT)
129 if (!set_start_bits(tcode, start_bits, caseless, utf8, cd))
139 /* Skip over callout */
142 tcode += 2 + 2*LINK_SIZE;
145 /* Skip over extended extraction bracket number */
151 /* Skip over lookbehind and negative lookahead assertions */
155 case OP_ASSERTBACK_NOT:
156 do tcode += GET(tcode, 1); while (*tcode == OP_ALT);
157 tcode += 1+LINK_SIZE;
160 /* Skip over an option setting, changing the caseless flag */
163 caseless = (tcode[1] & PCRE_CASELESS) != 0;
167 /* BRAZERO does the bracket, but carries on. */
171 if (!set_start_bits(++tcode, start_bits, caseless, utf8, cd))
173 /* =========================================================================
174 See the comment at the head of this function concerning the next line,
175 which was an old fudge for the benefit of OS/2.
177 ========================================================================= */
178 do tcode += GET(tcode,1); while (*tcode == OP_ALT);
179 tcode += 1+LINK_SIZE;
182 /* Single-char * or ? sets the bit and tries the next item */
188 set_bit(start_bits, tcode[1], caseless, cd);
191 if (utf8) while ((*tcode & 0xc0) == 0x80) tcode++;
195 /* Single-char upto sets the bit and tries the next */
199 set_bit(start_bits, tcode[3], caseless, cd);
202 if (utf8) while ((*tcode & 0xc0) == 0x80) tcode++;
206 /* At least one single char sets the bit and stops */
208 case OP_EXACT: /* Fall through */
215 set_bit(start_bits, tcode[1], caseless, cd);
219 /* Single character type sets the bits and stops */
222 for (c = 0; c < 32; c++)
223 start_bits[c] |= ~cd->cbits[c+cbit_digit];
228 for (c = 0; c < 32; c++)
229 start_bits[c] |= cd->cbits[c+cbit_digit];
233 /* The cbit_space table has vertical tab as whitespace; we have to
236 case OP_NOT_WHITESPACE:
237 for (c = 0; c < 32; c++)
239 int d = cd->cbits[c+cbit_space];
240 if (c == 1) d &= ~0x08;
246 /* The cbit_space table has vertical tab as whitespace; we have to
250 for (c = 0; c < 32; c++)
252 int d = cd->cbits[c+cbit_space];
253 if (c == 1) d &= ~0x08;
259 case OP_NOT_WORDCHAR:
260 for (c = 0; c < 32; c++)
261 start_bits[c] |= ~cd->cbits[c+cbit_word];
266 for (c = 0; c < 32; c++)
267 start_bits[c] |= cd->cbits[c+cbit_word];
271 /* One or more character type fudges the pointer and restarts, knowing
272 it will hit a single character type and stop there. */
283 /* Zero or more repeats of character types set the bits and then
288 tcode += 2; /* Fall through */
293 case OP_TYPEMINQUERY:
300 for (c = 0; c < 32; c++)
301 start_bits[c] |= ~cd->cbits[c+cbit_digit];
305 for (c = 0; c < 32; c++)
306 start_bits[c] |= cd->cbits[c+cbit_digit];
309 /* The cbit_space table has vertical tab as whitespace; we have to
312 case OP_NOT_WHITESPACE:
313 for (c = 0; c < 32; c++)
315 int d = cd->cbits[c+cbit_space];
316 if (c == 1) d &= ~0x08;
321 /* The cbit_space table has vertical tab as whitespace; we have to
325 for (c = 0; c < 32; c++)
327 int d = cd->cbits[c+cbit_space];
328 if (c == 1) d &= ~0x08;
333 case OP_NOT_WORDCHAR:
334 for (c = 0; c < 32; c++)
335 start_bits[c] |= ~cd->cbits[c+cbit_word];
339 for (c = 0; c < 32; c++)
340 start_bits[c] |= cd->cbits[c+cbit_word];
347 /* Character class where all the information is in a bit map: set the
348 bits and either carry on or not, according to the repeat count. If it was
349 a negative class, and we are operating with UTF-8 characters, any byte
350 with a value >= 0xc4 is a potentially valid starter because it starts a
351 character with a value > 255. */
356 start_bits[24] |= 0xf0; /* Bits for 0xc4 - 0xc8 */
357 memset(start_bits+25, 0xff, 7); /* Bits for 0xc9 - 0xff */
365 /* In UTF-8 mode, the bits in a bit map correspond to character
366 values, not to byte values. However, the bit map we are constructing is
367 for byte values. So we have to do a conversion for characters whose
368 value is > 127. In fact, there are only two possible starting bytes for
369 characters in the range 128 - 255. */
373 for (c = 0; c < 16; c++) start_bits[c] |= tcode[c];
374 for (c = 128; c < 256; c++)
376 if ((tcode[c/8] && (1 << (c&7))) != 0)
378 int d = (c >> 6) | 0xc0; /* Set bit for this starter */
379 start_bits[d/8] |= (1 << (d&7)); /* and then skip on to the */
380 c = (c & 0xc0) + 0x40 - 1; /* next relevant character. */
385 /* In non-UTF-8 mode, the two bit maps are completely compatible. */
389 for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];
392 /* Advance past the bit map, and act on what follows */
406 if (((tcode[1] << 8) + tcode[2]) == 0) tcode += 5;
407 else try_next = FALSE;
415 break; /* End of bitmap class handling */
417 } /* End of switch */
418 } /* End of try_next loop */
420 code += GET(code, 1); /* Advance to next branch */
422 while (*code == OP_ALT);
428 /*************************************************
429 * Study a compiled expression *
430 *************************************************/
432 /* This function is handed a compiled expression that it must study to produce
433 information that will speed up the matching. It returns a pcre_extra block
434 which then gets handed back to pcre_exec().
437 re points to the compiled expression
438 options contains option bits
439 errorptr points to where to place error messages;
440 set NULL unless error
442 Returns: pointer to a pcre_extra block, with study_data filled in and the
443 appropriate flag set;
444 NULL on error or if no optimization possible
447 PCRE_DATA_SCOPE pcre_extra *
448 pcre_study(const pcre *external_re, int options, const char **errorptr)
450 uschar start_bits[32];
452 pcre_study_data *study;
453 const uschar *tables;
455 compile_data compile_block;
456 const real_pcre *re = (const real_pcre *)external_re;
460 if (re == NULL || re->magic_number != MAGIC_NUMBER)
462 *errorptr = "argument is not a compiled regular expression";
466 if ((options & ~PUBLIC_STUDY_OPTIONS) != 0)
468 *errorptr = "unknown or incorrect option bit(s) set";
472 code = (uschar *)re + re->name_table_offset +
473 (re->name_count * re->name_entry_size);
475 /* For an anchored pattern, or an unanchored pattern that has a first char, or
476 a multiline pattern that matches only at "line starts", no further processing
479 if ((re->options & (PCRE_ANCHORED|PCRE_FIRSTSET|PCRE_STARTLINE)) != 0)
482 /* Set the character tables in the block that is passed around */
486 (void)pcre_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
489 compile_block.lcc = tables + lcc_offset;
490 compile_block.fcc = tables + fcc_offset;
491 compile_block.cbits = tables + cbits_offset;
492 compile_block.ctypes = tables + ctypes_offset;
494 /* See if we can find a fixed set of initial characters for the pattern. */
496 memset(start_bits, 0, 32 * sizeof(uschar));
497 if (!set_start_bits(code, start_bits, (re->options & PCRE_CASELESS) != 0,
498 (re->options & PCRE_UTF8) != 0, &compile_block)) return NULL;
500 /* Get a pcre_extra block and a pcre_study_data block. The study data is put in
501 the latter, which is pointed to by the former, which may also get additional
502 data set later by the calling program. At the moment, the size of
503 pcre_study_data is fixed. We nevertheless save it in a field for returning via
504 the pcre_fullinfo() function so that if it becomes variable in the future, we
505 don't have to change that code. */
507 extra = (pcre_extra *)(pcre_malloc)
508 (sizeof(pcre_extra) + sizeof(pcre_study_data));
512 *errorptr = "failed to get memory";
516 study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));
517 extra->flags = PCRE_EXTRA_STUDY_DATA;
518 extra->study_data = study;
520 study->size = sizeof(pcre_study_data);
521 study->options = PCRE_STUDY_MAPPED;
522 memcpy(study->start_bits, start_bits, sizeof(start_bits));
527 /* End of pcre_study.c */