1 | // Copyright 2003-2009 The RE2 Authors. All Rights Reserved. |
2 | // Use of this source code is governed by a BSD-style |
3 | // license that can be found in the LICENSE file. |
4 | |
5 | #ifndef RE2_RE2_H_ |
6 | #define RE2_RE2_H_ |
7 | |
8 | // C++ interface to the re2 regular-expression library. |
9 | // RE2 supports Perl-style regular expressions (with extensions like |
10 | // \d, \w, \s, ...). |
11 | // |
12 | // ----------------------------------------------------------------------- |
13 | // REGEXP SYNTAX: |
14 | // |
15 | // This module uses the re2 library and hence supports |
16 | // its syntax for regular expressions, which is similar to Perl's with |
17 | // some of the more complicated things thrown away. In particular, |
18 | // backreferences and generalized assertions are not available, nor is \Z. |
19 | // |
20 | // See https://github.com/google/re2/wiki/Syntax for the syntax |
21 | // supported by RE2, and a comparison with PCRE and PERL regexps. |
22 | // |
23 | // For those not familiar with Perl's regular expressions, |
24 | // here are some examples of the most commonly used extensions: |
25 | // |
26 | // "hello (\\w+) world" -- \w matches a "word" character |
27 | // "version (\\d+)" -- \d matches a digit |
28 | // "hello\\s+world" -- \s matches any whitespace character |
29 | // "\\b(\\w+)\\b" -- \b matches non-empty string at word boundary |
30 | // "(?i)hello" -- (?i) turns on case-insensitive matching |
31 | // "/\\*(.*?)\\*/" -- .*? matches . minimum no. of times possible |
32 | // |
33 | // ----------------------------------------------------------------------- |
34 | // MATCHING INTERFACE: |
35 | // |
36 | // The "FullMatch" operation checks that supplied text matches a |
37 | // supplied pattern exactly. |
38 | // |
39 | // Example: successful match |
40 | // CHECK(RE2::FullMatch("hello", "h.*o")); |
41 | // |
42 | // Example: unsuccessful match (requires full match): |
43 | // CHECK(!RE2::FullMatch("hello", "e")); |
44 | // |
45 | // ----------------------------------------------------------------------- |
46 | // UTF-8 AND THE MATCHING INTERFACE: |
47 | // |
48 | // By default, the pattern and input text are interpreted as UTF-8. |
49 | // The RE2::Latin1 option causes them to be interpreted as Latin-1. |
50 | // |
51 | // Example: |
52 | // CHECK(RE2::FullMatch(utf8_string, RE2(utf8_pattern))); |
53 | // CHECK(RE2::FullMatch(latin1_string, RE2(latin1_pattern, RE2::Latin1))); |
54 | // |
55 | // ----------------------------------------------------------------------- |
56 | // MATCHING WITH SUBSTRING EXTRACTION: |
57 | // |
58 | // You can supply extra pointer arguments to extract matched substrings. |
59 | // On match failure, none of the pointees will have been modified. |
60 | // On match success, the substrings will be converted (as necessary) and |
61 | // their values will be assigned to their pointees until all conversions |
62 | // have succeeded or one conversion has failed. |
63 | // On conversion failure, the pointees will be in an indeterminate state |
64 | // because the caller has no way of knowing which conversion failed. |
65 | // However, conversion cannot fail for types like string and StringPiece |
66 | // that do not inspect the substring contents. Hence, in the common case |
67 | // where all of the pointees are of such types, failure is always due to |
68 | // match failure and thus none of the pointees will have been modified. |
69 | // |
70 | // Example: extracts "ruby" into "s" and 1234 into "i" |
71 | // int i; |
72 | // std::string s; |
73 | // CHECK(RE2::FullMatch("ruby:1234", "(\\w+):(\\d+)", &s, &i)); |
74 | // |
75 | // Example: fails because string cannot be stored in integer |
76 | // CHECK(!RE2::FullMatch("ruby", "(.*)", &i)); |
77 | // |
78 | // Example: fails because there aren't enough sub-patterns |
79 | // CHECK(!RE2::FullMatch("ruby:1234", "\\w+:\\d+", &s)); |
80 | // |
81 | // Example: does not try to extract any extra sub-patterns |
82 | // CHECK(RE2::FullMatch("ruby:1234", "(\\w+):(\\d+)", &s)); |
83 | // |
84 | // Example: does not try to extract into NULL |
85 | // CHECK(RE2::FullMatch("ruby:1234", "(\\w+):(\\d+)", NULL, &i)); |
86 | // |
87 | // Example: integer overflow causes failure |
88 | // CHECK(!RE2::FullMatch("ruby:1234567891234", "\\w+:(\\d+)", &i)); |
89 | // |
90 | // NOTE(rsc): Asking for substrings slows successful matches quite a bit. |
91 | // This may get a little faster in the future, but right now is slower |
92 | // than PCRE. On the other hand, failed matches run *very* fast (faster |
93 | // than PCRE), as do matches without substring extraction. |
94 | // |
95 | // ----------------------------------------------------------------------- |
96 | // PARTIAL MATCHES |
97 | // |
98 | // You can use the "PartialMatch" operation when you want the pattern |
99 | // to match any substring of the text. |
100 | // |
101 | // Example: simple search for a string: |
102 | // CHECK(RE2::PartialMatch("hello", "ell")); |
103 | // |
104 | // Example: find first number in a string |
105 | // int number; |
106 | // CHECK(RE2::PartialMatch("x*100 + 20", "(\\d+)", &number)); |
107 | // CHECK_EQ(number, 100); |
108 | // |
109 | // ----------------------------------------------------------------------- |
110 | // PRE-COMPILED REGULAR EXPRESSIONS |
111 | // |
112 | // RE2 makes it easy to use any string as a regular expression, without |
113 | // requiring a separate compilation step. |
114 | // |
115 | // If speed is of the essence, you can create a pre-compiled "RE2" |
116 | // object from the pattern and use it multiple times. If you do so, |
117 | // you can typically parse text faster than with sscanf. |
118 | // |
119 | // Example: precompile pattern for faster matching: |
120 | // RE2 pattern("h.*o"); |
121 | // while (ReadLine(&str)) { |
122 | // if (RE2::FullMatch(str, pattern)) ...; |
123 | // } |
124 | // |
125 | // ----------------------------------------------------------------------- |
126 | // SCANNING TEXT INCREMENTALLY |
127 | // |
128 | // The "Consume" operation may be useful if you want to repeatedly |
129 | // match regular expressions at the front of a string and skip over |
130 | // them as they match. This requires use of the "StringPiece" type, |
131 | // which represents a sub-range of a real string. |
132 | // |
133 | // Example: read lines of the form "var = value" from a string. |
134 | // std::string contents = ...; // Fill string somehow |
135 | // StringPiece input(contents); // Wrap a StringPiece around it |
136 | // |
137 | // std::string var; |
138 | // int value; |
139 | // while (RE2::Consume(&input, "(\\w+) = (\\d+)\n", &var, &value)) { |
140 | // ...; |
141 | // } |
142 | // |
143 | // Each successful call to "Consume" will set "var/value", and also |
144 | // advance "input" so it points past the matched text. Note that if the |
145 | // regular expression matches an empty string, input will advance |
146 | // by 0 bytes. If the regular expression being used might match |
147 | // an empty string, the loop body must check for this case and either |
148 | // advance the string or break out of the loop. |
149 | // |
150 | // The "FindAndConsume" operation is similar to "Consume" but does not |
151 | // anchor your match at the beginning of the string. For example, you |
152 | // could extract all words from a string by repeatedly calling |
153 | // RE2::FindAndConsume(&input, "(\\w+)", &word) |
154 | // |
155 | // ----------------------------------------------------------------------- |
156 | // USING VARIABLE NUMBER OF ARGUMENTS |
157 | // |
158 | // The above operations require you to know the number of arguments |
159 | // when you write the code. This is not always possible or easy (for |
160 | // example, the regular expression may be calculated at run time). |
161 | // You can use the "N" version of the operations when the number of |
162 | // match arguments are determined at run time. |
163 | // |
164 | // Example: |
165 | // const RE2::Arg* args[10]; |
166 | // int n; |
167 | // // ... populate args with pointers to RE2::Arg values ... |
168 | // // ... set n to the number of RE2::Arg objects ... |
169 | // bool match = RE2::FullMatchN(input, pattern, args, n); |
170 | // |
171 | // The last statement is equivalent to |
172 | // |
173 | // bool match = RE2::FullMatch(input, pattern, |
174 | // *args[0], *args[1], ..., *args[n - 1]); |
175 | // |
176 | // ----------------------------------------------------------------------- |
177 | // PARSING HEX/OCTAL/C-RADIX NUMBERS |
178 | // |
179 | // By default, if you pass a pointer to a numeric value, the |
180 | // corresponding text is interpreted as a base-10 number. You can |
181 | // instead wrap the pointer with a call to one of the operators Hex(), |
182 | // Octal(), or CRadix() to interpret the text in another base. The |
183 | // CRadix operator interprets C-style "0" (base-8) and "0x" (base-16) |
184 | // prefixes, but defaults to base-10. |
185 | // |
186 | // Example: |
187 | // int a, b, c, d; |
188 | // CHECK(RE2::FullMatch("100 40 0100 0x40", "(.*) (.*) (.*) (.*)", |
189 | // RE2::Octal(&a), RE2::Hex(&b), RE2::CRadix(&c), RE2::CRadix(&d)); |
190 | // will leave 64 in a, b, c, and d. |
191 | |
192 | #include <stddef.h> |
193 | #include <stdint.h> |
194 | #include <algorithm> |
195 | #include <map> |
196 | #include <mutex> |
197 | #include <string> |
198 | |
199 | #include "re2/stringpiece.h" |
200 | |
201 | namespace re2 { |
202 | class Prog; |
203 | class Regexp; |
204 | } // namespace re2 |
205 | |
206 | namespace re2 { |
207 | |
208 | // Interface for regular expression matching. Also corresponds to a |
209 | // pre-compiled regular expression. An "RE2" object is safe for |
210 | // concurrent use by multiple threads. |
211 | class RE2 { |
212 | public: |
213 | // We convert user-passed pointers into special Arg objects |
214 | class Arg; |
215 | class Options; |
216 | |
217 | // Defined in set.h. |
218 | class Set; |
219 | |
220 | enum ErrorCode { |
221 | NoError = 0, |
222 | |
223 | // Unexpected error |
224 | ErrorInternal, |
225 | |
226 | // Parse errors |
227 | ErrorBadEscape, // bad escape sequence |
228 | ErrorBadCharClass, // bad character class |
229 | ErrorBadCharRange, // bad character class range |
230 | ErrorMissingBracket, // missing closing ] |
231 | ErrorMissingParen, // missing closing ) |
232 | ErrorTrailingBackslash, // trailing \ at end of regexp |
233 | ErrorRepeatArgument, // repeat argument missing, e.g. "*" |
234 | ErrorRepeatSize, // bad repetition argument |
235 | ErrorRepeatOp, // bad repetition operator |
236 | ErrorBadPerlOp, // bad perl operator |
237 | ErrorBadUTF8, // invalid UTF-8 in regexp |
238 | ErrorBadNamedCapture, // bad named capture group |
239 | ErrorPatternTooLarge // pattern too large (compile failed) |
240 | }; |
241 | |
242 | // Predefined common options. |
243 | // If you need more complicated things, instantiate |
244 | // an Option class, possibly passing one of these to |
245 | // the Option constructor, change the settings, and pass that |
246 | // Option class to the RE2 constructor. |
247 | enum CannedOptions { |
248 | DefaultOptions = 0, |
249 | Latin1, // treat input as Latin-1 (default UTF-8) |
250 | POSIX, // POSIX syntax, leftmost-longest match |
251 | Quiet // do not log about regexp parse errors |
252 | }; |
253 | |
254 | // Need to have the const char* and const std::string& forms for implicit |
255 | // conversions when passing string literals to FullMatch and PartialMatch. |
256 | // Otherwise the StringPiece form would be sufficient. |
257 | #ifndef SWIG |
258 | RE2(const char* pattern); |
259 | RE2(const std::string& pattern); |
260 | #endif |
261 | RE2(const StringPiece& pattern); |
262 | RE2(const StringPiece& pattern, const Options& options); |
263 | ~RE2(); |
264 | |
265 | // Returns whether RE2 was created properly. |
266 | bool ok() const { return error_code() == NoError; } |
267 | |
268 | // The string specification for this RE2. E.g. |
269 | // RE2 re("ab*c?d+"); |
270 | // re.pattern(); // "ab*c?d+" |
271 | const std::string& pattern() const { return pattern_; } |
272 | |
273 | // If RE2 could not be created properly, returns an error string. |
274 | // Else returns the empty string. |
275 | const std::string& error() const { return *error_; } |
276 | |
277 | // If RE2 could not be created properly, returns an error code. |
278 | // Else returns RE2::NoError (== 0). |
279 | ErrorCode error_code() const { return error_code_; } |
280 | |
281 | // If RE2 could not be created properly, returns the offending |
282 | // portion of the regexp. |
283 | const std::string& error_arg() const { return error_arg_; } |
284 | |
285 | // Returns the program size, a very approximate measure of a regexp's "cost". |
286 | // Larger numbers are more expensive than smaller numbers. |
287 | int ProgramSize() const; |
288 | int ReverseProgramSize() const; |
289 | |
290 | // EXPERIMENTAL! SUBJECT TO CHANGE! |
291 | // Outputs the program fanout as a histogram bucketed by powers of 2. |
292 | // Returns the number of the largest non-empty bucket. |
293 | int ProgramFanout(std::map<int, int>* histogram) const; |
294 | int ReverseProgramFanout(std::map<int, int>* histogram) const; |
295 | |
296 | // Returns the underlying Regexp; not for general use. |
297 | // Returns entire_regexp_ so that callers don't need |
298 | // to know about prefix_ and prefix_foldcase_. |
299 | re2::Regexp* Regexp() const { return entire_regexp_; } |
300 | |
301 | /***** The array-based matching interface ******/ |
302 | |
303 | // The functions here have names ending in 'N' and are used to implement |
304 | // the functions whose names are the prefix before the 'N'. It is sometimes |
305 | // useful to invoke them directly, but the syntax is awkward, so the 'N'-less |
306 | // versions should be preferred. |
307 | static bool FullMatchN(const StringPiece& text, const RE2& re, |
308 | const Arg* const args[], int n); |
309 | static bool PartialMatchN(const StringPiece& text, const RE2& re, |
310 | const Arg* const args[], int n); |
311 | static bool ConsumeN(StringPiece* input, const RE2& re, |
312 | const Arg* const args[], int n); |
313 | static bool FindAndConsumeN(StringPiece* input, const RE2& re, |
314 | const Arg* const args[], int n); |
315 | |
316 | #ifndef SWIG |
317 | private: |
318 | template <typename F, typename SP> |
319 | static inline bool Apply(F f, SP sp, const RE2& re) { |
320 | return f(sp, re, NULL, 0); |
321 | } |
322 | |
323 | template <typename F, typename SP, typename... A> |
324 | static inline bool Apply(F f, SP sp, const RE2& re, const A&... a) { |
325 | const Arg* const args[] = {&a...}; |
326 | const int n = sizeof...(a); |
327 | return f(sp, re, args, n); |
328 | } |
329 | |
330 | public: |
331 | // In order to allow FullMatch() et al. to be called with a varying number |
332 | // of arguments of varying types, we use two layers of variadic templates. |
333 | // The first layer constructs the temporary Arg objects. The second layer |
334 | // (above) constructs the array of pointers to the temporary Arg objects. |
335 | |
336 | /***** The useful part: the matching interface *****/ |
337 | |
338 | // Matches "text" against "re". If pointer arguments are |
339 | // supplied, copies matched sub-patterns into them. |
340 | // |
341 | // You can pass in a "const char*" or a "std::string" for "text". |
342 | // You can pass in a "const char*" or a "std::string" or a "RE2" for "re". |
343 | // |
344 | // The provided pointer arguments can be pointers to any scalar numeric |
345 | // type, or one of: |
346 | // std::string (matched piece is copied to string) |
347 | // StringPiece (StringPiece is mutated to point to matched piece) |
348 | // T (where "bool T::ParseFrom(const char*, size_t)" exists) |
349 | // (void*)NULL (the corresponding matched sub-pattern is not copied) |
350 | // |
351 | // Returns true iff all of the following conditions are satisfied: |
352 | // a. "text" matches "re" exactly |
353 | // b. The number of matched sub-patterns is >= number of supplied pointers |
354 | // c. The "i"th argument has a suitable type for holding the |
355 | // string captured as the "i"th sub-pattern. If you pass in |
356 | // NULL for the "i"th argument, or pass fewer arguments than |
357 | // number of sub-patterns, "i"th captured sub-pattern is |
358 | // ignored. |
359 | // |
360 | // CAVEAT: An optional sub-pattern that does not exist in the |
361 | // matched string is assigned the empty string. Therefore, the |
362 | // following will return false (because the empty string is not a |
363 | // valid number): |
364 | // int number; |
365 | // RE2::FullMatch("abc", "[a-z]+(\\d+)?", &number); |
366 | template <typename... A> |
367 | static bool FullMatch(const StringPiece& text, const RE2& re, A&&... a) { |
368 | return Apply(FullMatchN, text, re, Arg(std::forward<A>(a))...); |
369 | } |
370 | |
371 | // Exactly like FullMatch(), except that "re" is allowed to match |
372 | // a substring of "text". |
373 | template <typename... A> |
374 | static bool PartialMatch(const StringPiece& text, const RE2& re, A&&... a) { |
375 | return Apply(PartialMatchN, text, re, Arg(std::forward<A>(a))...); |
376 | } |
377 | |
378 | // Like FullMatch() and PartialMatch(), except that "re" has to match |
379 | // a prefix of the text, and "input" is advanced past the matched |
380 | // text. Note: "input" is modified iff this routine returns true |
381 | // and "re" matched a non-empty substring of "text". |
382 | template <typename... A> |
383 | static bool Consume(StringPiece* input, const RE2& re, A&&... a) { |
384 | return Apply(ConsumeN, input, re, Arg(std::forward<A>(a))...); |
385 | } |
386 | |
387 | // Like Consume(), but does not anchor the match at the beginning of |
388 | // the text. That is, "re" need not start its match at the beginning |
389 | // of "input". For example, "FindAndConsume(s, "(\\w+)", &word)" finds |
390 | // the next word in "s" and stores it in "word". |
391 | template <typename... A> |
392 | static bool FindAndConsume(StringPiece* input, const RE2& re, A&&... a) { |
393 | return Apply(FindAndConsumeN, input, re, Arg(std::forward<A>(a))...); |
394 | } |
395 | #endif |
396 | |
397 | // Replace the first match of "re" in "str" with "rewrite". |
398 | // Within "rewrite", backslash-escaped digits (\1 to \9) can be |
399 | // used to insert text matching corresponding parenthesized group |
400 | // from the pattern. \0 in "rewrite" refers to the entire matching |
401 | // text. E.g., |
402 | // |
403 | // std::string s = "yabba dabba doo"; |
404 | // CHECK(RE2::Replace(&s, "b+", "d")); |
405 | // |
406 | // will leave "s" containing "yada dabba doo" |
407 | // |
408 | // Returns true if the pattern matches and a replacement occurs, |
409 | // false otherwise. |
410 | static bool Replace(std::string* str, |
411 | const RE2& re, |
412 | const StringPiece& rewrite); |
413 | |
414 | // Like Replace(), except replaces successive non-overlapping occurrences |
415 | // of the pattern in the string with the rewrite. E.g. |
416 | // |
417 | // std::string s = "yabba dabba doo"; |
418 | // CHECK(RE2::GlobalReplace(&s, "b+", "d")); |
419 | // |
420 | // will leave "s" containing "yada dada doo" |
421 | // Replacements are not subject to re-matching. |
422 | // |
423 | // Because GlobalReplace only replaces non-overlapping matches, |
424 | // replacing "ana" within "banana" makes only one replacement, not two. |
425 | // |
426 | // Returns the number of replacements made. |
427 | static int GlobalReplace(std::string* str, |
428 | const RE2& re, |
429 | const StringPiece& rewrite); |
430 | |
431 | // Like Replace, except that if the pattern matches, "rewrite" |
432 | // is copied into "out" with substitutions. The non-matching |
433 | // portions of "text" are ignored. |
434 | // |
435 | // Returns true iff a match occurred and the extraction happened |
436 | // successfully; if no match occurs, the string is left unaffected. |
437 | // |
438 | // REQUIRES: "text" must not alias any part of "*out". |
439 | static bool (const StringPiece& text, |
440 | const RE2& re, |
441 | const StringPiece& rewrite, |
442 | std::string* out); |
443 | |
444 | // Escapes all potentially meaningful regexp characters in |
445 | // 'unquoted'. The returned string, used as a regular expression, |
446 | // will exactly match the original string. For example, |
447 | // 1.5-2.0? |
448 | // may become: |
449 | // 1\.5\-2\.0\? |
450 | static std::string QuoteMeta(const StringPiece& unquoted); |
451 | |
452 | // Computes range for any strings matching regexp. The min and max can in |
453 | // some cases be arbitrarily precise, so the caller gets to specify the |
454 | // maximum desired length of string returned. |
455 | // |
456 | // Assuming PossibleMatchRange(&min, &max, N) returns successfully, any |
457 | // string s that is an anchored match for this regexp satisfies |
458 | // min <= s && s <= max. |
459 | // |
460 | // Note that PossibleMatchRange() will only consider the first copy of an |
461 | // infinitely repeated element (i.e., any regexp element followed by a '*' or |
462 | // '+' operator). Regexps with "{N}" constructions are not affected, as those |
463 | // do not compile down to infinite repetitions. |
464 | // |
465 | // Returns true on success, false on error. |
466 | bool PossibleMatchRange(std::string* min, std::string* max, |
467 | int maxlen) const; |
468 | |
469 | // Generic matching interface |
470 | |
471 | // Type of match. |
472 | enum Anchor { |
473 | UNANCHORED, // No anchoring |
474 | ANCHOR_START, // Anchor at start only |
475 | ANCHOR_BOTH // Anchor at start and end |
476 | }; |
477 | |
478 | // Return the number of capturing subpatterns, or -1 if the |
479 | // regexp wasn't valid on construction. The overall match ($0) |
480 | // does not count: if the regexp is "(a)(b)", returns 2. |
481 | int NumberOfCapturingGroups() const { return num_captures_; } |
482 | |
483 | // Return a map from names to capturing indices. |
484 | // The map records the index of the leftmost group |
485 | // with the given name. |
486 | // Only valid until the re is deleted. |
487 | const std::map<std::string, int>& NamedCapturingGroups() const; |
488 | |
489 | // Return a map from capturing indices to names. |
490 | // The map has no entries for unnamed groups. |
491 | // Only valid until the re is deleted. |
492 | const std::map<int, std::string>& CapturingGroupNames() const; |
493 | |
494 | // General matching routine. |
495 | // Match against text starting at offset startpos |
496 | // and stopping the search at offset endpos. |
497 | // Returns true if match found, false if not. |
498 | // On a successful match, fills in submatch[] (up to nsubmatch entries) |
499 | // with information about submatches. |
500 | // I.e. matching RE2("(foo)|(bar)baz") on "barbazbla" will return true, with |
501 | // submatch[0] = "barbaz", submatch[1].data() = NULL, submatch[2] = "bar", |
502 | // submatch[3].data() = NULL, ..., up to submatch[nsubmatch-1].data() = NULL. |
503 | // Caveat: submatch[] may be clobbered even on match failure. |
504 | // |
505 | // Don't ask for more match information than you will use: |
506 | // runs much faster with nsubmatch == 1 than nsubmatch > 1, and |
507 | // runs even faster if nsubmatch == 0. |
508 | // Doesn't make sense to use nsubmatch > 1 + NumberOfCapturingGroups(), |
509 | // but will be handled correctly. |
510 | // |
511 | // Passing text == StringPiece(NULL, 0) will be handled like any other |
512 | // empty string, but note that on return, it will not be possible to tell |
513 | // whether submatch i matched the empty string or did not match: |
514 | // either way, submatch[i].data() == NULL. |
515 | bool Match(const StringPiece& text, |
516 | size_t startpos, |
517 | size_t endpos, |
518 | Anchor re_anchor, |
519 | StringPiece* submatch, |
520 | int nsubmatch) const; |
521 | |
522 | // Check that the given rewrite string is suitable for use with this |
523 | // regular expression. It checks that: |
524 | // * The regular expression has enough parenthesized subexpressions |
525 | // to satisfy all of the \N tokens in rewrite |
526 | // * The rewrite string doesn't have any syntax errors. E.g., |
527 | // '\' followed by anything other than a digit or '\'. |
528 | // A true return value guarantees that Replace() and Extract() won't |
529 | // fail because of a bad rewrite string. |
530 | bool CheckRewriteString(const StringPiece& rewrite, |
531 | std::string* error) const; |
532 | |
533 | // Returns the maximum submatch needed for the rewrite to be done by |
534 | // Replace(). E.g. if rewrite == "foo \\2,\\1", returns 2. |
535 | static int MaxSubmatch(const StringPiece& rewrite); |
536 | |
537 | // Append the "rewrite" string, with backslash subsitutions from "vec", |
538 | // to string "out". |
539 | // Returns true on success. This method can fail because of a malformed |
540 | // rewrite string. CheckRewriteString guarantees that the rewrite will |
541 | // be sucessful. |
542 | bool Rewrite(std::string* out, |
543 | const StringPiece& rewrite, |
544 | const StringPiece* vec, |
545 | int veclen) const; |
546 | |
547 | // Constructor options |
548 | class Options { |
549 | public: |
550 | // The options are (defaults in parentheses): |
551 | // |
552 | // utf8 (true) text and pattern are UTF-8; otherwise Latin-1 |
553 | // posix_syntax (false) restrict regexps to POSIX egrep syntax |
554 | // longest_match (false) search for longest match, not first match |
555 | // log_errors (true) log syntax and execution errors to ERROR |
556 | // max_mem (see below) approx. max memory footprint of RE2 |
557 | // literal (false) interpret string as literal, not regexp |
558 | // never_nl (false) never match \n, even if it is in regexp |
559 | // dot_nl (false) dot matches everything including new line |
560 | // never_capture (false) parse all parens as non-capturing |
561 | // case_sensitive (true) match is case-sensitive (regexp can override |
562 | // with (?i) unless in posix_syntax mode) |
563 | // |
564 | // The following options are only consulted when posix_syntax == true. |
565 | // When posix_syntax == false, these features are always enabled and |
566 | // cannot be turned off; to perform multi-line matching in that case, |
567 | // begin the regexp with (?m). |
568 | // perl_classes (false) allow Perl's \d \s \w \D \S \W |
569 | // word_boundary (false) allow Perl's \b \B (word boundary and not) |
570 | // one_line (false) ^ and $ only match beginning and end of text |
571 | // |
572 | // The max_mem option controls how much memory can be used |
573 | // to hold the compiled form of the regexp (the Prog) and |
574 | // its cached DFA graphs. Code Search placed limits on the number |
575 | // of Prog instructions and DFA states: 10,000 for both. |
576 | // In RE2, those limits would translate to about 240 KB per Prog |
577 | // and perhaps 2.5 MB per DFA (DFA state sizes vary by regexp; RE2 does a |
578 | // better job of keeping them small than Code Search did). |
579 | // Each RE2 has two Progs (one forward, one reverse), and each Prog |
580 | // can have two DFAs (one first match, one longest match). |
581 | // That makes 4 DFAs: |
582 | // |
583 | // forward, first-match - used for UNANCHORED or ANCHOR_START searches |
584 | // if opt.longest_match() == false |
585 | // forward, longest-match - used for all ANCHOR_BOTH searches, |
586 | // and the other two kinds if |
587 | // opt.longest_match() == true |
588 | // reverse, first-match - never used |
589 | // reverse, longest-match - used as second phase for unanchored searches |
590 | // |
591 | // The RE2 memory budget is statically divided between the two |
592 | // Progs and then the DFAs: two thirds to the forward Prog |
593 | // and one third to the reverse Prog. The forward Prog gives half |
594 | // of what it has left over to each of its DFAs. The reverse Prog |
595 | // gives it all to its longest-match DFA. |
596 | // |
597 | // Once a DFA fills its budget, it flushes its cache and starts over. |
598 | // If this happens too often, RE2 falls back on the NFA implementation. |
599 | |
600 | // For now, make the default budget something close to Code Search. |
601 | static const int kDefaultMaxMem = 8<<20; |
602 | |
603 | enum Encoding { |
604 | EncodingUTF8 = 1, |
605 | EncodingLatin1 |
606 | }; |
607 | |
608 | Options() : |
609 | encoding_(EncodingUTF8), |
610 | posix_syntax_(false), |
611 | longest_match_(false), |
612 | log_errors_(true), |
613 | max_mem_(kDefaultMaxMem), |
614 | literal_(false), |
615 | never_nl_(false), |
616 | dot_nl_(false), |
617 | never_capture_(false), |
618 | case_sensitive_(true), |
619 | perl_classes_(false), |
620 | word_boundary_(false), |
621 | one_line_(false) { |
622 | } |
623 | |
624 | /*implicit*/ Options(CannedOptions); |
625 | |
626 | Encoding encoding() const { return encoding_; } |
627 | void set_encoding(Encoding encoding) { encoding_ = encoding; } |
628 | |
629 | // Legacy interface to encoding. |
630 | // TODO(rsc): Remove once clients have been converted. |
631 | bool utf8() const { return encoding_ == EncodingUTF8; } |
632 | void set_utf8(bool b) { |
633 | if (b) { |
634 | encoding_ = EncodingUTF8; |
635 | } else { |
636 | encoding_ = EncodingLatin1; |
637 | } |
638 | } |
639 | |
640 | bool posix_syntax() const { return posix_syntax_; } |
641 | void set_posix_syntax(bool b) { posix_syntax_ = b; } |
642 | |
643 | bool longest_match() const { return longest_match_; } |
644 | void set_longest_match(bool b) { longest_match_ = b; } |
645 | |
646 | bool log_errors() const { return log_errors_; } |
647 | void set_log_errors(bool b) { log_errors_ = b; } |
648 | |
649 | int64_t max_mem() const { return max_mem_; } |
650 | void set_max_mem(int64_t m) { max_mem_ = m; } |
651 | |
652 | bool literal() const { return literal_; } |
653 | void set_literal(bool b) { literal_ = b; } |
654 | |
655 | bool never_nl() const { return never_nl_; } |
656 | void set_never_nl(bool b) { never_nl_ = b; } |
657 | |
658 | bool dot_nl() const { return dot_nl_; } |
659 | void set_dot_nl(bool b) { dot_nl_ = b; } |
660 | |
661 | bool never_capture() const { return never_capture_; } |
662 | void set_never_capture(bool b) { never_capture_ = b; } |
663 | |
664 | bool case_sensitive() const { return case_sensitive_; } |
665 | void set_case_sensitive(bool b) { case_sensitive_ = b; } |
666 | |
667 | bool perl_classes() const { return perl_classes_; } |
668 | void set_perl_classes(bool b) { perl_classes_ = b; } |
669 | |
670 | bool word_boundary() const { return word_boundary_; } |
671 | void set_word_boundary(bool b) { word_boundary_ = b; } |
672 | |
673 | bool one_line() const { return one_line_; } |
674 | void set_one_line(bool b) { one_line_ = b; } |
675 | |
676 | void Copy(const Options& src) { |
677 | *this = src; |
678 | } |
679 | |
680 | int ParseFlags() const; |
681 | |
682 | private: |
683 | Encoding encoding_; |
684 | bool posix_syntax_; |
685 | bool longest_match_; |
686 | bool log_errors_; |
687 | int64_t max_mem_; |
688 | bool literal_; |
689 | bool never_nl_; |
690 | bool dot_nl_; |
691 | bool never_capture_; |
692 | bool case_sensitive_; |
693 | bool perl_classes_; |
694 | bool word_boundary_; |
695 | bool one_line_; |
696 | }; |
697 | |
698 | // Returns the options set in the constructor. |
699 | const Options& options() const { return options_; } |
700 | |
701 | // Argument converters; see below. |
702 | static inline Arg CRadix(short* x); |
703 | static inline Arg CRadix(unsigned short* x); |
704 | static inline Arg CRadix(int* x); |
705 | static inline Arg CRadix(unsigned int* x); |
706 | static inline Arg CRadix(long* x); |
707 | static inline Arg CRadix(unsigned long* x); |
708 | static inline Arg CRadix(long long* x); |
709 | static inline Arg CRadix(unsigned long long* x); |
710 | |
711 | static inline Arg Hex(short* x); |
712 | static inline Arg Hex(unsigned short* x); |
713 | static inline Arg Hex(int* x); |
714 | static inline Arg Hex(unsigned int* x); |
715 | static inline Arg Hex(long* x); |
716 | static inline Arg Hex(unsigned long* x); |
717 | static inline Arg Hex(long long* x); |
718 | static inline Arg Hex(unsigned long long* x); |
719 | |
720 | static inline Arg Octal(short* x); |
721 | static inline Arg Octal(unsigned short* x); |
722 | static inline Arg Octal(int* x); |
723 | static inline Arg Octal(unsigned int* x); |
724 | static inline Arg Octal(long* x); |
725 | static inline Arg Octal(unsigned long* x); |
726 | static inline Arg Octal(long long* x); |
727 | static inline Arg Octal(unsigned long long* x); |
728 | |
729 | private: |
730 | void Init(const StringPiece& pattern, const Options& options); |
731 | |
732 | bool DoMatch(const StringPiece& text, |
733 | Anchor re_anchor, |
734 | size_t* consumed, |
735 | const Arg* const args[], |
736 | int n) const; |
737 | |
738 | re2::Prog* ReverseProg() const; |
739 | |
740 | std::string pattern_; // string regular expression |
741 | Options options_; // option flags |
742 | std::string prefix_; // required prefix (before regexp_) |
743 | bool prefix_foldcase_; // prefix is ASCII case-insensitive |
744 | re2::Regexp* entire_regexp_; // parsed regular expression |
745 | re2::Regexp* suffix_regexp_; // parsed regular expression, prefix removed |
746 | re2::Prog* prog_; // compiled program for regexp |
747 | int num_captures_; // Number of capturing groups |
748 | bool is_one_pass_; // can use prog_->SearchOnePass? |
749 | |
750 | mutable re2::Prog* rprog_; // reverse program for regexp |
751 | mutable const std::string* error_; // Error indicator |
752 | // (or points to empty string) |
753 | mutable ErrorCode error_code_; // Error code |
754 | mutable std::string error_arg_; // Fragment of regexp showing error |
755 | |
756 | // Map from capture names to indices |
757 | mutable const std::map<std::string, int>* named_groups_; |
758 | |
759 | // Map from capture indices to names |
760 | mutable const std::map<int, std::string>* group_names_; |
761 | |
762 | // Onces for lazy computations. |
763 | mutable std::once_flag rprog_once_; |
764 | mutable std::once_flag named_groups_once_; |
765 | mutable std::once_flag group_names_once_; |
766 | |
767 | RE2(const RE2&) = delete; |
768 | RE2& operator=(const RE2&) = delete; |
769 | }; |
770 | |
771 | /***** Implementation details *****/ |
772 | |
773 | // Hex/Octal/Binary? |
774 | |
775 | // Special class for parsing into objects that define a ParseFrom() method |
776 | template <class T> |
777 | class _RE2_MatchObject { |
778 | public: |
779 | static inline bool Parse(const char* str, size_t n, void* dest) { |
780 | if (dest == NULL) return true; |
781 | T* object = reinterpret_cast<T*>(dest); |
782 | return object->ParseFrom(str, n); |
783 | } |
784 | }; |
785 | |
786 | class RE2::Arg { |
787 | public: |
788 | // Empty constructor so we can declare arrays of RE2::Arg |
789 | Arg(); |
790 | |
791 | // Constructor specially designed for NULL arguments |
792 | Arg(void*); |
793 | Arg(std::nullptr_t); |
794 | |
795 | typedef bool (*Parser)(const char* str, size_t n, void* dest); |
796 | |
797 | // Type-specific parsers |
798 | #define MAKE_PARSER(type, name) \ |
799 | Arg(type* p) : arg_(p), parser_(name) {} \ |
800 | Arg(type* p, Parser parser) : arg_(p), parser_(parser) {} |
801 | |
802 | MAKE_PARSER(char, parse_char) |
803 | MAKE_PARSER(signed char, parse_schar) |
804 | MAKE_PARSER(unsigned char, parse_uchar) |
805 | MAKE_PARSER(float, parse_float) |
806 | MAKE_PARSER(double, parse_double) |
807 | MAKE_PARSER(std::string, parse_string) |
808 | MAKE_PARSER(StringPiece, parse_stringpiece) |
809 | |
810 | MAKE_PARSER(short, parse_short) |
811 | MAKE_PARSER(unsigned short, parse_ushort) |
812 | MAKE_PARSER(int, parse_int) |
813 | MAKE_PARSER(unsigned int, parse_uint) |
814 | MAKE_PARSER(long, parse_long) |
815 | MAKE_PARSER(unsigned long, parse_ulong) |
816 | MAKE_PARSER(long long, parse_longlong) |
817 | MAKE_PARSER(unsigned long long, parse_ulonglong) |
818 | |
819 | #undef MAKE_PARSER |
820 | |
821 | // Generic constructor templates |
822 | template <class T> Arg(T* p) |
823 | : arg_(p), parser_(_RE2_MatchObject<T>::Parse) { } |
824 | template <class T> Arg(T* p, Parser parser) |
825 | : arg_(p), parser_(parser) { } |
826 | |
827 | // Parse the data |
828 | bool Parse(const char* str, size_t n) const; |
829 | |
830 | private: |
831 | void* arg_; |
832 | Parser parser_; |
833 | |
834 | static bool parse_null (const char* str, size_t n, void* dest); |
835 | static bool parse_char (const char* str, size_t n, void* dest); |
836 | static bool parse_schar (const char* str, size_t n, void* dest); |
837 | static bool parse_uchar (const char* str, size_t n, void* dest); |
838 | static bool parse_float (const char* str, size_t n, void* dest); |
839 | static bool parse_double (const char* str, size_t n, void* dest); |
840 | static bool parse_string (const char* str, size_t n, void* dest); |
841 | static bool parse_stringpiece (const char* str, size_t n, void* dest); |
842 | |
843 | #define DECLARE_INTEGER_PARSER(name) \ |
844 | private: \ |
845 | static bool parse_##name(const char* str, size_t n, void* dest); \ |
846 | static bool parse_##name##_radix(const char* str, size_t n, void* dest, \ |
847 | int radix); \ |
848 | \ |
849 | public: \ |
850 | static bool parse_##name##_hex(const char* str, size_t n, void* dest); \ |
851 | static bool parse_##name##_octal(const char* str, size_t n, void* dest); \ |
852 | static bool parse_##name##_cradix(const char* str, size_t n, void* dest); |
853 | |
854 | DECLARE_INTEGER_PARSER(short) |
855 | DECLARE_INTEGER_PARSER(ushort) |
856 | DECLARE_INTEGER_PARSER(int) |
857 | DECLARE_INTEGER_PARSER(uint) |
858 | DECLARE_INTEGER_PARSER(long) |
859 | DECLARE_INTEGER_PARSER(ulong) |
860 | DECLARE_INTEGER_PARSER(longlong) |
861 | DECLARE_INTEGER_PARSER(ulonglong) |
862 | |
863 | #undef DECLARE_INTEGER_PARSER |
864 | |
865 | }; |
866 | |
867 | inline RE2::Arg::Arg() : arg_(NULL), parser_(parse_null) { } |
868 | inline RE2::Arg::Arg(void* p) : arg_(p), parser_(parse_null) { } |
869 | inline RE2::Arg::Arg(std::nullptr_t p) : arg_(p), parser_(parse_null) { } |
870 | |
871 | inline bool RE2::Arg::Parse(const char* str, size_t n) const { |
872 | return (*parser_)(str, n, arg_); |
873 | } |
874 | |
875 | // This part of the parser, appropriate only for ints, deals with bases |
876 | #define MAKE_INTEGER_PARSER(type, name) \ |
877 | inline RE2::Arg RE2::Hex(type* ptr) { \ |
878 | return RE2::Arg(ptr, RE2::Arg::parse_##name##_hex); \ |
879 | } \ |
880 | inline RE2::Arg RE2::Octal(type* ptr) { \ |
881 | return RE2::Arg(ptr, RE2::Arg::parse_##name##_octal); \ |
882 | } \ |
883 | inline RE2::Arg RE2::CRadix(type* ptr) { \ |
884 | return RE2::Arg(ptr, RE2::Arg::parse_##name##_cradix); \ |
885 | } |
886 | |
887 | MAKE_INTEGER_PARSER(short, short) |
888 | MAKE_INTEGER_PARSER(unsigned short, ushort) |
889 | MAKE_INTEGER_PARSER(int, int) |
890 | MAKE_INTEGER_PARSER(unsigned int, uint) |
891 | MAKE_INTEGER_PARSER(long, long) |
892 | MAKE_INTEGER_PARSER(unsigned long, ulong) |
893 | MAKE_INTEGER_PARSER(long long, longlong) |
894 | MAKE_INTEGER_PARSER(unsigned long long, ulonglong) |
895 | |
896 | #undef MAKE_INTEGER_PARSER |
897 | |
898 | #ifndef SWIG |
899 | |
900 | // Silence warnings about missing initializers for members of LazyRE2. |
901 | // Note that we test for Clang first because it defines __GNUC__ as well. |
902 | #if defined(__clang__) |
903 | #elif defined(__GNUC__) && __GNUC__ >= 6 |
904 | #pragma GCC diagnostic ignored "-Wmissing-field-initializers" |
905 | #endif |
906 | |
907 | // Helper for writing global or static RE2s safely. |
908 | // Write |
909 | // static LazyRE2 re = {".*"}; |
910 | // and then use *re instead of writing |
911 | // static RE2 re(".*"); |
912 | // The former is more careful about multithreaded |
913 | // situations than the latter. |
914 | // |
915 | // N.B. This class never deletes the RE2 object that |
916 | // it constructs: that's a feature, so that it can be used |
917 | // for global and function static variables. |
918 | class LazyRE2 { |
919 | private: |
920 | struct NoArg {}; |
921 | |
922 | public: |
923 | typedef RE2 element_type; // support std::pointer_traits |
924 | |
925 | // Constructor omitted to preserve braced initialization in C++98. |
926 | |
927 | // Pretend to be a pointer to Type (never NULL due to on-demand creation): |
928 | RE2& operator*() const { return *get(); } |
929 | RE2* operator->() const { return get(); } |
930 | |
931 | // Named accessor/initializer: |
932 | RE2* get() const { |
933 | std::call_once(once_, &LazyRE2::Init, this); |
934 | return ptr_; |
935 | } |
936 | |
937 | // All data fields must be public to support {"foo"} initialization. |
938 | const char* pattern_; |
939 | RE2::CannedOptions options_; |
940 | NoArg barrier_against_excess_initializers_; |
941 | |
942 | mutable RE2* ptr_; |
943 | mutable std::once_flag once_; |
944 | |
945 | private: |
946 | static void Init(const LazyRE2* lazy_re2) { |
947 | lazy_re2->ptr_ = new RE2(lazy_re2->pattern_, lazy_re2->options_); |
948 | } |
949 | |
950 | void operator=(const LazyRE2&); // disallowed |
951 | }; |
952 | #endif // SWIG |
953 | |
954 | } // namespace re2 |
955 | |
956 | using re2::RE2; |
957 | using re2::LazyRE2; |
958 | |
959 | #endif // RE2_RE2_H_ |
960 | |