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