29 |
// |
// |
30 |
// Author: Sanjay Ghemawat |
// Author: Sanjay Ghemawat |
31 |
|
|
32 |
|
#ifdef HAVE_CONFIG_H |
33 |
|
#include <config.h> |
34 |
|
#endif |
35 |
|
|
36 |
|
#ifdef _WIN32 |
37 |
|
#define HAVE_STRTOQ 1 |
38 |
|
#define strtoll _strtoui64 |
39 |
|
#define strtoull _strtoi64 |
40 |
|
#endif |
41 |
|
|
42 |
#include <stdlib.h> |
#include <stdlib.h> |
43 |
#include <stdio.h> |
#include <stdio.h> |
44 |
#include <ctype.h> |
#include <ctype.h> |
46 |
#include <assert.h> |
#include <assert.h> |
47 |
#include <errno.h> |
#include <errno.h> |
48 |
#include <string> |
#include <string> |
49 |
#include "config.h" |
#include <algorithm> |
50 |
// We need this to compile the proper dll on windows/msys. This is copied |
|
51 |
// from pcre_internal.h. It would probably be better just to include that. |
#include "pcrecpp_internal.h" |
52 |
#define PCRE_DEFINITION /* Win32 __declspec(export) trigger for .dll */ |
#include <pcre.h> |
|
#include "pcre.h" |
|
|
#include "pcre_stringpiece.h" |
|
53 |
#include "pcrecpp.h" |
#include "pcrecpp.h" |
54 |
|
#include "pcre_stringpiece.h" |
55 |
|
|
56 |
|
|
57 |
namespace pcrecpp { |
namespace pcrecpp { |
61 |
static const int kVecSize = (1 + kMaxArgs) * 3; // results + PCRE workspace |
static const int kVecSize = (1 + kMaxArgs) * 3; // results + PCRE workspace |
62 |
|
|
63 |
// Special object that stands-in for no argument |
// Special object that stands-in for no argument |
64 |
Arg no_arg((void*)NULL); |
PCRECPP_EXP_DEFN Arg no_arg((void*)NULL); |
65 |
|
|
66 |
// If a regular expression has no error, its error_ field points here |
// If a regular expression has no error, its error_ field points here |
67 |
static const string empty_string; |
static const string empty_string; |
69 |
// 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 |
70 |
static RE_Options default_options; |
static RE_Options default_options; |
71 |
|
|
72 |
void RE::Init(const char* pat, const RE_Options* options) { |
void RE::Init(const string& pat, const RE_Options* options) { |
73 |
pattern_ = pat; |
pattern_ = pat; |
74 |
if (options == NULL) { |
if (options == NULL) { |
75 |
options_ = default_options; |
options_ = default_options; |
82 |
|
|
83 |
re_partial_ = Compile(UNANCHORED); |
re_partial_ = Compile(UNANCHORED); |
84 |
if (re_partial_ != NULL) { |
if (re_partial_ != NULL) { |
85 |
// 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); |
|
|
} |
|
86 |
} |
} |
87 |
} |
} |
88 |
|
|
89 |
|
void RE::Cleanup() { |
90 |
|
if (re_full_ != NULL) (*pcre_free)(re_full_); |
91 |
|
if (re_partial_ != NULL) (*pcre_free)(re_partial_); |
92 |
|
if (error_ != &empty_string) delete error_; |
93 |
|
} |
94 |
|
|
95 |
|
|
96 |
RE::~RE() { |
RE::~RE() { |
97 |
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_; |
|
98 |
} |
} |
99 |
|
|
100 |
|
|
101 |
pcre* RE::Compile(Anchor anchor) { |
pcre* RE::Compile(Anchor anchor) { |
102 |
// First, convert RE_Options into pcre options |
// First, convert RE_Options into pcre options |
103 |
int pcre_options = 0; |
int pcre_options = 0; |
104 |
if (options_.utf8()) |
pcre_options = options_.all_options(); |
|
pcre_options |= PCRE_UTF8; |
|
105 |
|
|
106 |
// Special treatment for anchoring. This is needed because at |
// Special treatment for anchoring. This is needed because at |
107 |
// runtime pcre only provides an option for anchoring at the |
// runtime pcre only provides an option for anchoring at the |
335 |
return true; |
return true; |
336 |
} |
} |
337 |
|
|
338 |
|
// Returns PCRE_NEWLINE_CRLF, PCRE_NEWLINE_CR, or PCRE_NEWLINE_LF. |
339 |
|
// Note that PCRE_NEWLINE_CRLF is defined to be P_N_CR | P_N_LF. |
340 |
|
static int NewlineMode(int pcre_options) { |
341 |
|
// TODO: if we can make it threadsafe, cache this var |
342 |
|
int newline_mode = 0; |
343 |
|
/* if (newline_mode) return newline_mode; */ // do this once it's cached |
344 |
|
if (pcre_options & (PCRE_NEWLINE_CRLF|PCRE_NEWLINE_CR|PCRE_NEWLINE_LF)) { |
345 |
|
newline_mode = (pcre_options & |
346 |
|
(PCRE_NEWLINE_CRLF|PCRE_NEWLINE_CR|PCRE_NEWLINE_LF)); |
347 |
|
} else { |
348 |
|
int newline; |
349 |
|
pcre_config(PCRE_CONFIG_NEWLINE, &newline); |
350 |
|
if (newline == 10) |
351 |
|
newline_mode = PCRE_NEWLINE_LF; |
352 |
|
else if (newline == 13) |
353 |
|
newline_mode = PCRE_NEWLINE_CR; |
354 |
|
else if (newline == 3338) |
355 |
|
newline_mode = PCRE_NEWLINE_CRLF; |
356 |
|
else |
357 |
|
assert("" == "Unexpected return value from pcre_config(NEWLINE)"); |
358 |
|
} |
359 |
|
return newline_mode; |
360 |
|
} |
361 |
|
|
362 |
int RE::GlobalReplace(const StringPiece& rewrite, |
int RE::GlobalReplace(const StringPiece& rewrite, |
363 |
string *str) const { |
string *str) const { |
364 |
int count = 0; |
int count = 0; |
377 |
if (matchstart == matchend && matchstart == lastend) { |
if (matchstart == matchend && matchstart == lastend) { |
378 |
// advance one character if we matched an empty string at the same |
// advance one character if we matched an empty string at the same |
379 |
// place as the last match occurred |
// place as the last match occurred |
380 |
if (start < static_cast<int>(str->length())) |
matchend = start + 1; |
381 |
out.push_back((*str)[start]); |
// If the current char is CR and we're in CRLF mode, skip LF too. |
382 |
start++; |
// Note it's better to call pcre_fullinfo() than to examine |
383 |
|
// all_options(), since options_ could have changed bewteen |
384 |
|
// compile-time and now, but this is simpler and safe enough. |
385 |
|
if (start+1 < static_cast<int>(str->length()) && |
386 |
|
(*str)[start] == '\r' && (*str)[start+1] == '\n' && |
387 |
|
NewlineMode(options_.all_options()) == PCRE_NEWLINE_CRLF) { |
388 |
|
matchend++; |
389 |
|
} |
390 |
|
// We also need to advance more than one char if we're in utf8 mode. |
391 |
|
#ifdef SUPPORT_UTF8 |
392 |
|
if (options_.utf8()) { |
393 |
|
while (matchend < static_cast<int>(str->length()) && |
394 |
|
((*str)[matchend] & 0xc0) == 0x80) |
395 |
|
matchend++; |
396 |
|
} |
397 |
|
#endif |
398 |
|
if (matchend <= static_cast<int>(str->length())) |
399 |
|
out.append(*str, start, matchend - start); |
400 |
|
start = matchend; |
401 |
} else { |
} else { |
402 |
out.append(*str, start, matchstart - start); |
out.append(*str, start, matchstart - start); |
403 |
Rewrite(&out, rewrite, *str, vec, matches); |
Rewrite(&out, rewrite, *str, vec, matches); |
423 |
int matches = TryMatch(text, 0, UNANCHORED, vec, kVecSize); |
int matches = TryMatch(text, 0, UNANCHORED, vec, kVecSize); |
424 |
if (matches == 0) |
if (matches == 0) |
425 |
return false; |
return false; |
426 |
out->clear(); |
out->erase(); |
427 |
return Rewrite(out, rewrite, text, vec, matches); |
return Rewrite(out, rewrite, text, vec, matches); |
428 |
} |
} |
429 |
|
|
430 |
|
/*static*/ string RE::QuoteMeta(const StringPiece& unquoted) { |
431 |
|
string result; |
432 |
|
|
433 |
|
// Escape any ascii character not in [A-Za-z_0-9]. |
434 |
|
// |
435 |
|
// Note that it's legal to escape a character even if it has no |
436 |
|
// special meaning in a regular expression -- so this function does |
437 |
|
// that. (This also makes it identical to the perl function of the |
438 |
|
// same name; see `perldoc -f quotemeta`.) |
439 |
|
for (int ii = 0; ii < unquoted.size(); ++ii) { |
440 |
|
// Note that using 'isalnum' here raises the benchmark time from |
441 |
|
// 32ns to 58ns: |
442 |
|
if ((unquoted[ii] < 'a' || unquoted[ii] > 'z') && |
443 |
|
(unquoted[ii] < 'A' || unquoted[ii] > 'Z') && |
444 |
|
(unquoted[ii] < '0' || unquoted[ii] > '9') && |
445 |
|
unquoted[ii] != '_' && |
446 |
|
// If this is the part of a UTF8 or Latin1 character, we need |
447 |
|
// to copy this byte without escaping. Experimentally this is |
448 |
|
// what works correctly with the regexp library. |
449 |
|
!(unquoted[ii] & 128)) { |
450 |
|
result += '\\'; |
451 |
|
} |
452 |
|
result += unquoted[ii]; |
453 |
|
} |
454 |
|
|
455 |
|
return result; |
456 |
|
} |
457 |
|
|
458 |
/***** Actual matching and rewriting code *****/ |
/***** Actual matching and rewriting code *****/ |
459 |
|
|
460 |
int RE::TryMatch(const StringPiece& text, |
int RE::TryMatch(const StringPiece& text, |
468 |
return 0; |
return 0; |
469 |
} |
} |
470 |
|
|
471 |
pcre_extra extra = { 0 }; |
pcre_extra extra = { 0, 0, 0, 0, 0, 0 }; |
472 |
if (options_.match_limit() > 0) { |
if (options_.match_limit() > 0) { |
473 |
extra.flags = PCRE_EXTRA_MATCH_LIMIT; |
extra.flags |= PCRE_EXTRA_MATCH_LIMIT; |
474 |
extra.match_limit = options_.match_limit(); |
extra.match_limit = options_.match_limit(); |
475 |
} |
} |
476 |
|
if (options_.match_limit_recursion() > 0) { |
477 |
|
extra.flags |= PCRE_EXTRA_MATCH_LIMIT_RECURSION; |
478 |
|
extra.match_limit_recursion = options_.match_limit_recursion(); |
479 |
|
} |
480 |
int rc = pcre_exec(re, // The regular expression object |
int rc = pcre_exec(re, // The regular expression object |
481 |
&extra, |
&extra, |
482 |
text.data(), |
(text.data() == NULL) ? "" : text.data(), |
483 |
text.size(), |
text.size(), |
484 |
startpos, |
startpos, |
485 |
(anchor == UNANCHORED) ? 0 : PCRE_ANCHORED, |
(anchor == UNANCHORED) ? 0 : PCRE_ANCHORED, |
501 |
rc = vecsize / 2; |
rc = vecsize / 2; |
502 |
} |
} |
503 |
|
|
|
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 |
|
|
} |
|
|
|
|
504 |
return rc; |
return rc; |
505 |
} |
} |
506 |
|
|
519 |
|
|
520 |
*consumed = vec[1]; |
*consumed = vec[1]; |
521 |
|
|
522 |
if (args == NULL) { |
if (n == 0 || args == NULL) { |
523 |
// We are not interested in results |
// We are not interested in results |
524 |
return true; |
return true; |
525 |
} |
} |
526 |
|
|
527 |
|
if (NumberOfCapturingGroups() < n) { |
528 |
|
// RE has fewer capturing groups than number of arg pointers passed in |
529 |
|
return false; |
530 |
|
} |
531 |
|
|
532 |
// If we got here, we must have matched the whole pattern. |
// If we got here, we must have matched the whole pattern. |
533 |
// 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 |
534 |
// -- see the comment for TryMatch. |
// -- see the comment for TryMatch. |
592 |
|
|
593 |
// Return the number of capturing subpatterns, or -1 if the |
// Return the number of capturing subpatterns, or -1 if the |
594 |
// regexp wasn't valid on construction. |
// regexp wasn't valid on construction. |
595 |
int RE::NumberOfCapturingGroups() { |
int RE::NumberOfCapturingGroups() const { |
596 |
if (re_partial_ == NULL) return -1; |
if (re_partial_ == NULL) return -1; |
597 |
|
|
598 |
int result; |
int result; |
688 |
if (n == 0) return false; |
if (n == 0) return false; |
689 |
char buf[kMaxNumberLength+1]; |
char buf[kMaxNumberLength+1]; |
690 |
str = TerminateNumber(buf, str, n); |
str = TerminateNumber(buf, str, n); |
691 |
|
if (str[0] == '-') return false; // strtoul() on a negative number?! |
692 |
char* end; |
char* end; |
693 |
errno = 0; |
errno = 0; |
694 |
unsigned long r = strtoul(str, &end, radix); |
unsigned long r = strtoul(str, &end, radix); |
778 |
if (n == 0) return false; |
if (n == 0) return false; |
779 |
char buf[kMaxNumberLength+1]; |
char buf[kMaxNumberLength+1]; |
780 |
str = TerminateNumber(buf, str, n); |
str = TerminateNumber(buf, str, n); |
781 |
|
if (str[0] == '-') return false; // strtoull() on a negative number?! |
782 |
char* end; |
char* end; |
783 |
errno = 0; |
errno = 0; |
784 |
#if defined HAVE_STRTOQ |
#if defined HAVE_STRTOQ |
833 |
return parse_##name##_radix(str, n, dest, 0); \ |
return parse_##name##_radix(str, n, dest, 0); \ |
834 |
} |
} |
835 |
|
|
836 |
DEFINE_INTEGER_PARSERS(short); |
DEFINE_INTEGER_PARSERS(short) /* */ |
837 |
DEFINE_INTEGER_PARSERS(ushort); |
DEFINE_INTEGER_PARSERS(ushort) /* */ |
838 |
DEFINE_INTEGER_PARSERS(int); |
DEFINE_INTEGER_PARSERS(int) /* Don't use semicolons after these */ |
839 |
DEFINE_INTEGER_PARSERS(uint); |
DEFINE_INTEGER_PARSERS(uint) /* statements because they can cause */ |
840 |
DEFINE_INTEGER_PARSERS(long); |
DEFINE_INTEGER_PARSERS(long) /* compiler warnings if the checking */ |
841 |
DEFINE_INTEGER_PARSERS(ulong); |
DEFINE_INTEGER_PARSERS(ulong) /* level is turned up high enough. */ |
842 |
DEFINE_INTEGER_PARSERS(longlong); |
DEFINE_INTEGER_PARSERS(longlong) /* */ |
843 |
DEFINE_INTEGER_PARSERS(ulonglong); |
DEFINE_INTEGER_PARSERS(ulonglong) /* */ |
844 |
|
|
845 |
#undef DEFINE_INTEGER_PARSERS |
#undef DEFINE_INTEGER_PARSERS |
846 |
|
|