Ian Hodson | 2ee91b4 | 2012-05-14 12:29:36 +0100 | [diff] [blame] | 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 http://code.google.com/p/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 | |
| 183 | #include <stdint.h> |
| 184 | #include <map> |
| 185 | #include <string> |
| 186 | #include "re2/stringpiece.h" |
| 187 | #include "re2/variadic_function.h" |
| 188 | |
| 189 | namespace re2 { |
Alexander Gutkin | 0d4c523 | 2013-02-28 13:47:27 +0000 | [diff] [blame] | 190 | |
Ian Hodson | 2ee91b4 | 2012-05-14 12:29:36 +0100 | [diff] [blame] | 191 | using std::string; |
| 192 | using std::map; |
| 193 | class Mutex; |
| 194 | class Prog; |
| 195 | class Regexp; |
| 196 | |
Alexander Gutkin | 0d4c523 | 2013-02-28 13:47:27 +0000 | [diff] [blame] | 197 | // The following enum should be used only as a constructor argument to indicate |
| 198 | // that the variable has static storage class, and that the constructor should |
| 199 | // do nothing to its state. It indicates to the reader that it is legal to |
| 200 | // declare a static instance of the class, provided the constructor is given |
| 201 | // the LINKER_INITIALIZED argument. Normally, it is unsafe to declare a |
| 202 | // static variable that has a constructor or a destructor because invocation |
| 203 | // order is undefined. However, IF the type can be initialized by filling with |
| 204 | // zeroes (which the loader does for static variables), AND the type's |
| 205 | // destructor does nothing to the storage, then a constructor for static |
| 206 | // initialization can be declared as |
| 207 | // explicit MyClass(LinkerInitialized x) {} |
| 208 | // and invoked as |
| 209 | // static MyClass my_variable_name(LINKER_INITIALIZED); |
| 210 | enum LinkerInitialized { LINKER_INITIALIZED }; |
| 211 | |
Ian Hodson | 2ee91b4 | 2012-05-14 12:29:36 +0100 | [diff] [blame] | 212 | // Interface for regular expression matching. Also corresponds to a |
| 213 | // pre-compiled regular expression. An "RE2" object is safe for |
| 214 | // concurrent use by multiple threads. |
| 215 | class RE2 { |
| 216 | public: |
| 217 | // We convert user-passed pointers into special Arg objects |
| 218 | class Arg; |
| 219 | class Options; |
| 220 | |
| 221 | // Defined in set.h. |
| 222 | class Set; |
| 223 | |
| 224 | enum ErrorCode { |
| 225 | NoError = 0, |
| 226 | |
| 227 | // Unexpected error |
| 228 | ErrorInternal, |
| 229 | |
| 230 | // Parse errors |
| 231 | ErrorBadEscape, // bad escape sequence |
| 232 | ErrorBadCharClass, // bad character class |
| 233 | ErrorBadCharRange, // bad character class range |
| 234 | ErrorMissingBracket, // missing closing ] |
| 235 | ErrorMissingParen, // missing closing ) |
| 236 | ErrorTrailingBackslash, // trailing \ at end of regexp |
| 237 | ErrorRepeatArgument, // repeat argument missing, e.g. "*" |
| 238 | ErrorRepeatSize, // bad repetition argument |
| 239 | ErrorRepeatOp, // bad repetition operator |
| 240 | ErrorBadPerlOp, // bad perl operator |
| 241 | ErrorBadUTF8, // invalid UTF-8 in regexp |
| 242 | ErrorBadNamedCapture, // bad named capture group |
| 243 | ErrorPatternTooLarge, // pattern too large (compile failed) |
| 244 | }; |
| 245 | |
| 246 | // Predefined common options. |
| 247 | // If you need more complicated things, instantiate |
Alexander Gutkin | 0d4c523 | 2013-02-28 13:47:27 +0000 | [diff] [blame] | 248 | // an Option class, possibly passing one of these to |
| 249 | // the Option constructor, change the settings, and pass that |
| 250 | // Option class to the RE2 constructor. |
| 251 | enum CannedOptions { |
| 252 | DefaultOptions = 0, |
| 253 | Latin1, // treat input as Latin-1 (default UTF-8) |
| 254 | POSIX, // POSIX syntax, leftmost-longest match |
| 255 | Quiet // do not log about regexp parse errors |
| 256 | }; |
Ian Hodson | 2ee91b4 | 2012-05-14 12:29:36 +0100 | [diff] [blame] | 257 | |
| 258 | // Need to have the const char* and const string& forms for implicit |
| 259 | // conversions when passing string literals to FullMatch and PartialMatch. |
| 260 | // Otherwise the StringPiece form would be sufficient. |
| 261 | #ifndef SWIG |
| 262 | RE2(const char* pattern); |
| 263 | RE2(const string& pattern); |
| 264 | #endif |
| 265 | RE2(const StringPiece& pattern); |
| 266 | RE2(const StringPiece& pattern, const Options& option); |
| 267 | ~RE2(); |
| 268 | |
| 269 | // Returns whether RE2 was created properly. |
| 270 | bool ok() const { return error_code() == NoError; } |
| 271 | |
| 272 | // The string specification for this RE2. E.g. |
| 273 | // RE2 re("ab*c?d+"); |
| 274 | // re.pattern(); // "ab*c?d+" |
| 275 | const string& pattern() const { return pattern_; } |
| 276 | |
| 277 | // If RE2 could not be created properly, returns an error string. |
| 278 | // Else returns the empty string. |
| 279 | const string& error() const { return *error_; } |
| 280 | |
| 281 | // If RE2 could not be created properly, returns an error code. |
| 282 | // Else returns RE2::NoError (== 0). |
| 283 | ErrorCode error_code() const { return error_code_; } |
| 284 | |
| 285 | // If RE2 could not be created properly, returns the offending |
| 286 | // portion of the regexp. |
| 287 | const string& error_arg() const { return error_arg_; } |
| 288 | |
| 289 | // Returns the program size, a very approximate measure of a regexp's "cost". |
| 290 | // Larger numbers are more expensive than smaller numbers. |
| 291 | int ProgramSize() const; |
| 292 | |
| 293 | // Returns the underlying Regexp; not for general use. |
| 294 | // Returns entire_regexp_ so that callers don't need |
| 295 | // to know about prefix_ and prefix_foldcase_. |
| 296 | re2::Regexp* Regexp() const { return entire_regexp_; } |
| 297 | |
| 298 | /***** The useful part: the matching interface *****/ |
| 299 | |
| 300 | // Matches "text" against "pattern". If pointer arguments are |
| 301 | // supplied, copies matched sub-patterns into them. |
| 302 | // |
| 303 | // You can pass in a "const char*" or a "string" for "text". |
| 304 | // You can pass in a "const char*" or a "string" or a "RE2" for "pattern". |
| 305 | // |
| 306 | // The provided pointer arguments can be pointers to any scalar numeric |
| 307 | // type, or one of: |
| 308 | // string (matched piece is copied to string) |
| 309 | // StringPiece (StringPiece is mutated to point to matched piece) |
| 310 | // T (where "bool T::ParseFrom(const char*, int)" exists) |
| 311 | // (void*)NULL (the corresponding matched sub-pattern is not copied) |
| 312 | // |
| 313 | // Returns true iff all of the following conditions are satisfied: |
| 314 | // a. "text" matches "pattern" exactly |
| 315 | // b. The number of matched sub-patterns is >= number of supplied pointers |
| 316 | // c. The "i"th argument has a suitable type for holding the |
| 317 | // string captured as the "i"th sub-pattern. If you pass in |
| 318 | // NULL for the "i"th argument, or pass fewer arguments than |
| 319 | // number of sub-patterns, "i"th captured sub-pattern is |
| 320 | // ignored. |
| 321 | // |
| 322 | // CAVEAT: An optional sub-pattern that does not exist in the |
| 323 | // matched string is assigned the empty string. Therefore, the |
| 324 | // following will return false (because the empty string is not a |
| 325 | // valid number): |
| 326 | // int number; |
| 327 | // RE2::FullMatch("abc", "[a-z]+(\\d+)?", &number); |
| 328 | static bool FullMatchN(const StringPiece& text, const RE2& re, |
| 329 | const Arg* const args[], int argc); |
| 330 | static const VariadicFunction2< |
| 331 | bool, const StringPiece&, const RE2&, Arg, RE2::FullMatchN> FullMatch; |
| 332 | |
| 333 | // Exactly like FullMatch(), except that "pattern" is allowed to match |
| 334 | // a substring of "text". |
| 335 | static bool PartialMatchN(const StringPiece& text, const RE2& re, // 3..16 args |
| 336 | const Arg* const args[], int argc); |
| 337 | static const VariadicFunction2< |
| 338 | bool, const StringPiece&, const RE2&, Arg, RE2::PartialMatchN> PartialMatch; |
| 339 | |
| 340 | // Like FullMatch() and PartialMatch(), except that pattern has to |
| 341 | // match a prefix of "text", and "input" is advanced past the matched |
| 342 | // text. Note: "input" is modified iff this routine returns true. |
| 343 | static bool ConsumeN(StringPiece* input, const RE2& pattern, // 3..16 args |
| 344 | const Arg* const args[], int argc); |
| 345 | static const VariadicFunction2< |
| 346 | bool, StringPiece*, const RE2&, Arg, RE2::ConsumeN> Consume; |
| 347 | |
| 348 | // Like Consume(..), but does not anchor the match at the beginning of the |
| 349 | // string. That is, "pattern" need not start its match at the beginning of |
| 350 | // "input". For example, "FindAndConsume(s, "(\\w+)", &word)" finds the next |
| 351 | // word in "s" and stores it in "word". |
| 352 | static bool FindAndConsumeN(StringPiece* input, const RE2& pattern, |
| 353 | const Arg* const args[], int argc); |
| 354 | static const VariadicFunction2< |
| 355 | bool, StringPiece*, const RE2&, Arg, RE2::FindAndConsumeN> FindAndConsume; |
| 356 | |
| 357 | // Replace the first match of "pattern" in "str" with "rewrite". |
| 358 | // Within "rewrite", backslash-escaped digits (\1 to \9) can be |
| 359 | // used to insert text matching corresponding parenthesized group |
| 360 | // from the pattern. \0 in "rewrite" refers to the entire matching |
| 361 | // text. E.g., |
| 362 | // |
| 363 | // string s = "yabba dabba doo"; |
| 364 | // CHECK(RE2::Replace(&s, "b+", "d")); |
| 365 | // |
| 366 | // will leave "s" containing "yada dabba doo" |
| 367 | // |
| 368 | // Returns true if the pattern matches and a replacement occurs, |
| 369 | // false otherwise. |
| 370 | static bool Replace(string *str, |
| 371 | const RE2& pattern, |
| 372 | const StringPiece& rewrite); |
| 373 | |
| 374 | // Like Replace(), except replaces successive non-overlapping occurrences |
| 375 | // of the pattern in the string with the rewrite. E.g. |
| 376 | // |
| 377 | // string s = "yabba dabba doo"; |
| 378 | // CHECK(RE2::GlobalReplace(&s, "b+", "d")); |
| 379 | // |
| 380 | // will leave "s" containing "yada dada doo" |
| 381 | // Replacements are not subject to re-matching. |
| 382 | // |
| 383 | // Because GlobalReplace only replaces non-overlapping matches, |
| 384 | // replacing "ana" within "banana" makes only one replacement, not two. |
| 385 | // |
| 386 | // Returns the number of replacements made. |
| 387 | static int GlobalReplace(string *str, |
| 388 | const RE2& pattern, |
| 389 | const StringPiece& rewrite); |
| 390 | |
| 391 | // Like Replace, except that if the pattern matches, "rewrite" |
| 392 | // is copied into "out" with substitutions. The non-matching |
| 393 | // portions of "text" are ignored. |
| 394 | // |
| 395 | // Returns true iff a match occurred and the extraction happened |
| 396 | // successfully; if no match occurs, the string is left unaffected. |
| 397 | static bool Extract(const StringPiece &text, |
| 398 | const RE2& pattern, |
| 399 | const StringPiece &rewrite, |
| 400 | string *out); |
| 401 | |
| 402 | // Escapes all potentially meaningful regexp characters in |
| 403 | // 'unquoted'. The returned string, used as a regular expression, |
| 404 | // will exactly match the original string. For example, |
| 405 | // 1.5-2.0? |
| 406 | // may become: |
| 407 | // 1\.5\-2\.0\? |
| 408 | static string QuoteMeta(const StringPiece& unquoted); |
| 409 | |
| 410 | // Computes range for any strings matching regexp. The min and max can in |
| 411 | // some cases be arbitrarily precise, so the caller gets to specify the |
| 412 | // maximum desired length of string returned. |
| 413 | // |
| 414 | // Assuming PossibleMatchRange(&min, &max, N) returns successfully, any |
| 415 | // string s that is an anchored match for this regexp satisfies |
| 416 | // min <= s && s <= max. |
| 417 | // |
| 418 | // Note that PossibleMatchRange() will only consider the first copy of an |
| 419 | // infinitely repeated element (i.e., any regexp element followed by a '*' or |
| 420 | // '+' operator). Regexps with "{N}" constructions are not affected, as those |
| 421 | // do not compile down to infinite repetitions. |
| 422 | // |
| 423 | // Returns true on success, false on error. |
| 424 | bool PossibleMatchRange(string* min, string* max, int maxlen) const; |
| 425 | |
| 426 | // Generic matching interface |
| 427 | |
| 428 | // Type of match. |
| 429 | enum Anchor { |
| 430 | UNANCHORED, // No anchoring |
| 431 | ANCHOR_START, // Anchor at start only |
| 432 | ANCHOR_BOTH, // Anchor at start and end |
| 433 | }; |
| 434 | |
| 435 | // Return the number of capturing subpatterns, or -1 if the |
| 436 | // regexp wasn't valid on construction. The overall match ($0) |
| 437 | // does not count: if the regexp is "(a)(b)", returns 2. |
| 438 | int NumberOfCapturingGroups() const; |
| 439 | |
| 440 | |
| 441 | // Return a map from names to capturing indices. |
| 442 | // The map records the index of the leftmost group |
| 443 | // with the given name. |
| 444 | // Only valid until the re is deleted. |
| 445 | const map<string, int>& NamedCapturingGroups() const; |
| 446 | |
| 447 | // Return a map from capturing indices to names. |
| 448 | // The map has no entries for unnamed groups. |
| 449 | // Only valid until the re is deleted. |
| 450 | const map<int, string>& CapturingGroupNames() const; |
| 451 | |
| 452 | // General matching routine. |
| 453 | // Match against text starting at offset startpos |
| 454 | // and stopping the search at offset endpos. |
| 455 | // Returns true if match found, false if not. |
| 456 | // On a successful match, fills in match[] (up to nmatch entries) |
| 457 | // with information about submatches. |
| 458 | // I.e. matching RE2("(foo)|(bar)baz") on "barbazbla" will return true, |
| 459 | // setting match[0] = "barbaz", match[1] = NULL, match[2] = "bar", |
| 460 | // match[3] = NULL, ..., up to match[nmatch-1] = NULL. |
| 461 | // |
| 462 | // Don't ask for more match information than you will use: |
| 463 | // runs much faster with nmatch == 1 than nmatch > 1, and |
| 464 | // runs even faster if nmatch == 0. |
| 465 | // Doesn't make sense to use nmatch > 1 + NumberOfCapturingGroups(), |
| 466 | // but will be handled correctly. |
| 467 | // |
| 468 | // Passing text == StringPiece(NULL, 0) will be handled like any other |
| 469 | // empty string, but note that on return, it will not be possible to tell |
| 470 | // whether submatch i matched the empty string or did not match: |
| 471 | // either way, match[i] == NULL. |
| 472 | bool Match(const StringPiece& text, |
| 473 | int startpos, |
| 474 | int endpos, |
| 475 | Anchor anchor, |
| 476 | StringPiece *match, |
| 477 | int nmatch) const; |
| 478 | |
| 479 | // Check that the given rewrite string is suitable for use with this |
| 480 | // regular expression. It checks that: |
| 481 | // * The regular expression has enough parenthesized subexpressions |
| 482 | // to satisfy all of the \N tokens in rewrite |
| 483 | // * The rewrite string doesn't have any syntax errors. E.g., |
| 484 | // '\' followed by anything other than a digit or '\'. |
| 485 | // A true return value guarantees that Replace() and Extract() won't |
| 486 | // fail because of a bad rewrite string. |
| 487 | bool CheckRewriteString(const StringPiece& rewrite, string* error) const; |
| 488 | |
Alexander Gutkin | 0d4c523 | 2013-02-28 13:47:27 +0000 | [diff] [blame] | 489 | // Returns the maximum submatch needed for the rewrite to be done by |
| 490 | // Replace(). E.g. if rewrite == "foo \\2,\\1", returns 2. |
| 491 | static int MaxSubmatch(const StringPiece& rewrite); |
| 492 | |
| 493 | // Append the "rewrite" string, with backslash subsitutions from "vec", |
| 494 | // to string "out". |
| 495 | // Returns true on success. This method can fail because of a malformed |
| 496 | // rewrite string. CheckRewriteString guarantees that the rewrite will |
| 497 | // be sucessful. |
| 498 | bool Rewrite(string *out, |
| 499 | const StringPiece &rewrite, |
| 500 | const StringPiece* vec, |
| 501 | int veclen) const; |
| 502 | |
Ian Hodson | 2ee91b4 | 2012-05-14 12:29:36 +0100 | [diff] [blame] | 503 | // Constructor options |
| 504 | class Options { |
| 505 | public: |
| 506 | // The options are (defaults in parentheses): |
| 507 | // |
| 508 | // utf8 (true) text and pattern are UTF-8; otherwise Latin-1 |
| 509 | // posix_syntax (false) restrict regexps to POSIX egrep syntax |
| 510 | // longest_match (false) search for longest match, not first match |
| 511 | // log_errors (true) log syntax and execution errors to ERROR |
| 512 | // max_mem (see below) approx. max memory footprint of RE2 |
| 513 | // literal (false) interpret string as literal, not regexp |
| 514 | // never_nl (false) never match \n, even if it is in regexp |
Alexander Gutkin | 0d4c523 | 2013-02-28 13:47:27 +0000 | [diff] [blame] | 515 | // never_capture (false) parse all parens as non-capturing |
Ian Hodson | 2ee91b4 | 2012-05-14 12:29:36 +0100 | [diff] [blame] | 516 | // case_sensitive (true) match is case-sensitive (regexp can override |
| 517 | // with (?i) unless in posix_syntax mode) |
| 518 | // |
| 519 | // The following options are only consulted when posix_syntax == true. |
| 520 | // (When posix_syntax == false these features are always enabled and |
| 521 | // cannot be turned off.) |
| 522 | // perl_classes (false) allow Perl's \d \s \w \D \S \W |
| 523 | // word_boundary (false) allow Perl's \b \B (word boundary and not) |
| 524 | // one_line (false) ^ and $ only match beginning and end of text |
| 525 | // |
| 526 | // The max_mem option controls how much memory can be used |
| 527 | // to hold the compiled form of the regexp (the Prog) and |
| 528 | // its cached DFA graphs. Code Search placed limits on the number |
| 529 | // of Prog instructions and DFA states: 10,000 for both. |
| 530 | // In RE2, those limits would translate to about 240 KB per Prog |
| 531 | // and perhaps 2.5 MB per DFA (DFA state sizes vary by regexp; RE2 does a |
| 532 | // better job of keeping them small than Code Search did). |
| 533 | // Each RE2 has two Progs (one forward, one reverse), and each Prog |
| 534 | // can have two DFAs (one first match, one longest match). |
| 535 | // That makes 4 DFAs: |
| 536 | // |
| 537 | // forward, first-match - used for UNANCHORED or ANCHOR_LEFT searches |
| 538 | // if opt.longest_match() == false |
| 539 | // forward, longest-match - used for all ANCHOR_BOTH searches, |
| 540 | // and the other two kinds if |
| 541 | // opt.longest_match() == true |
| 542 | // reverse, first-match - never used |
| 543 | // reverse, longest-match - used as second phase for unanchored searches |
| 544 | // |
| 545 | // The RE2 memory budget is statically divided between the two |
| 546 | // Progs and then the DFAs: two thirds to the forward Prog |
| 547 | // and one third to the reverse Prog. The forward Prog gives half |
| 548 | // of what it has left over to each of its DFAs. The reverse Prog |
| 549 | // gives it all to its longest-match DFA. |
| 550 | // |
| 551 | // Once a DFA fills its budget, it flushes its cache and starts over. |
| 552 | // If this happens too often, RE2 falls back on the NFA implementation. |
| 553 | |
| 554 | // For now, make the default budget something close to Code Search. |
| 555 | static const int kDefaultMaxMem = 8<<20; |
| 556 | |
| 557 | enum Encoding { |
| 558 | EncodingUTF8 = 1, |
| 559 | EncodingLatin1 |
| 560 | }; |
| 561 | |
| 562 | Options() : |
| 563 | encoding_(EncodingUTF8), |
| 564 | posix_syntax_(false), |
| 565 | longest_match_(false), |
| 566 | log_errors_(true), |
| 567 | max_mem_(kDefaultMaxMem), |
| 568 | literal_(false), |
| 569 | never_nl_(false), |
Alexander Gutkin | 0d4c523 | 2013-02-28 13:47:27 +0000 | [diff] [blame] | 570 | never_capture_(false), |
Ian Hodson | 2ee91b4 | 2012-05-14 12:29:36 +0100 | [diff] [blame] | 571 | case_sensitive_(true), |
| 572 | perl_classes_(false), |
| 573 | word_boundary_(false), |
| 574 | one_line_(false) { |
| 575 | } |
Alexander Gutkin | 0d4c523 | 2013-02-28 13:47:27 +0000 | [diff] [blame] | 576 | |
| 577 | /*implicit*/ Options(CannedOptions); |
Ian Hodson | 2ee91b4 | 2012-05-14 12:29:36 +0100 | [diff] [blame] | 578 | |
| 579 | Encoding encoding() const { return encoding_; } |
| 580 | void set_encoding(Encoding encoding) { encoding_ = encoding; } |
| 581 | |
| 582 | // Legacy interface to encoding. |
| 583 | // TODO(rsc): Remove once clients have been converted. |
| 584 | bool utf8() const { return encoding_ == EncodingUTF8; } |
| 585 | void set_utf8(bool b) { |
| 586 | if (b) { |
| 587 | encoding_ = EncodingUTF8; |
| 588 | } else { |
| 589 | encoding_ = EncodingLatin1; |
| 590 | } |
| 591 | } |
| 592 | |
| 593 | bool posix_syntax() const { return posix_syntax_; } |
| 594 | void set_posix_syntax(bool b) { posix_syntax_ = b; } |
| 595 | |
| 596 | bool longest_match() const { return longest_match_; } |
| 597 | void set_longest_match(bool b) { longest_match_ = b; } |
| 598 | |
| 599 | bool log_errors() const { return log_errors_; } |
| 600 | void set_log_errors(bool b) { log_errors_ = b; } |
| 601 | |
| 602 | int max_mem() const { return max_mem_; } |
| 603 | void set_max_mem(int m) { max_mem_ = m; } |
| 604 | |
| 605 | bool literal() const { return literal_; } |
| 606 | void set_literal(bool b) { literal_ = b; } |
| 607 | |
| 608 | bool never_nl() const { return never_nl_; } |
| 609 | void set_never_nl(bool b) { never_nl_ = b; } |
| 610 | |
Alexander Gutkin | 0d4c523 | 2013-02-28 13:47:27 +0000 | [diff] [blame] | 611 | bool never_capture() const { return never_capture_; } |
| 612 | void set_never_capture(bool b) { never_capture_ = b; } |
| 613 | |
Ian Hodson | 2ee91b4 | 2012-05-14 12:29:36 +0100 | [diff] [blame] | 614 | bool case_sensitive() const { return case_sensitive_; } |
| 615 | void set_case_sensitive(bool b) { case_sensitive_ = b; } |
| 616 | |
| 617 | bool perl_classes() const { return perl_classes_; } |
| 618 | void set_perl_classes(bool b) { perl_classes_ = b; } |
| 619 | |
| 620 | bool word_boundary() const { return word_boundary_; } |
| 621 | void set_word_boundary(bool b) { word_boundary_ = b; } |
| 622 | |
| 623 | bool one_line() const { return one_line_; } |
| 624 | void set_one_line(bool b) { one_line_ = b; } |
| 625 | |
| 626 | void Copy(const Options& src) { |
| 627 | encoding_ = src.encoding_; |
| 628 | posix_syntax_ = src.posix_syntax_; |
| 629 | longest_match_ = src.longest_match_; |
| 630 | log_errors_ = src.log_errors_; |
| 631 | max_mem_ = src.max_mem_; |
| 632 | literal_ = src.literal_; |
| 633 | never_nl_ = src.never_nl_; |
Alexander Gutkin | 0d4c523 | 2013-02-28 13:47:27 +0000 | [diff] [blame] | 634 | never_capture_ = src.never_capture_; |
Ian Hodson | 2ee91b4 | 2012-05-14 12:29:36 +0100 | [diff] [blame] | 635 | case_sensitive_ = src.case_sensitive_; |
| 636 | perl_classes_ = src.perl_classes_; |
| 637 | word_boundary_ = src.word_boundary_; |
| 638 | one_line_ = src.one_line_; |
| 639 | } |
| 640 | |
| 641 | int ParseFlags() const; |
| 642 | |
| 643 | private: |
Ian Hodson | 2ee91b4 | 2012-05-14 12:29:36 +0100 | [diff] [blame] | 644 | Encoding encoding_; |
| 645 | bool posix_syntax_; |
| 646 | bool longest_match_; |
| 647 | bool log_errors_; |
| 648 | int64_t max_mem_; |
| 649 | bool literal_; |
| 650 | bool never_nl_; |
Alexander Gutkin | 0d4c523 | 2013-02-28 13:47:27 +0000 | [diff] [blame] | 651 | bool never_capture_; |
Ian Hodson | 2ee91b4 | 2012-05-14 12:29:36 +0100 | [diff] [blame] | 652 | bool case_sensitive_; |
| 653 | bool perl_classes_; |
| 654 | bool word_boundary_; |
| 655 | bool one_line_; |
| 656 | |
| 657 | //DISALLOW_EVIL_CONSTRUCTORS(Options); |
| 658 | Options(const Options&); |
| 659 | void operator=(const Options&); |
| 660 | }; |
| 661 | |
| 662 | // Returns the options set in the constructor. |
| 663 | const Options& options() const { return options_; }; |
| 664 | |
| 665 | // Argument converters; see below. |
| 666 | static inline Arg CRadix(short* x); |
| 667 | static inline Arg CRadix(unsigned short* x); |
| 668 | static inline Arg CRadix(int* x); |
| 669 | static inline Arg CRadix(unsigned int* x); |
| 670 | static inline Arg CRadix(long* x); |
| 671 | static inline Arg CRadix(unsigned long* x); |
| 672 | static inline Arg CRadix(long long* x); |
| 673 | static inline Arg CRadix(unsigned long long* x); |
| 674 | |
| 675 | static inline Arg Hex(short* x); |
| 676 | static inline Arg Hex(unsigned short* x); |
| 677 | static inline Arg Hex(int* x); |
| 678 | static inline Arg Hex(unsigned int* x); |
| 679 | static inline Arg Hex(long* x); |
| 680 | static inline Arg Hex(unsigned long* x); |
| 681 | static inline Arg Hex(long long* x); |
| 682 | static inline Arg Hex(unsigned long long* x); |
| 683 | |
| 684 | static inline Arg Octal(short* x); |
| 685 | static inline Arg Octal(unsigned short* x); |
| 686 | static inline Arg Octal(int* x); |
| 687 | static inline Arg Octal(unsigned int* x); |
| 688 | static inline Arg Octal(long* x); |
| 689 | static inline Arg Octal(unsigned long* x); |
| 690 | static inline Arg Octal(long long* x); |
| 691 | static inline Arg Octal(unsigned long long* x); |
| 692 | |
| 693 | private: |
| 694 | void Init(const StringPiece& pattern, const Options& options); |
| 695 | |
Ian Hodson | 2ee91b4 | 2012-05-14 12:29:36 +0100 | [diff] [blame] | 696 | bool DoMatch(const StringPiece& text, |
| 697 | Anchor anchor, |
| 698 | int* consumed, |
| 699 | const Arg* const args[], |
| 700 | int n) const; |
| 701 | |
| 702 | re2::Prog* ReverseProg() const; |
| 703 | |
| 704 | mutable Mutex* mutex_; |
| 705 | string pattern_; // string regular expression |
| 706 | Options options_; // option flags |
| 707 | string prefix_; // required prefix (before regexp_) |
| 708 | bool prefix_foldcase_; // prefix is ASCII case-insensitive |
| 709 | re2::Regexp* entire_regexp_; // parsed regular expression |
| 710 | re2::Regexp* suffix_regexp_; // parsed regular expression, prefix removed |
| 711 | re2::Prog* prog_; // compiled program for regexp |
| 712 | mutable re2::Prog* rprog_; // reverse program for regexp |
| 713 | bool is_one_pass_; // can use prog_->SearchOnePass? |
| 714 | mutable const string* error_; // Error indicator |
| 715 | // (or points to empty string) |
| 716 | mutable ErrorCode error_code_; // Error code |
| 717 | mutable string error_arg_; // Fragment of regexp showing error |
| 718 | mutable int num_captures_; // Number of capturing groups |
| 719 | |
| 720 | // Map from capture names to indices |
| 721 | mutable const map<string, int>* named_groups_; |
| 722 | |
| 723 | // Map from capture indices to names |
| 724 | mutable const map<int, string>* group_names_; |
| 725 | |
| 726 | //DISALLOW_EVIL_CONSTRUCTORS(RE2); |
| 727 | RE2(const RE2&); |
| 728 | void operator=(const RE2&); |
| 729 | }; |
| 730 | |
| 731 | /***** Implementation details *****/ |
| 732 | |
| 733 | // Hex/Octal/Binary? |
| 734 | |
| 735 | // Special class for parsing into objects that define a ParseFrom() method |
| 736 | template <class T> |
| 737 | class _RE2_MatchObject { |
| 738 | public: |
| 739 | static inline bool Parse(const char* str, int n, void* dest) { |
| 740 | if (dest == NULL) return true; |
| 741 | T* object = reinterpret_cast<T*>(dest); |
| 742 | return object->ParseFrom(str, n); |
| 743 | } |
| 744 | }; |
| 745 | |
| 746 | class RE2::Arg { |
| 747 | public: |
| 748 | // Empty constructor so we can declare arrays of RE2::Arg |
| 749 | Arg(); |
| 750 | |
| 751 | // Constructor specially designed for NULL arguments |
| 752 | Arg(void*); |
| 753 | |
| 754 | typedef bool (*Parser)(const char* str, int n, void* dest); |
| 755 | |
| 756 | // Type-specific parsers |
| 757 | #define MAKE_PARSER(type,name) \ |
| 758 | Arg(type* p) : arg_(p), parser_(name) { } \ |
| 759 | Arg(type* p, Parser parser) : arg_(p), parser_(parser) { } \ |
| 760 | |
| 761 | |
| 762 | MAKE_PARSER(char, parse_char); |
| 763 | MAKE_PARSER(signed char, parse_char); |
| 764 | MAKE_PARSER(unsigned char, parse_uchar); |
| 765 | MAKE_PARSER(short, parse_short); |
| 766 | MAKE_PARSER(unsigned short, parse_ushort); |
| 767 | MAKE_PARSER(int, parse_int); |
| 768 | MAKE_PARSER(unsigned int, parse_uint); |
| 769 | MAKE_PARSER(long, parse_long); |
| 770 | MAKE_PARSER(unsigned long, parse_ulong); |
| 771 | MAKE_PARSER(long long, parse_longlong); |
| 772 | MAKE_PARSER(unsigned long long, parse_ulonglong); |
| 773 | MAKE_PARSER(float, parse_float); |
| 774 | MAKE_PARSER(double, parse_double); |
| 775 | MAKE_PARSER(string, parse_string); |
| 776 | MAKE_PARSER(StringPiece, parse_stringpiece); |
| 777 | |
| 778 | #undef MAKE_PARSER |
| 779 | |
| 780 | // Generic constructor |
| 781 | template <class T> Arg(T*, Parser parser); |
| 782 | // Generic constructor template |
| 783 | template <class T> Arg(T* p) |
| 784 | : arg_(p), parser_(_RE2_MatchObject<T>::Parse) { |
| 785 | } |
| 786 | |
| 787 | // Parse the data |
| 788 | bool Parse(const char* str, int n) const; |
| 789 | |
| 790 | private: |
| 791 | void* arg_; |
| 792 | Parser parser_; |
| 793 | |
| 794 | static bool parse_null (const char* str, int n, void* dest); |
| 795 | static bool parse_char (const char* str, int n, void* dest); |
| 796 | static bool parse_uchar (const char* str, int n, void* dest); |
| 797 | static bool parse_float (const char* str, int n, void* dest); |
| 798 | static bool parse_double (const char* str, int n, void* dest); |
| 799 | static bool parse_string (const char* str, int n, void* dest); |
| 800 | static bool parse_stringpiece (const char* str, int n, void* dest); |
| 801 | |
| 802 | #define DECLARE_INTEGER_PARSER(name) \ |
| 803 | private: \ |
| 804 | static bool parse_ ## name(const char* str, int n, void* dest); \ |
| 805 | static bool parse_ ## name ## _radix( \ |
| 806 | const char* str, int n, void* dest, int radix); \ |
| 807 | public: \ |
| 808 | static bool parse_ ## name ## _hex(const char* str, int n, void* dest); \ |
| 809 | static bool parse_ ## name ## _octal(const char* str, int n, void* dest); \ |
| 810 | static bool parse_ ## name ## _cradix(const char* str, int n, void* dest) |
| 811 | |
| 812 | DECLARE_INTEGER_PARSER(short); |
| 813 | DECLARE_INTEGER_PARSER(ushort); |
| 814 | DECLARE_INTEGER_PARSER(int); |
| 815 | DECLARE_INTEGER_PARSER(uint); |
| 816 | DECLARE_INTEGER_PARSER(long); |
| 817 | DECLARE_INTEGER_PARSER(ulong); |
| 818 | DECLARE_INTEGER_PARSER(longlong); |
| 819 | DECLARE_INTEGER_PARSER(ulonglong); |
| 820 | |
| 821 | #undef DECLARE_INTEGER_PARSER |
| 822 | }; |
| 823 | |
| 824 | inline RE2::Arg::Arg() : arg_(NULL), parser_(parse_null) { } |
| 825 | inline RE2::Arg::Arg(void* p) : arg_(p), parser_(parse_null) { } |
| 826 | |
| 827 | inline bool RE2::Arg::Parse(const char* str, int n) const { |
| 828 | return (*parser_)(str, n, arg_); |
| 829 | } |
| 830 | |
| 831 | // This part of the parser, appropriate only for ints, deals with bases |
| 832 | #define MAKE_INTEGER_PARSER(type, name) \ |
| 833 | inline RE2::Arg RE2::Hex(type* ptr) { \ |
| 834 | return RE2::Arg(ptr, RE2::Arg::parse_ ## name ## _hex); } \ |
| 835 | inline RE2::Arg RE2::Octal(type* ptr) { \ |
| 836 | return RE2::Arg(ptr, RE2::Arg::parse_ ## name ## _octal); } \ |
| 837 | inline RE2::Arg RE2::CRadix(type* ptr) { \ |
| 838 | return RE2::Arg(ptr, RE2::Arg::parse_ ## name ## _cradix); } |
| 839 | |
| 840 | MAKE_INTEGER_PARSER(short, short); |
| 841 | MAKE_INTEGER_PARSER(unsigned short, ushort); |
| 842 | MAKE_INTEGER_PARSER(int, int); |
| 843 | MAKE_INTEGER_PARSER(unsigned int, uint); |
| 844 | MAKE_INTEGER_PARSER(long, long); |
| 845 | MAKE_INTEGER_PARSER(unsigned long, ulong); |
| 846 | MAKE_INTEGER_PARSER(long long, longlong); |
| 847 | MAKE_INTEGER_PARSER(unsigned long long, ulonglong); |
| 848 | |
| 849 | #undef MAKE_INTEGER_PARSER |
| 850 | |
| 851 | } // namespace re2 |
| 852 | |
| 853 | using re2::RE2; |
| 854 | |
| 855 | #endif /* RE2_RE2_H */ |