| 29 |
// |
// |
| 30 |
// Author: Sanjay Ghemawat |
// Author: Sanjay Ghemawat |
| 31 |
|
|
| 32 |
|
#ifdef HAVE_CONFIG_H |
| 33 |
|
#include <config.h> |
| 34 |
|
#endif |
| 35 |
|
|
| 36 |
#include <stdlib.h> |
#include <stdlib.h> |
| 37 |
#include <stdio.h> |
#include <stdio.h> |
| 38 |
#include <ctype.h> |
#include <ctype.h> |
| 40 |
#include <assert.h> |
#include <assert.h> |
| 41 |
#include <errno.h> |
#include <errno.h> |
| 42 |
#include <string> |
#include <string> |
| 43 |
#include "config.h" |
#include <algorithm> |
| 44 |
// We need this to compile the proper dll on windows/msys. This is copied |
|
| 45 |
// from pcre_internal.h. It would probably be better just to include that. |
#include "pcrecpp_internal.h" |
| 46 |
#define PCRE_DEFINITION /* Win32 __declspec(export) trigger for .dll */ |
#include <pcre.h> |
|
#include "pcre.h" |
|
|
#include "pcre_stringpiece.h" |
|
| 47 |
#include "pcrecpp.h" |
#include "pcrecpp.h" |
| 48 |
|
#include "pcre_stringpiece.h" |
| 49 |
|
|
| 50 |
|
|
| 51 |
namespace pcrecpp { |
namespace pcrecpp { |
| 55 |
static const int kVecSize = (1 + kMaxArgs) * 3; // results + PCRE workspace |
static const int kVecSize = (1 + kMaxArgs) * 3; // results + PCRE workspace |
| 56 |
|
|
| 57 |
// Special object that stands-in for no argument |
// Special object that stands-in for no argument |
| 58 |
Arg no_arg((void*)NULL); |
PCRECPP_EXP_DECL Arg no_arg((void*)NULL); |
| 59 |
|
|
| 60 |
// If a regular expression has no error, its error_ field points here |
// If a regular expression has no error, its error_ field points here |
| 61 |
static const string empty_string; |
static const string empty_string; |
| 63 |
// If the user doesn't ask for any options, we just use this one |
// If the user doesn't ask for any options, we just use this one |
| 64 |
static RE_Options default_options; |
static RE_Options default_options; |
| 65 |
|
|
| 66 |
void RE::Init(const char* pat, const RE_Options* options) { |
void RE::Init(const string& pat, const RE_Options* options) { |
| 67 |
pattern_ = pat; |
pattern_ = pat; |
| 68 |
if (options == NULL) { |
if (options == NULL) { |
| 69 |
options_ = default_options; |
options_ = default_options; |
| 76 |
|
|
| 77 |
re_partial_ = Compile(UNANCHORED); |
re_partial_ = Compile(UNANCHORED); |
| 78 |
if (re_partial_ != NULL) { |
if (re_partial_ != NULL) { |
| 79 |
// Check for complicated patterns. The following change is |
re_full_ = Compile(ANCHOR_BOTH); |
|
// conservative in that it may treat some "simple" patterns |
|
|
// as "complex" (e.g., if the vertical bar is in a character |
|
|
// class or is escaped). But it seems good enough. |
|
|
if (strchr(pat, '|') == NULL) { |
|
|
// Simple pattern: we can use position-based checks to perform |
|
|
// fully anchored matches |
|
|
re_full_ = re_partial_; |
|
|
} else { |
|
|
// We need a special pattern for anchored matches |
|
|
re_full_ = Compile(ANCHOR_BOTH); |
|
|
} |
|
| 80 |
} |
} |
| 81 |
} |
} |
| 82 |
|
|
| 83 |
|
void RE::Cleanup() { |
| 84 |
|
if (re_full_ != NULL) (*pcre_free)(re_full_); |
| 85 |
|
if (re_partial_ != NULL) (*pcre_free)(re_partial_); |
| 86 |
|
if (error_ != &empty_string) delete error_; |
| 87 |
|
} |
| 88 |
|
|
| 89 |
|
|
| 90 |
RE::~RE() { |
RE::~RE() { |
| 91 |
if (re_full_ != NULL && re_full_ != re_partial_) (*pcre_free)(re_full_); |
Cleanup(); |
|
if (re_partial_ != NULL) (*pcre_free)(re_partial_); |
|
|
if (error_ != &empty_string) delete error_; |
|
| 92 |
} |
} |
| 93 |
|
|
| 94 |
|
|
| 95 |
pcre* RE::Compile(Anchor anchor) { |
pcre* RE::Compile(Anchor anchor) { |
| 96 |
// First, convert RE_Options into pcre options |
// First, convert RE_Options into pcre options |
| 97 |
int pcre_options = 0; |
int pcre_options = 0; |
| 98 |
if (options_.utf8()) |
pcre_options = options_.all_options(); |
|
pcre_options |= PCRE_UTF8; |
|
| 99 |
|
|
| 100 |
// Special treatment for anchoring. This is needed because at |
// Special treatment for anchoring. This is needed because at |
| 101 |
// runtime pcre only provides an option for anchoring at the |
// runtime pcre only provides an option for anchoring at the |
| 329 |
return true; |
return true; |
| 330 |
} |
} |
| 331 |
|
|
| 332 |
|
// Returns PCRE_NEWLINE_CRLF, PCRE_NEWLINE_CR, or PCRE_NEWLINE_LF. |
| 333 |
|
// Note that PCRE_NEWLINE_CRLF is defined to be P_N_CR | P_N_LF. |
| 334 |
|
static int NewlineMode(int pcre_options) { |
| 335 |
|
// TODO: if we can make it threadsafe, cache this var |
| 336 |
|
int newline_mode = 0; |
| 337 |
|
/* if (newline_mode) return newline_mode; */ // do this once it's cached |
| 338 |
|
if (pcre_options & (PCRE_NEWLINE_CRLF|PCRE_NEWLINE_CR|PCRE_NEWLINE_LF)) { |
| 339 |
|
newline_mode = (pcre_options & |
| 340 |
|
(PCRE_NEWLINE_CRLF|PCRE_NEWLINE_CR|PCRE_NEWLINE_LF)); |
| 341 |
|
} else { |
| 342 |
|
int newline; |
| 343 |
|
pcre_config(PCRE_CONFIG_NEWLINE, &newline); |
| 344 |
|
if (newline == 10) |
| 345 |
|
newline_mode = PCRE_NEWLINE_LF; |
| 346 |
|
else if (newline == 13) |
| 347 |
|
newline_mode = PCRE_NEWLINE_CR; |
| 348 |
|
else if (newline == 3338) |
| 349 |
|
newline_mode = PCRE_NEWLINE_CRLF; |
| 350 |
|
else |
| 351 |
|
assert("" == "Unexpected return value from pcre_config(NEWLINE)"); |
| 352 |
|
} |
| 353 |
|
return newline_mode; |
| 354 |
|
} |
| 355 |
|
|
| 356 |
int RE::GlobalReplace(const StringPiece& rewrite, |
int RE::GlobalReplace(const StringPiece& rewrite, |
| 357 |
string *str) const { |
string *str) const { |
| 358 |
int count = 0; |
int count = 0; |
| 371 |
if (matchstart == matchend && matchstart == lastend) { |
if (matchstart == matchend && matchstart == lastend) { |
| 372 |
// advance one character if we matched an empty string at the same |
// advance one character if we matched an empty string at the same |
| 373 |
// place as the last match occurred |
// place as the last match occurred |
| 374 |
if (start < static_cast<int>(str->length())) |
matchend = start + 1; |
| 375 |
out.push_back((*str)[start]); |
// If the current char is CR and we're in CRLF mode, skip LF too. |
| 376 |
start++; |
// Note it's better to call pcre_fullinfo() than to examine |
| 377 |
|
// all_options(), since options_ could have changed bewteen |
| 378 |
|
// compile-time and now, but this is simpler and safe enough. |
| 379 |
|
if (start+1 < static_cast<int>(str->length()) && |
| 380 |
|
(*str)[start] == '\r' && (*str)[start+1] == '\n' && |
| 381 |
|
NewlineMode(options_.all_options()) == PCRE_NEWLINE_CRLF) { |
| 382 |
|
matchend++; |
| 383 |
|
} |
| 384 |
|
// We also need to advance more than one char if we're in utf8 mode. |
| 385 |
|
#ifdef SUPPORT_UTF8 |
| 386 |
|
if (options_.utf8()) { |
| 387 |
|
while (matchend < static_cast<int>(str->length()) && |
| 388 |
|
((*str)[matchend] & 0xc0) == 0x80) |
| 389 |
|
matchend++; |
| 390 |
|
} |
| 391 |
|
#endif |
| 392 |
|
if (matchend <= static_cast<int>(str->length())) |
| 393 |
|
out.append(*str, start, matchend - start); |
| 394 |
|
start = matchend; |
| 395 |
} else { |
} else { |
| 396 |
out.append(*str, start, matchstart - start); |
out.append(*str, start, matchstart - start); |
| 397 |
Rewrite(&out, rewrite, *str, vec, matches); |
Rewrite(&out, rewrite, *str, vec, matches); |
| 417 |
int matches = TryMatch(text, 0, UNANCHORED, vec, kVecSize); |
int matches = TryMatch(text, 0, UNANCHORED, vec, kVecSize); |
| 418 |
if (matches == 0) |
if (matches == 0) |
| 419 |
return false; |
return false; |
| 420 |
out->clear(); |
out->erase(); |
| 421 |
return Rewrite(out, rewrite, text, vec, matches); |
return Rewrite(out, rewrite, text, vec, matches); |
| 422 |
} |
} |
| 423 |
|
|
| 424 |
|
/*static*/ string RE::QuoteMeta(const StringPiece& unquoted) { |
| 425 |
|
string result; |
| 426 |
|
|
| 427 |
|
// Escape any ascii character not in [A-Za-z_0-9]. |
| 428 |
|
// |
| 429 |
|
// Note that it's legal to escape a character even if it has no |
| 430 |
|
// special meaning in a regular expression -- so this function does |
| 431 |
|
// that. (This also makes it identical to the perl function of the |
| 432 |
|
// same name; see `perldoc -f quotemeta`.) |
| 433 |
|
for (int ii = 0; ii < unquoted.size(); ++ii) { |
| 434 |
|
// Note that using 'isalnum' here raises the benchmark time from |
| 435 |
|
// 32ns to 58ns: |
| 436 |
|
if ((unquoted[ii] < 'a' || unquoted[ii] > 'z') && |
| 437 |
|
(unquoted[ii] < 'A' || unquoted[ii] > 'Z') && |
| 438 |
|
(unquoted[ii] < '0' || unquoted[ii] > '9') && |
| 439 |
|
unquoted[ii] != '_' && |
| 440 |
|
// If this is the part of a UTF8 or Latin1 character, we need |
| 441 |
|
// to copy this byte without escaping. Experimentally this is |
| 442 |
|
// what works correctly with the regexp library. |
| 443 |
|
!(unquoted[ii] & 128)) { |
| 444 |
|
result += '\\'; |
| 445 |
|
} |
| 446 |
|
result += unquoted[ii]; |
| 447 |
|
} |
| 448 |
|
|
| 449 |
|
return result; |
| 450 |
|
} |
| 451 |
|
|
| 452 |
/***** Actual matching and rewriting code *****/ |
/***** Actual matching and rewriting code *****/ |
| 453 |
|
|
| 454 |
int RE::TryMatch(const StringPiece& text, |
int RE::TryMatch(const StringPiece& text, |
| 462 |
return 0; |
return 0; |
| 463 |
} |
} |
| 464 |
|
|
| 465 |
pcre_extra extra = { 0 }; |
pcre_extra extra = { 0, 0, 0, 0, 0, 0 }; |
| 466 |
if (options_.match_limit() > 0) { |
if (options_.match_limit() > 0) { |
| 467 |
extra.flags = PCRE_EXTRA_MATCH_LIMIT; |
extra.flags |= PCRE_EXTRA_MATCH_LIMIT; |
| 468 |
extra.match_limit = options_.match_limit(); |
extra.match_limit = options_.match_limit(); |
| 469 |
} |
} |
| 470 |
|
if (options_.match_limit_recursion() > 0) { |
| 471 |
|
extra.flags |= PCRE_EXTRA_MATCH_LIMIT_RECURSION; |
| 472 |
|
extra.match_limit_recursion = options_.match_limit_recursion(); |
| 473 |
|
} |
| 474 |
int rc = pcre_exec(re, // The regular expression object |
int rc = pcre_exec(re, // The regular expression object |
| 475 |
&extra, |
&extra, |
| 476 |
text.data(), |
(text.data() == NULL) ? "" : text.data(), |
| 477 |
text.size(), |
text.size(), |
| 478 |
startpos, |
startpos, |
| 479 |
(anchor == UNANCHORED) ? 0 : PCRE_ANCHORED, |
(anchor == UNANCHORED) ? 0 : PCRE_ANCHORED, |
| 495 |
rc = vecsize / 2; |
rc = vecsize / 2; |
| 496 |
} |
} |
| 497 |
|
|
|
if ((anchor == ANCHOR_BOTH) && (re_full_ == re_partial_)) { |
|
|
// We need an extra check to make sure that the match extended |
|
|
// to the end of the input string |
|
|
assert(vec[0] == 0); // PCRE_ANCHORED forces starting match |
|
|
if (vec[1] != text.size()) return 0; // Did not get ending match |
|
|
} |
|
|
|
|
| 498 |
return rc; |
return rc; |
| 499 |
} |
} |
| 500 |
|
|
| 513 |
|
|
| 514 |
*consumed = vec[1]; |
*consumed = vec[1]; |
| 515 |
|
|
| 516 |
if (args == NULL) { |
if (n == 0 || args == NULL) { |
| 517 |
// We are not interested in results |
// We are not interested in results |
| 518 |
return true; |
return true; |
| 519 |
} |
} |
| 520 |
|
|
| 521 |
|
if (NumberOfCapturingGroups() < n) { |
| 522 |
|
// RE has fewer capturing groups than number of arg pointers passed in |
| 523 |
|
return false; |
| 524 |
|
} |
| 525 |
|
|
| 526 |
// If we got here, we must have matched the whole pattern. |
// If we got here, we must have matched the whole pattern. |
| 527 |
// We do not need (can not do) any more checks on the value of 'matches' here |
// We do not need (can not do) any more checks on the value of 'matches' here |
| 528 |
// -- see the comment for TryMatch. |
// -- see the comment for TryMatch. |
| 586 |
|
|
| 587 |
// Return the number of capturing subpatterns, or -1 if the |
// Return the number of capturing subpatterns, or -1 if the |
| 588 |
// regexp wasn't valid on construction. |
// regexp wasn't valid on construction. |
| 589 |
int RE::NumberOfCapturingGroups() { |
int RE::NumberOfCapturingGroups() const { |
| 590 |
if (re_partial_ == NULL) return -1; |
if (re_partial_ == NULL) return -1; |
| 591 |
|
|
| 592 |
int result; |
int result; |
| 682 |
if (n == 0) return false; |
if (n == 0) return false; |
| 683 |
char buf[kMaxNumberLength+1]; |
char buf[kMaxNumberLength+1]; |
| 684 |
str = TerminateNumber(buf, str, n); |
str = TerminateNumber(buf, str, n); |
| 685 |
|
if (str[0] == '-') return false; // strtoul() on a negative number?! |
| 686 |
char* end; |
char* end; |
| 687 |
errno = 0; |
errno = 0; |
| 688 |
unsigned long r = strtoul(str, &end, radix); |
unsigned long r = strtoul(str, &end, radix); |
| 772 |
if (n == 0) return false; |
if (n == 0) return false; |
| 773 |
char buf[kMaxNumberLength+1]; |
char buf[kMaxNumberLength+1]; |
| 774 |
str = TerminateNumber(buf, str, n); |
str = TerminateNumber(buf, str, n); |
| 775 |
|
if (str[0] == '-') return false; // strtoull() on a negative number?! |
| 776 |
char* end; |
char* end; |
| 777 |
errno = 0; |
errno = 0; |
| 778 |
#if defined HAVE_STRTOQ |
#if defined HAVE_STRTOQ |
| 827 |
return parse_##name##_radix(str, n, dest, 0); \ |
return parse_##name##_radix(str, n, dest, 0); \ |
| 828 |
} |
} |
| 829 |
|
|
| 830 |
DEFINE_INTEGER_PARSERS(short); |
DEFINE_INTEGER_PARSERS(short) /* */ |
| 831 |
DEFINE_INTEGER_PARSERS(ushort); |
DEFINE_INTEGER_PARSERS(ushort) /* */ |
| 832 |
DEFINE_INTEGER_PARSERS(int); |
DEFINE_INTEGER_PARSERS(int) /* Don't use semicolons after these */ |
| 833 |
DEFINE_INTEGER_PARSERS(uint); |
DEFINE_INTEGER_PARSERS(uint) /* statements because they can cause */ |
| 834 |
DEFINE_INTEGER_PARSERS(long); |
DEFINE_INTEGER_PARSERS(long) /* compiler warnings if the checking */ |
| 835 |
DEFINE_INTEGER_PARSERS(ulong); |
DEFINE_INTEGER_PARSERS(ulong) /* level is turned up high enough. */ |
| 836 |
DEFINE_INTEGER_PARSERS(longlong); |
DEFINE_INTEGER_PARSERS(longlong) /* */ |
| 837 |
DEFINE_INTEGER_PARSERS(ulonglong); |
DEFINE_INTEGER_PARSERS(ulonglong) /* */ |
| 838 |
|
|
| 839 |
#undef DEFINE_INTEGER_PARSERS |
#undef DEFINE_INTEGER_PARSERS |
| 840 |
|
|