Tristan Matthews | 0461646 | 2013-11-14 16:09:34 -0500 | [diff] [blame] | 1 | // Copyright (c) 2010, Google Inc. |
| 2 | // All rights reserved. |
| 3 | // |
| 4 | // Redistribution and use in source and binary forms, with or without |
| 5 | // modification, are permitted provided that the following conditions are |
| 6 | // met: |
| 7 | // |
| 8 | // * Redistributions of source code must retain the above copyright |
| 9 | // notice, this list of conditions and the following disclaimer. |
| 10 | // * Redistributions in binary form must reproduce the above |
| 11 | // copyright notice, this list of conditions and the following disclaimer |
| 12 | // in the documentation and/or other materials provided with the |
| 13 | // distribution. |
| 14 | // * Neither the name of Google Inc. nor the names of its |
| 15 | // contributors may be used to endorse or promote products derived from |
| 16 | // this software without specific prior written permission. |
| 17 | // |
| 18 | // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
| 19 | // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
| 20 | // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR |
| 21 | // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT |
| 22 | // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, |
| 23 | // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT |
| 24 | // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
| 25 | // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
| 26 | // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
| 27 | // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
| 28 | // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 29 | // |
| 30 | // Author: Sanjay Ghemawat |
| 31 | |
| 32 | #ifdef HAVE_CONFIG_H |
| 33 | #include "config.h" |
| 34 | #endif |
| 35 | |
| 36 | #include <stdlib.h> |
| 37 | #include <stdio.h> |
| 38 | #include <ctype.h> |
| 39 | #include <limits.h> /* for SHRT_MIN, USHRT_MAX, etc */ |
| 40 | #include <string.h> /* for memcpy */ |
| 41 | #include <assert.h> |
| 42 | #include <errno.h> |
| 43 | #include <string> |
| 44 | #include <algorithm> |
| 45 | |
| 46 | #include "pcrecpp_internal.h" |
| 47 | #include "pcre.h" |
| 48 | #include "pcrecpp.h" |
| 49 | #include "pcre_stringpiece.h" |
| 50 | |
| 51 | |
| 52 | namespace pcrecpp { |
| 53 | |
| 54 | // Maximum number of args we can set |
| 55 | static const int kMaxArgs = 16; |
| 56 | static const int kVecSize = (1 + kMaxArgs) * 3; // results + PCRE workspace |
| 57 | |
| 58 | // Special object that stands-in for no argument |
| 59 | Arg RE::no_arg((void*)NULL); |
| 60 | |
| 61 | // This is for ABI compatibility with old versions of pcre (pre-7.6), |
| 62 | // which defined a global no_arg variable instead of putting it in the |
| 63 | // RE class. This works on GCC >= 3, at least. It definitely works |
| 64 | // for ELF, but may not for other object formats (Mach-O, for |
| 65 | // instance, does not support aliases.) We could probably have a more |
| 66 | // inclusive test if we ever needed it. (Note that not only the |
| 67 | // __attribute__ syntax, but also __USER_LABEL_PREFIX__, are |
| 68 | // gnu-specific.) |
| 69 | #if defined(__GNUC__) && __GNUC__ >= 3 && defined(__ELF__) |
| 70 | # define ULP_AS_STRING(x) ULP_AS_STRING_INTERNAL(x) |
| 71 | # define ULP_AS_STRING_INTERNAL(x) #x |
| 72 | # define USER_LABEL_PREFIX_STR ULP_AS_STRING(__USER_LABEL_PREFIX__) |
| 73 | extern Arg no_arg |
| 74 | __attribute__((alias(USER_LABEL_PREFIX_STR "_ZN7pcrecpp2RE6no_argE"))); |
| 75 | #endif |
| 76 | |
| 77 | // If a regular expression has no error, its error_ field points here |
| 78 | static const string empty_string; |
| 79 | |
| 80 | // If the user doesn't ask for any options, we just use this one |
| 81 | static RE_Options default_options; |
| 82 | |
| 83 | void RE::Init(const string& pat, const RE_Options* options) { |
| 84 | pattern_ = pat; |
| 85 | if (options == NULL) { |
| 86 | options_ = default_options; |
| 87 | } else { |
| 88 | options_ = *options; |
| 89 | } |
| 90 | error_ = &empty_string; |
| 91 | re_full_ = NULL; |
| 92 | re_partial_ = NULL; |
| 93 | |
| 94 | re_partial_ = Compile(UNANCHORED); |
| 95 | if (re_partial_ != NULL) { |
| 96 | re_full_ = Compile(ANCHOR_BOTH); |
| 97 | } |
| 98 | } |
| 99 | |
| 100 | void RE::Cleanup() { |
| 101 | if (re_full_ != NULL) (*pcre_free)(re_full_); |
| 102 | if (re_partial_ != NULL) (*pcre_free)(re_partial_); |
| 103 | if (error_ != &empty_string) delete error_; |
| 104 | } |
| 105 | |
| 106 | |
| 107 | RE::~RE() { |
| 108 | Cleanup(); |
| 109 | } |
| 110 | |
| 111 | |
| 112 | pcre* RE::Compile(Anchor anchor) { |
| 113 | // First, convert RE_Options into pcre options |
| 114 | int pcre_options = 0; |
| 115 | pcre_options = options_.all_options(); |
| 116 | |
| 117 | // Special treatment for anchoring. This is needed because at |
| 118 | // runtime pcre only provides an option for anchoring at the |
| 119 | // beginning of a string (unless you use offset). |
| 120 | // |
| 121 | // There are three types of anchoring we want: |
| 122 | // UNANCHORED Compile the original pattern, and use |
| 123 | // a pcre unanchored match. |
| 124 | // ANCHOR_START Compile the original pattern, and use |
| 125 | // a pcre anchored match. |
| 126 | // ANCHOR_BOTH Tack a "\z" to the end of the original pattern |
| 127 | // and use a pcre anchored match. |
| 128 | |
| 129 | const char* compile_error; |
| 130 | int eoffset; |
| 131 | pcre* re; |
| 132 | if (anchor != ANCHOR_BOTH) { |
| 133 | re = pcre_compile(pattern_.c_str(), pcre_options, |
| 134 | &compile_error, &eoffset, NULL); |
| 135 | } else { |
| 136 | // Tack a '\z' at the end of RE. Parenthesize it first so that |
| 137 | // the '\z' applies to all top-level alternatives in the regexp. |
| 138 | string wrapped = "(?:"; // A non-counting grouping operator |
| 139 | wrapped += pattern_; |
| 140 | wrapped += ")\\z"; |
| 141 | re = pcre_compile(wrapped.c_str(), pcre_options, |
| 142 | &compile_error, &eoffset, NULL); |
| 143 | } |
| 144 | if (re == NULL) { |
| 145 | if (error_ == &empty_string) error_ = new string(compile_error); |
| 146 | } |
| 147 | return re; |
| 148 | } |
| 149 | |
| 150 | /***** Matching interfaces *****/ |
| 151 | |
| 152 | bool RE::FullMatch(const StringPiece& text, |
| 153 | const Arg& ptr1, |
| 154 | const Arg& ptr2, |
| 155 | const Arg& ptr3, |
| 156 | const Arg& ptr4, |
| 157 | const Arg& ptr5, |
| 158 | const Arg& ptr6, |
| 159 | const Arg& ptr7, |
| 160 | const Arg& ptr8, |
| 161 | const Arg& ptr9, |
| 162 | const Arg& ptr10, |
| 163 | const Arg& ptr11, |
| 164 | const Arg& ptr12, |
| 165 | const Arg& ptr13, |
| 166 | const Arg& ptr14, |
| 167 | const Arg& ptr15, |
| 168 | const Arg& ptr16) const { |
| 169 | const Arg* args[kMaxArgs]; |
| 170 | int n = 0; |
| 171 | if (&ptr1 == &no_arg) goto done; args[n++] = &ptr1; |
| 172 | if (&ptr2 == &no_arg) goto done; args[n++] = &ptr2; |
| 173 | if (&ptr3 == &no_arg) goto done; args[n++] = &ptr3; |
| 174 | if (&ptr4 == &no_arg) goto done; args[n++] = &ptr4; |
| 175 | if (&ptr5 == &no_arg) goto done; args[n++] = &ptr5; |
| 176 | if (&ptr6 == &no_arg) goto done; args[n++] = &ptr6; |
| 177 | if (&ptr7 == &no_arg) goto done; args[n++] = &ptr7; |
| 178 | if (&ptr8 == &no_arg) goto done; args[n++] = &ptr8; |
| 179 | if (&ptr9 == &no_arg) goto done; args[n++] = &ptr9; |
| 180 | if (&ptr10 == &no_arg) goto done; args[n++] = &ptr10; |
| 181 | if (&ptr11 == &no_arg) goto done; args[n++] = &ptr11; |
| 182 | if (&ptr12 == &no_arg) goto done; args[n++] = &ptr12; |
| 183 | if (&ptr13 == &no_arg) goto done; args[n++] = &ptr13; |
| 184 | if (&ptr14 == &no_arg) goto done; args[n++] = &ptr14; |
| 185 | if (&ptr15 == &no_arg) goto done; args[n++] = &ptr15; |
| 186 | if (&ptr16 == &no_arg) goto done; args[n++] = &ptr16; |
| 187 | done: |
| 188 | |
| 189 | int consumed; |
| 190 | int vec[kVecSize]; |
| 191 | return DoMatchImpl(text, ANCHOR_BOTH, &consumed, args, n, vec, kVecSize); |
| 192 | } |
| 193 | |
| 194 | bool RE::PartialMatch(const StringPiece& text, |
| 195 | const Arg& ptr1, |
| 196 | const Arg& ptr2, |
| 197 | const Arg& ptr3, |
| 198 | const Arg& ptr4, |
| 199 | const Arg& ptr5, |
| 200 | const Arg& ptr6, |
| 201 | const Arg& ptr7, |
| 202 | const Arg& ptr8, |
| 203 | const Arg& ptr9, |
| 204 | const Arg& ptr10, |
| 205 | const Arg& ptr11, |
| 206 | const Arg& ptr12, |
| 207 | const Arg& ptr13, |
| 208 | const Arg& ptr14, |
| 209 | const Arg& ptr15, |
| 210 | const Arg& ptr16) const { |
| 211 | const Arg* args[kMaxArgs]; |
| 212 | int n = 0; |
| 213 | if (&ptr1 == &no_arg) goto done; args[n++] = &ptr1; |
| 214 | if (&ptr2 == &no_arg) goto done; args[n++] = &ptr2; |
| 215 | if (&ptr3 == &no_arg) goto done; args[n++] = &ptr3; |
| 216 | if (&ptr4 == &no_arg) goto done; args[n++] = &ptr4; |
| 217 | if (&ptr5 == &no_arg) goto done; args[n++] = &ptr5; |
| 218 | if (&ptr6 == &no_arg) goto done; args[n++] = &ptr6; |
| 219 | if (&ptr7 == &no_arg) goto done; args[n++] = &ptr7; |
| 220 | if (&ptr8 == &no_arg) goto done; args[n++] = &ptr8; |
| 221 | if (&ptr9 == &no_arg) goto done; args[n++] = &ptr9; |
| 222 | if (&ptr10 == &no_arg) goto done; args[n++] = &ptr10; |
| 223 | if (&ptr11 == &no_arg) goto done; args[n++] = &ptr11; |
| 224 | if (&ptr12 == &no_arg) goto done; args[n++] = &ptr12; |
| 225 | if (&ptr13 == &no_arg) goto done; args[n++] = &ptr13; |
| 226 | if (&ptr14 == &no_arg) goto done; args[n++] = &ptr14; |
| 227 | if (&ptr15 == &no_arg) goto done; args[n++] = &ptr15; |
| 228 | if (&ptr16 == &no_arg) goto done; args[n++] = &ptr16; |
| 229 | done: |
| 230 | |
| 231 | int consumed; |
| 232 | int vec[kVecSize]; |
| 233 | return DoMatchImpl(text, UNANCHORED, &consumed, args, n, vec, kVecSize); |
| 234 | } |
| 235 | |
| 236 | bool RE::Consume(StringPiece* input, |
| 237 | const Arg& ptr1, |
| 238 | const Arg& ptr2, |
| 239 | const Arg& ptr3, |
| 240 | const Arg& ptr4, |
| 241 | const Arg& ptr5, |
| 242 | const Arg& ptr6, |
| 243 | const Arg& ptr7, |
| 244 | const Arg& ptr8, |
| 245 | const Arg& ptr9, |
| 246 | const Arg& ptr10, |
| 247 | const Arg& ptr11, |
| 248 | const Arg& ptr12, |
| 249 | const Arg& ptr13, |
| 250 | const Arg& ptr14, |
| 251 | const Arg& ptr15, |
| 252 | const Arg& ptr16) const { |
| 253 | const Arg* args[kMaxArgs]; |
| 254 | int n = 0; |
| 255 | if (&ptr1 == &no_arg) goto done; args[n++] = &ptr1; |
| 256 | if (&ptr2 == &no_arg) goto done; args[n++] = &ptr2; |
| 257 | if (&ptr3 == &no_arg) goto done; args[n++] = &ptr3; |
| 258 | if (&ptr4 == &no_arg) goto done; args[n++] = &ptr4; |
| 259 | if (&ptr5 == &no_arg) goto done; args[n++] = &ptr5; |
| 260 | if (&ptr6 == &no_arg) goto done; args[n++] = &ptr6; |
| 261 | if (&ptr7 == &no_arg) goto done; args[n++] = &ptr7; |
| 262 | if (&ptr8 == &no_arg) goto done; args[n++] = &ptr8; |
| 263 | if (&ptr9 == &no_arg) goto done; args[n++] = &ptr9; |
| 264 | if (&ptr10 == &no_arg) goto done; args[n++] = &ptr10; |
| 265 | if (&ptr11 == &no_arg) goto done; args[n++] = &ptr11; |
| 266 | if (&ptr12 == &no_arg) goto done; args[n++] = &ptr12; |
| 267 | if (&ptr13 == &no_arg) goto done; args[n++] = &ptr13; |
| 268 | if (&ptr14 == &no_arg) goto done; args[n++] = &ptr14; |
| 269 | if (&ptr15 == &no_arg) goto done; args[n++] = &ptr15; |
| 270 | if (&ptr16 == &no_arg) goto done; args[n++] = &ptr16; |
| 271 | done: |
| 272 | |
| 273 | int consumed; |
| 274 | int vec[kVecSize]; |
| 275 | if (DoMatchImpl(*input, ANCHOR_START, &consumed, |
| 276 | args, n, vec, kVecSize)) { |
| 277 | input->remove_prefix(consumed); |
| 278 | return true; |
| 279 | } else { |
| 280 | return false; |
| 281 | } |
| 282 | } |
| 283 | |
| 284 | bool RE::FindAndConsume(StringPiece* input, |
| 285 | const Arg& ptr1, |
| 286 | const Arg& ptr2, |
| 287 | const Arg& ptr3, |
| 288 | const Arg& ptr4, |
| 289 | const Arg& ptr5, |
| 290 | const Arg& ptr6, |
| 291 | const Arg& ptr7, |
| 292 | const Arg& ptr8, |
| 293 | const Arg& ptr9, |
| 294 | const Arg& ptr10, |
| 295 | const Arg& ptr11, |
| 296 | const Arg& ptr12, |
| 297 | const Arg& ptr13, |
| 298 | const Arg& ptr14, |
| 299 | const Arg& ptr15, |
| 300 | const Arg& ptr16) const { |
| 301 | const Arg* args[kMaxArgs]; |
| 302 | int n = 0; |
| 303 | if (&ptr1 == &no_arg) goto done; args[n++] = &ptr1; |
| 304 | if (&ptr2 == &no_arg) goto done; args[n++] = &ptr2; |
| 305 | if (&ptr3 == &no_arg) goto done; args[n++] = &ptr3; |
| 306 | if (&ptr4 == &no_arg) goto done; args[n++] = &ptr4; |
| 307 | if (&ptr5 == &no_arg) goto done; args[n++] = &ptr5; |
| 308 | if (&ptr6 == &no_arg) goto done; args[n++] = &ptr6; |
| 309 | if (&ptr7 == &no_arg) goto done; args[n++] = &ptr7; |
| 310 | if (&ptr8 == &no_arg) goto done; args[n++] = &ptr8; |
| 311 | if (&ptr9 == &no_arg) goto done; args[n++] = &ptr9; |
| 312 | if (&ptr10 == &no_arg) goto done; args[n++] = &ptr10; |
| 313 | if (&ptr11 == &no_arg) goto done; args[n++] = &ptr11; |
| 314 | if (&ptr12 == &no_arg) goto done; args[n++] = &ptr12; |
| 315 | if (&ptr13 == &no_arg) goto done; args[n++] = &ptr13; |
| 316 | if (&ptr14 == &no_arg) goto done; args[n++] = &ptr14; |
| 317 | if (&ptr15 == &no_arg) goto done; args[n++] = &ptr15; |
| 318 | if (&ptr16 == &no_arg) goto done; args[n++] = &ptr16; |
| 319 | done: |
| 320 | |
| 321 | int consumed; |
| 322 | int vec[kVecSize]; |
| 323 | if (DoMatchImpl(*input, UNANCHORED, &consumed, |
| 324 | args, n, vec, kVecSize)) { |
| 325 | input->remove_prefix(consumed); |
| 326 | return true; |
| 327 | } else { |
| 328 | return false; |
| 329 | } |
| 330 | } |
| 331 | |
| 332 | bool RE::Replace(const StringPiece& rewrite, |
| 333 | string *str) const { |
| 334 | int vec[kVecSize]; |
| 335 | int matches = TryMatch(*str, 0, UNANCHORED, true, vec, kVecSize); |
| 336 | if (matches == 0) |
| 337 | return false; |
| 338 | |
| 339 | string s; |
| 340 | if (!Rewrite(&s, rewrite, *str, vec, matches)) |
| 341 | return false; |
| 342 | |
| 343 | assert(vec[0] >= 0); |
| 344 | assert(vec[1] >= 0); |
| 345 | str->replace(vec[0], vec[1] - vec[0], s); |
| 346 | return true; |
| 347 | } |
| 348 | |
| 349 | // Returns PCRE_NEWLINE_CRLF, PCRE_NEWLINE_CR, or PCRE_NEWLINE_LF. |
| 350 | // Note that PCRE_NEWLINE_CRLF is defined to be P_N_CR | P_N_LF. |
| 351 | // Modified by PH to add PCRE_NEWLINE_ANY and PCRE_NEWLINE_ANYCRLF. |
| 352 | |
| 353 | static int NewlineMode(int pcre_options) { |
| 354 | // TODO: if we can make it threadsafe, cache this var |
| 355 | int newline_mode = 0; |
| 356 | /* if (newline_mode) return newline_mode; */ // do this once it's cached |
| 357 | if (pcre_options & (PCRE_NEWLINE_CRLF|PCRE_NEWLINE_CR|PCRE_NEWLINE_LF| |
| 358 | PCRE_NEWLINE_ANY|PCRE_NEWLINE_ANYCRLF)) { |
| 359 | newline_mode = (pcre_options & |
| 360 | (PCRE_NEWLINE_CRLF|PCRE_NEWLINE_CR|PCRE_NEWLINE_LF| |
| 361 | PCRE_NEWLINE_ANY|PCRE_NEWLINE_ANYCRLF)); |
| 362 | } else { |
| 363 | int newline; |
| 364 | pcre_config(PCRE_CONFIG_NEWLINE, &newline); |
| 365 | if (newline == 10) |
| 366 | newline_mode = PCRE_NEWLINE_LF; |
| 367 | else if (newline == 13) |
| 368 | newline_mode = PCRE_NEWLINE_CR; |
| 369 | else if (newline == 3338) |
| 370 | newline_mode = PCRE_NEWLINE_CRLF; |
| 371 | else if (newline == -1) |
| 372 | newline_mode = PCRE_NEWLINE_ANY; |
| 373 | else if (newline == -2) |
| 374 | newline_mode = PCRE_NEWLINE_ANYCRLF; |
| 375 | else |
| 376 | assert(NULL == "Unexpected return value from pcre_config(NEWLINE)"); |
| 377 | } |
| 378 | return newline_mode; |
| 379 | } |
| 380 | |
| 381 | int RE::GlobalReplace(const StringPiece& rewrite, |
| 382 | string *str) const { |
| 383 | int count = 0; |
| 384 | int vec[kVecSize]; |
| 385 | string out; |
| 386 | int start = 0; |
| 387 | int lastend = -1; |
| 388 | bool last_match_was_empty_string = false; |
| 389 | |
| 390 | while (start <= static_cast<int>(str->length())) { |
| 391 | // If the previous match was for the empty string, we shouldn't |
| 392 | // just match again: we'll match in the same way and get an |
| 393 | // infinite loop. Instead, we do the match in a special way: |
| 394 | // anchored -- to force another try at the same position -- |
| 395 | // and with a flag saying that this time, ignore empty matches. |
| 396 | // If this special match returns, that means there's a non-empty |
| 397 | // match at this position as well, and we can continue. If not, |
| 398 | // we do what perl does, and just advance by one. |
| 399 | // Notice that perl prints '@@@' for this; |
| 400 | // perl -le '$_ = "aa"; s/b*|aa/@/g; print' |
| 401 | int matches; |
| 402 | if (last_match_was_empty_string) { |
| 403 | matches = TryMatch(*str, start, ANCHOR_START, false, vec, kVecSize); |
| 404 | if (matches <= 0) { |
| 405 | int matchend = start + 1; // advance one character. |
| 406 | // If the current char is CR and we're in CRLF mode, skip LF too. |
| 407 | // Note it's better to call pcre_fullinfo() than to examine |
| 408 | // all_options(), since options_ could have changed bewteen |
| 409 | // compile-time and now, but this is simpler and safe enough. |
| 410 | // Modified by PH to add ANY and ANYCRLF. |
| 411 | if (matchend < static_cast<int>(str->length()) && |
| 412 | (*str)[start] == '\r' && (*str)[matchend] == '\n' && |
| 413 | (NewlineMode(options_.all_options()) == PCRE_NEWLINE_CRLF || |
| 414 | NewlineMode(options_.all_options()) == PCRE_NEWLINE_ANY || |
| 415 | NewlineMode(options_.all_options()) == PCRE_NEWLINE_ANYCRLF)) { |
| 416 | matchend++; |
| 417 | } |
| 418 | // We also need to advance more than one char if we're in utf8 mode. |
| 419 | #ifdef SUPPORT_UTF8 |
| 420 | if (options_.utf8()) { |
| 421 | while (matchend < static_cast<int>(str->length()) && |
| 422 | ((*str)[matchend] & 0xc0) == 0x80) |
| 423 | matchend++; |
| 424 | } |
| 425 | #endif |
| 426 | if (start < static_cast<int>(str->length())) |
| 427 | out.append(*str, start, matchend - start); |
| 428 | start = matchend; |
| 429 | last_match_was_empty_string = false; |
| 430 | continue; |
| 431 | } |
| 432 | } else { |
| 433 | matches = TryMatch(*str, start, UNANCHORED, true, vec, kVecSize); |
| 434 | if (matches <= 0) |
| 435 | break; |
| 436 | } |
| 437 | int matchstart = vec[0], matchend = vec[1]; |
| 438 | assert(matchstart >= start); |
| 439 | assert(matchend >= matchstart); |
| 440 | out.append(*str, start, matchstart - start); |
| 441 | Rewrite(&out, rewrite, *str, vec, matches); |
| 442 | start = matchend; |
| 443 | lastend = matchend; |
| 444 | count++; |
| 445 | last_match_was_empty_string = (matchstart == matchend); |
| 446 | } |
| 447 | |
| 448 | if (count == 0) |
| 449 | return 0; |
| 450 | |
| 451 | if (start < static_cast<int>(str->length())) |
| 452 | out.append(*str, start, str->length() - start); |
| 453 | swap(out, *str); |
| 454 | return count; |
| 455 | } |
| 456 | |
| 457 | bool RE::Extract(const StringPiece& rewrite, |
| 458 | const StringPiece& text, |
| 459 | string *out) const { |
| 460 | int vec[kVecSize]; |
| 461 | int matches = TryMatch(text, 0, UNANCHORED, true, vec, kVecSize); |
| 462 | if (matches == 0) |
| 463 | return false; |
| 464 | out->erase(); |
| 465 | return Rewrite(out, rewrite, text, vec, matches); |
| 466 | } |
| 467 | |
| 468 | /*static*/ string RE::QuoteMeta(const StringPiece& unquoted) { |
| 469 | string result; |
| 470 | |
| 471 | // Escape any ascii character not in [A-Za-z_0-9]. |
| 472 | // |
| 473 | // Note that it's legal to escape a character even if it has no |
| 474 | // special meaning in a regular expression -- so this function does |
| 475 | // that. (This also makes it identical to the perl function of the |
| 476 | // same name; see `perldoc -f quotemeta`.) The one exception is |
| 477 | // escaping NUL: rather than doing backslash + NUL, like perl does, |
| 478 | // we do '\0', because pcre itself doesn't take embedded NUL chars. |
| 479 | for (int ii = 0; ii < unquoted.size(); ++ii) { |
| 480 | // Note that using 'isalnum' here raises the benchmark time from |
| 481 | // 32ns to 58ns: |
| 482 | if (unquoted[ii] == '\0') { |
| 483 | result += "\\0"; |
| 484 | } else if ((unquoted[ii] < 'a' || unquoted[ii] > 'z') && |
| 485 | (unquoted[ii] < 'A' || unquoted[ii] > 'Z') && |
| 486 | (unquoted[ii] < '0' || unquoted[ii] > '9') && |
| 487 | unquoted[ii] != '_' && |
| 488 | // If this is the part of a UTF8 or Latin1 character, we need |
| 489 | // to copy this byte without escaping. Experimentally this is |
| 490 | // what works correctly with the regexp library. |
| 491 | !(unquoted[ii] & 128)) { |
| 492 | result += '\\'; |
| 493 | result += unquoted[ii]; |
| 494 | } else { |
| 495 | result += unquoted[ii]; |
| 496 | } |
| 497 | } |
| 498 | |
| 499 | return result; |
| 500 | } |
| 501 | |
| 502 | /***** Actual matching and rewriting code *****/ |
| 503 | |
| 504 | int RE::TryMatch(const StringPiece& text, |
| 505 | int startpos, |
| 506 | Anchor anchor, |
| 507 | bool empty_ok, |
| 508 | int *vec, |
| 509 | int vecsize) const { |
| 510 | pcre* re = (anchor == ANCHOR_BOTH) ? re_full_ : re_partial_; |
| 511 | if (re == NULL) { |
| 512 | //fprintf(stderr, "Matching against invalid re: %s\n", error_->c_str()); |
| 513 | return 0; |
| 514 | } |
| 515 | |
| 516 | pcre_extra extra = { 0, 0, 0, 0, 0, 0 }; |
| 517 | if (options_.match_limit() > 0) { |
| 518 | extra.flags |= PCRE_EXTRA_MATCH_LIMIT; |
| 519 | extra.match_limit = options_.match_limit(); |
| 520 | } |
| 521 | if (options_.match_limit_recursion() > 0) { |
| 522 | extra.flags |= PCRE_EXTRA_MATCH_LIMIT_RECURSION; |
| 523 | extra.match_limit_recursion = options_.match_limit_recursion(); |
| 524 | } |
| 525 | |
| 526 | int options = 0; |
| 527 | if (anchor != UNANCHORED) |
| 528 | options |= PCRE_ANCHORED; |
| 529 | if (!empty_ok) |
| 530 | options |= PCRE_NOTEMPTY; |
| 531 | |
| 532 | int rc = pcre_exec(re, // The regular expression object |
| 533 | &extra, |
| 534 | (text.data() == NULL) ? "" : text.data(), |
| 535 | text.size(), |
| 536 | startpos, |
| 537 | options, |
| 538 | vec, |
| 539 | vecsize); |
| 540 | |
| 541 | // Handle errors |
| 542 | if (rc == PCRE_ERROR_NOMATCH) { |
| 543 | return 0; |
| 544 | } else if (rc < 0) { |
| 545 | //fprintf(stderr, "Unexpected return code: %d when matching '%s'\n", |
| 546 | // re, pattern_.c_str()); |
| 547 | return 0; |
| 548 | } else if (rc == 0) { |
| 549 | // pcre_exec() returns 0 as a special case when the number of |
| 550 | // capturing subpatterns exceeds the size of the vector. |
| 551 | // When this happens, there is a match and the output vector |
| 552 | // is filled, but we miss out on the positions of the extra subpatterns. |
| 553 | rc = vecsize / 2; |
| 554 | } |
| 555 | |
| 556 | return rc; |
| 557 | } |
| 558 | |
| 559 | bool RE::DoMatchImpl(const StringPiece& text, |
| 560 | Anchor anchor, |
| 561 | int* consumed, |
| 562 | const Arg* const* args, |
| 563 | int n, |
| 564 | int* vec, |
| 565 | int vecsize) const { |
| 566 | assert((1 + n) * 3 <= vecsize); // results + PCRE workspace |
| 567 | int matches = TryMatch(text, 0, anchor, true, vec, vecsize); |
| 568 | assert(matches >= 0); // TryMatch never returns negatives |
| 569 | if (matches == 0) |
| 570 | return false; |
| 571 | |
| 572 | *consumed = vec[1]; |
| 573 | |
| 574 | if (n == 0 || args == NULL) { |
| 575 | // We are not interested in results |
| 576 | return true; |
| 577 | } |
| 578 | |
| 579 | if (NumberOfCapturingGroups() < n) { |
| 580 | // RE has fewer capturing groups than number of arg pointers passed in |
| 581 | return false; |
| 582 | } |
| 583 | |
| 584 | // If we got here, we must have matched the whole pattern. |
| 585 | // We do not need (can not do) any more checks on the value of 'matches' here |
| 586 | // -- see the comment for TryMatch. |
| 587 | for (int i = 0; i < n; i++) { |
| 588 | const int start = vec[2*(i+1)]; |
| 589 | const int limit = vec[2*(i+1)+1]; |
| 590 | if (!args[i]->Parse(text.data() + start, limit-start)) { |
| 591 | // TODO: Should we indicate what the error was? |
| 592 | return false; |
| 593 | } |
| 594 | } |
| 595 | |
| 596 | return true; |
| 597 | } |
| 598 | |
| 599 | bool RE::DoMatch(const StringPiece& text, |
| 600 | Anchor anchor, |
| 601 | int* consumed, |
| 602 | const Arg* const args[], |
| 603 | int n) const { |
| 604 | assert(n >= 0); |
| 605 | size_t const vecsize = (1 + n) * 3; // results + PCRE workspace |
| 606 | // (as for kVecSize) |
| 607 | int space[21]; // use stack allocation for small vecsize (common case) |
| 608 | int* vec = vecsize <= 21 ? space : new int[vecsize]; |
| 609 | bool retval = DoMatchImpl(text, anchor, consumed, args, n, vec, (int)vecsize); |
| 610 | if (vec != space) delete [] vec; |
| 611 | return retval; |
| 612 | } |
| 613 | |
| 614 | bool RE::Rewrite(string *out, const StringPiece &rewrite, |
| 615 | const StringPiece &text, int *vec, int veclen) const { |
| 616 | for (const char *s = rewrite.data(), *end = s + rewrite.size(); |
| 617 | s < end; s++) { |
| 618 | int c = *s; |
| 619 | if (c == '\\') { |
| 620 | c = *++s; |
| 621 | if (isdigit(c)) { |
| 622 | int n = (c - '0'); |
| 623 | if (n >= veclen) { |
| 624 | //fprintf(stderr, requested group %d in regexp %.*s\n", |
| 625 | // n, rewrite.size(), rewrite.data()); |
| 626 | return false; |
| 627 | } |
| 628 | int start = vec[2 * n]; |
| 629 | if (start >= 0) |
| 630 | out->append(text.data() + start, vec[2 * n + 1] - start); |
| 631 | } else if (c == '\\') { |
| 632 | *out += '\\'; |
| 633 | } else { |
| 634 | //fprintf(stderr, "invalid rewrite pattern: %.*s\n", |
| 635 | // rewrite.size(), rewrite.data()); |
| 636 | return false; |
| 637 | } |
| 638 | } else { |
| 639 | *out += c; |
| 640 | } |
| 641 | } |
| 642 | return true; |
| 643 | } |
| 644 | |
| 645 | // Return the number of capturing subpatterns, or -1 if the |
| 646 | // regexp wasn't valid on construction. |
| 647 | int RE::NumberOfCapturingGroups() const { |
| 648 | if (re_partial_ == NULL) return -1; |
| 649 | |
| 650 | int result; |
| 651 | int pcre_retval = pcre_fullinfo(re_partial_, // The regular expression object |
| 652 | NULL, // We did not study the pattern |
| 653 | PCRE_INFO_CAPTURECOUNT, |
| 654 | &result); |
| 655 | assert(pcre_retval == 0); |
| 656 | return result; |
| 657 | } |
| 658 | |
| 659 | /***** Parsers for various types *****/ |
| 660 | |
| 661 | bool Arg::parse_null(const char* str, int n, void* dest) { |
| 662 | // We fail if somebody asked us to store into a non-NULL void* pointer |
| 663 | return (dest == NULL); |
| 664 | } |
| 665 | |
| 666 | bool Arg::parse_string(const char* str, int n, void* dest) { |
| 667 | if (dest == NULL) return true; |
| 668 | reinterpret_cast<string*>(dest)->assign(str, n); |
| 669 | return true; |
| 670 | } |
| 671 | |
| 672 | bool Arg::parse_stringpiece(const char* str, int n, void* dest) { |
| 673 | if (dest == NULL) return true; |
| 674 | reinterpret_cast<StringPiece*>(dest)->set(str, n); |
| 675 | return true; |
| 676 | } |
| 677 | |
| 678 | bool Arg::parse_char(const char* str, int n, void* dest) { |
| 679 | if (n != 1) return false; |
| 680 | if (dest == NULL) return true; |
| 681 | *(reinterpret_cast<char*>(dest)) = str[0]; |
| 682 | return true; |
| 683 | } |
| 684 | |
| 685 | bool Arg::parse_uchar(const char* str, int n, void* dest) { |
| 686 | if (n != 1) return false; |
| 687 | if (dest == NULL) return true; |
| 688 | *(reinterpret_cast<unsigned char*>(dest)) = str[0]; |
| 689 | return true; |
| 690 | } |
| 691 | |
| 692 | // Largest number spec that we are willing to parse |
| 693 | static const int kMaxNumberLength = 32; |
| 694 | |
| 695 | // REQUIRES "buf" must have length at least kMaxNumberLength+1 |
| 696 | // REQUIRES "n > 0" |
| 697 | // Copies "str" into "buf" and null-terminates if necessary. |
| 698 | // Returns one of: |
| 699 | // a. "str" if no termination is needed |
| 700 | // b. "buf" if the string was copied and null-terminated |
| 701 | // c. "" if the input was invalid and has no hope of being parsed |
| 702 | static const char* TerminateNumber(char* buf, const char* str, int n) { |
| 703 | if ((n > 0) && isspace(*str)) { |
| 704 | // We are less forgiving than the strtoxxx() routines and do not |
| 705 | // allow leading spaces. |
| 706 | return ""; |
| 707 | } |
| 708 | |
| 709 | // See if the character right after the input text may potentially |
| 710 | // look like a digit. |
| 711 | if (isdigit(str[n]) || |
| 712 | ((str[n] >= 'a') && (str[n] <= 'f')) || |
| 713 | ((str[n] >= 'A') && (str[n] <= 'F'))) { |
| 714 | if (n > kMaxNumberLength) return ""; // Input too big to be a valid number |
| 715 | memcpy(buf, str, n); |
| 716 | buf[n] = '\0'; |
| 717 | return buf; |
| 718 | } else { |
| 719 | // We can parse right out of the supplied string, so return it. |
| 720 | return str; |
| 721 | } |
| 722 | } |
| 723 | |
| 724 | bool Arg::parse_long_radix(const char* str, |
| 725 | int n, |
| 726 | void* dest, |
| 727 | int radix) { |
| 728 | if (n == 0) return false; |
| 729 | char buf[kMaxNumberLength+1]; |
| 730 | str = TerminateNumber(buf, str, n); |
| 731 | char* end; |
| 732 | errno = 0; |
| 733 | long r = strtol(str, &end, radix); |
| 734 | if (end != str + n) return false; // Leftover junk |
| 735 | if (errno) return false; |
| 736 | if (dest == NULL) return true; |
| 737 | *(reinterpret_cast<long*>(dest)) = r; |
| 738 | return true; |
| 739 | } |
| 740 | |
| 741 | bool Arg::parse_ulong_radix(const char* str, |
| 742 | int n, |
| 743 | void* dest, |
| 744 | int radix) { |
| 745 | if (n == 0) return false; |
| 746 | char buf[kMaxNumberLength+1]; |
| 747 | str = TerminateNumber(buf, str, n); |
| 748 | if (str[0] == '-') return false; // strtoul() on a negative number?! |
| 749 | char* end; |
| 750 | errno = 0; |
| 751 | unsigned long r = strtoul(str, &end, radix); |
| 752 | if (end != str + n) return false; // Leftover junk |
| 753 | if (errno) return false; |
| 754 | if (dest == NULL) return true; |
| 755 | *(reinterpret_cast<unsigned long*>(dest)) = r; |
| 756 | return true; |
| 757 | } |
| 758 | |
| 759 | bool Arg::parse_short_radix(const char* str, |
| 760 | int n, |
| 761 | void* dest, |
| 762 | int radix) { |
| 763 | long r; |
| 764 | if (!parse_long_radix(str, n, &r, radix)) return false; // Could not parse |
| 765 | if (r < SHRT_MIN || r > SHRT_MAX) return false; // Out of range |
| 766 | if (dest == NULL) return true; |
| 767 | *(reinterpret_cast<short*>(dest)) = static_cast<short>(r); |
| 768 | return true; |
| 769 | } |
| 770 | |
| 771 | bool Arg::parse_ushort_radix(const char* str, |
| 772 | int n, |
| 773 | void* dest, |
| 774 | int radix) { |
| 775 | unsigned long r; |
| 776 | if (!parse_ulong_radix(str, n, &r, radix)) return false; // Could not parse |
| 777 | if (r > USHRT_MAX) return false; // Out of range |
| 778 | if (dest == NULL) return true; |
| 779 | *(reinterpret_cast<unsigned short*>(dest)) = static_cast<unsigned short>(r); |
| 780 | return true; |
| 781 | } |
| 782 | |
| 783 | bool Arg::parse_int_radix(const char* str, |
| 784 | int n, |
| 785 | void* dest, |
| 786 | int radix) { |
| 787 | long r; |
| 788 | if (!parse_long_radix(str, n, &r, radix)) return false; // Could not parse |
| 789 | if (r < INT_MIN || r > INT_MAX) return false; // Out of range |
| 790 | if (dest == NULL) return true; |
| 791 | *(reinterpret_cast<int*>(dest)) = r; |
| 792 | return true; |
| 793 | } |
| 794 | |
| 795 | bool Arg::parse_uint_radix(const char* str, |
| 796 | int n, |
| 797 | void* dest, |
| 798 | int radix) { |
| 799 | unsigned long r; |
| 800 | if (!parse_ulong_radix(str, n, &r, radix)) return false; // Could not parse |
| 801 | if (r > UINT_MAX) return false; // Out of range |
| 802 | if (dest == NULL) return true; |
| 803 | *(reinterpret_cast<unsigned int*>(dest)) = r; |
| 804 | return true; |
| 805 | } |
| 806 | |
| 807 | bool Arg::parse_longlong_radix(const char* str, |
| 808 | int n, |
| 809 | void* dest, |
| 810 | int radix) { |
| 811 | #ifndef HAVE_LONG_LONG |
| 812 | return false; |
| 813 | #else |
| 814 | if (n == 0) return false; |
| 815 | char buf[kMaxNumberLength+1]; |
| 816 | str = TerminateNumber(buf, str, n); |
| 817 | char* end; |
| 818 | errno = 0; |
| 819 | #if defined HAVE_STRTOQ |
| 820 | long long r = strtoq(str, &end, radix); |
| 821 | #elif defined HAVE_STRTOLL |
| 822 | long long r = strtoll(str, &end, radix); |
| 823 | #elif defined HAVE__STRTOI64 |
| 824 | long long r = _strtoi64(str, &end, radix); |
| 825 | #elif defined HAVE_STRTOIMAX |
| 826 | long long r = strtoimax(str, &end, radix); |
| 827 | #else |
| 828 | #error parse_longlong_radix: cannot convert input to a long-long |
| 829 | #endif |
| 830 | if (end != str + n) return false; // Leftover junk |
| 831 | if (errno) return false; |
| 832 | if (dest == NULL) return true; |
| 833 | *(reinterpret_cast<long long*>(dest)) = r; |
| 834 | return true; |
| 835 | #endif /* HAVE_LONG_LONG */ |
| 836 | } |
| 837 | |
| 838 | bool Arg::parse_ulonglong_radix(const char* str, |
| 839 | int n, |
| 840 | void* dest, |
| 841 | int radix) { |
| 842 | #ifndef HAVE_UNSIGNED_LONG_LONG |
| 843 | return false; |
| 844 | #else |
| 845 | if (n == 0) return false; |
| 846 | char buf[kMaxNumberLength+1]; |
| 847 | str = TerminateNumber(buf, str, n); |
| 848 | if (str[0] == '-') return false; // strtoull() on a negative number?! |
| 849 | char* end; |
| 850 | errno = 0; |
| 851 | #if defined HAVE_STRTOQ |
| 852 | unsigned long long r = strtouq(str, &end, radix); |
| 853 | #elif defined HAVE_STRTOLL |
| 854 | unsigned long long r = strtoull(str, &end, radix); |
| 855 | #elif defined HAVE__STRTOI64 |
| 856 | unsigned long long r = _strtoui64(str, &end, radix); |
| 857 | #elif defined HAVE_STRTOIMAX |
| 858 | unsigned long long r = strtoumax(str, &end, radix); |
| 859 | #else |
| 860 | #error parse_ulonglong_radix: cannot convert input to a long-long |
| 861 | #endif |
| 862 | if (end != str + n) return false; // Leftover junk |
| 863 | if (errno) return false; |
| 864 | if (dest == NULL) return true; |
| 865 | *(reinterpret_cast<unsigned long long*>(dest)) = r; |
| 866 | return true; |
| 867 | #endif /* HAVE_UNSIGNED_LONG_LONG */ |
| 868 | } |
| 869 | |
| 870 | bool Arg::parse_double(const char* str, int n, void* dest) { |
| 871 | if (n == 0) return false; |
| 872 | static const int kMaxLength = 200; |
| 873 | char buf[kMaxLength]; |
| 874 | if (n >= kMaxLength) return false; |
| 875 | memcpy(buf, str, n); |
| 876 | buf[n] = '\0'; |
| 877 | errno = 0; |
| 878 | char* end; |
| 879 | double r = strtod(buf, &end); |
| 880 | if (end != buf + n) return false; // Leftover junk |
| 881 | if (errno) return false; |
| 882 | if (dest == NULL) return true; |
| 883 | *(reinterpret_cast<double*>(dest)) = r; |
| 884 | return true; |
| 885 | } |
| 886 | |
| 887 | bool Arg::parse_float(const char* str, int n, void* dest) { |
| 888 | double r; |
| 889 | if (!parse_double(str, n, &r)) return false; |
| 890 | if (dest == NULL) return true; |
| 891 | *(reinterpret_cast<float*>(dest)) = static_cast<float>(r); |
| 892 | return true; |
| 893 | } |
| 894 | |
| 895 | |
| 896 | #define DEFINE_INTEGER_PARSERS(name) \ |
| 897 | bool Arg::parse_##name(const char* str, int n, void* dest) { \ |
| 898 | return parse_##name##_radix(str, n, dest, 10); \ |
| 899 | } \ |
| 900 | bool Arg::parse_##name##_hex(const char* str, int n, void* dest) { \ |
| 901 | return parse_##name##_radix(str, n, dest, 16); \ |
| 902 | } \ |
| 903 | bool Arg::parse_##name##_octal(const char* str, int n, void* dest) { \ |
| 904 | return parse_##name##_radix(str, n, dest, 8); \ |
| 905 | } \ |
| 906 | bool Arg::parse_##name##_cradix(const char* str, int n, void* dest) { \ |
| 907 | return parse_##name##_radix(str, n, dest, 0); \ |
| 908 | } |
| 909 | |
| 910 | DEFINE_INTEGER_PARSERS(short) /* */ |
| 911 | DEFINE_INTEGER_PARSERS(ushort) /* */ |
| 912 | DEFINE_INTEGER_PARSERS(int) /* Don't use semicolons after these */ |
| 913 | DEFINE_INTEGER_PARSERS(uint) /* statements because they can cause */ |
| 914 | DEFINE_INTEGER_PARSERS(long) /* compiler warnings if the checking */ |
| 915 | DEFINE_INTEGER_PARSERS(ulong) /* level is turned up high enough. */ |
| 916 | DEFINE_INTEGER_PARSERS(longlong) /* */ |
| 917 | DEFINE_INTEGER_PARSERS(ulonglong) /* */ |
| 918 | |
| 919 | #undef DEFINE_INTEGER_PARSERS |
| 920 | |
| 921 | } // namespace pcrecpp |