Alexander Gutkin | 0d4c523 | 2013-02-28 13:47:27 +0000 | [diff] [blame] | 1 | // Copyright 2008 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 | // Regular expression engine tester -- test all the implementations against each other. |
| 6 | |
| 7 | #include "util/util.h" |
| 8 | #include "util/flags.h" |
| 9 | #include "re2/testing/tester.h" |
| 10 | #include "re2/prog.h" |
| 11 | #include "re2/re2.h" |
| 12 | #include "re2/regexp.h" |
| 13 | |
| 14 | DEFINE_bool(dump_prog, false, "dump regexp program"); |
| 15 | DEFINE_bool(log_okay, false, "log successful runs"); |
| 16 | DEFINE_bool(dump_rprog, false, "dump reversed regexp program"); |
| 17 | |
| 18 | DEFINE_int32(max_regexp_failures, 100, |
| 19 | "maximum number of regexp test failures (-1 = unlimited)"); |
| 20 | |
| 21 | DEFINE_string(regexp_engines, "", "pattern to select regexp engines to test"); |
| 22 | |
| 23 | namespace re2 { |
| 24 | |
| 25 | enum { |
| 26 | kMaxSubmatch = 1+16, // $0...$16 |
| 27 | }; |
| 28 | |
| 29 | const char* engine_types[kEngineMax] = { |
| 30 | "Backtrack", |
| 31 | "NFA", |
| 32 | "DFA", |
| 33 | "DFA1", |
| 34 | "OnePass", |
| 35 | "BitState", |
| 36 | "RE2", |
| 37 | "RE2a", |
| 38 | "RE2b", |
| 39 | "PCRE", |
| 40 | }; |
| 41 | |
| 42 | // Returns the name string for the type t. |
| 43 | static string EngineString(Engine t) { |
| 44 | if (t < 0 || t >= arraysize(engine_types) || engine_types[t] == NULL) { |
| 45 | return StringPrintf("type%d", static_cast<int>(t)); |
| 46 | } |
| 47 | return engine_types[t]; |
| 48 | } |
| 49 | |
| 50 | // Returns bit mask of engines to use. |
| 51 | static uint32 Engines() { |
| 52 | static uint32 cached_engines; |
| 53 | static bool did_parse; |
| 54 | |
| 55 | if (did_parse) |
| 56 | return cached_engines; |
| 57 | |
| 58 | if (FLAGS_regexp_engines.empty()) { |
| 59 | cached_engines = ~0; |
| 60 | } else { |
| 61 | for (Engine i = static_cast<Engine>(0); i < kEngineMax; i++) |
| 62 | if (strstr(EngineString(i).c_str(), FLAGS_regexp_engines.c_str())) |
| 63 | cached_engines |= 1<<i; |
| 64 | } |
| 65 | |
| 66 | if (cached_engines == 0) |
| 67 | LOG(INFO) << "Warning: no engines enabled."; |
| 68 | if (!UsingPCRE) |
| 69 | cached_engines &= ~(1<<kEnginePCRE); |
| 70 | for (Engine i = static_cast<Engine>(0); i < kEngineMax; i++) { |
| 71 | if (cached_engines & (1<<i)) |
| 72 | LOG(INFO) << EngineString(i) << " enabled"; |
| 73 | } |
| 74 | did_parse = true; |
| 75 | return cached_engines; |
| 76 | } |
| 77 | |
| 78 | // The result of running a match. |
| 79 | struct TestInstance::Result { |
| 80 | bool skipped; // test skipped: wasn't applicable |
| 81 | bool matched; // found a match |
| 82 | bool untrusted; // don't really trust the answer |
| 83 | bool have_submatch; // computed all submatch info |
| 84 | bool have_submatch0; // computed just submatch[0] |
| 85 | StringPiece submatch[kMaxSubmatch]; |
| 86 | }; |
| 87 | |
| 88 | typedef TestInstance::Result Result; |
| 89 | |
| 90 | // Formats a single capture range s in text in the form (a,b) |
| 91 | // where a and b are the starting and ending offsets of s in text. |
| 92 | static string FormatCapture(const StringPiece& text, const StringPiece& s) { |
| 93 | if (s.begin() == NULL) |
| 94 | return "(?,?)"; |
| 95 | return StringPrintf("(%d,%d)", |
| 96 | static_cast<int>(s.begin() - text.begin()), |
| 97 | static_cast<int>(s.end() - text.begin())); |
| 98 | } |
| 99 | |
| 100 | // Returns whether text contains non-ASCII (>= 0x80) bytes. |
| 101 | static bool NonASCII(const StringPiece& text) { |
| 102 | for (int i = 0; i < text.size(); i++) |
| 103 | if ((uint8)text[i] >= 0x80) |
| 104 | return true; |
| 105 | return false; |
| 106 | } |
| 107 | |
| 108 | // Returns string representation of match kind. |
| 109 | static string FormatKind(Prog::MatchKind kind) { |
| 110 | switch (kind) { |
| 111 | case Prog::kFullMatch: |
| 112 | return "full match"; |
| 113 | case Prog::kLongestMatch: |
| 114 | return "longest match"; |
| 115 | case Prog::kFirstMatch: |
| 116 | return "first match"; |
| 117 | case Prog::kManyMatch: |
| 118 | return "many match"; |
| 119 | } |
| 120 | return "???"; |
| 121 | } |
| 122 | |
| 123 | // Returns string representation of anchor kind. |
| 124 | static string FormatAnchor(Prog::Anchor anchor) { |
| 125 | switch (anchor) { |
| 126 | case Prog::kAnchored: |
| 127 | return "anchored"; |
| 128 | case Prog::kUnanchored: |
| 129 | return "unanchored"; |
| 130 | } |
| 131 | return "???"; |
| 132 | } |
| 133 | |
| 134 | struct ParseMode { |
| 135 | Regexp::ParseFlags parse_flags; |
| 136 | string desc; |
| 137 | }; |
| 138 | |
| 139 | static const Regexp::ParseFlags single_line = |
| 140 | Regexp::LikePerl; |
| 141 | static const Regexp::ParseFlags multi_line = |
| 142 | static_cast<Regexp::ParseFlags>(Regexp::LikePerl & ~Regexp::OneLine); |
| 143 | |
| 144 | static ParseMode parse_modes[] = { |
| 145 | { single_line, "single-line" }, |
| 146 | { single_line|Regexp::Latin1, "single-line, latin1" }, |
| 147 | { multi_line, "multiline" }, |
| 148 | { multi_line|Regexp::NonGreedy, "multiline, nongreedy" }, |
| 149 | { multi_line|Regexp::Latin1, "multiline, latin1" }, |
| 150 | }; |
| 151 | |
| 152 | static string FormatMode(Regexp::ParseFlags flags) { |
| 153 | for (int i = 0; i < arraysize(parse_modes); i++) |
| 154 | if (parse_modes[i].parse_flags == flags) |
| 155 | return parse_modes[i].desc; |
| 156 | return StringPrintf("%#x", static_cast<uint>(flags)); |
| 157 | } |
| 158 | |
| 159 | // Constructs and saves all the matching engines that |
| 160 | // will be required for the given tests. |
| 161 | TestInstance::TestInstance(const StringPiece& regexp_str, Prog::MatchKind kind, |
| 162 | Regexp::ParseFlags flags) |
| 163 | : regexp_str_(regexp_str), |
| 164 | kind_(kind), |
| 165 | flags_(flags), |
| 166 | error_(false), |
| 167 | regexp_(NULL), |
| 168 | num_captures_(0), |
| 169 | prog_(NULL), |
| 170 | rprog_(NULL), |
| 171 | re_(NULL), |
| 172 | re2_(NULL) { |
| 173 | |
| 174 | VLOG(1) << CEscape(regexp_str); |
| 175 | |
| 176 | // Compile regexp to prog. |
| 177 | // Always required - needed for backtracking (reference implementation). |
| 178 | RegexpStatus status; |
| 179 | regexp_ = Regexp::Parse(regexp_str, flags, &status); |
| 180 | if (regexp_ == NULL) { |
| 181 | LOG(INFO) << "Cannot parse: " << CEscape(regexp_str_) |
| 182 | << " mode: " << FormatMode(flags); |
| 183 | error_ = true; |
| 184 | return; |
| 185 | } |
| 186 | num_captures_ = regexp_->NumCaptures(); |
| 187 | prog_ = regexp_->CompileToProg(0); |
| 188 | if (prog_ == NULL) { |
| 189 | LOG(INFO) << "Cannot compile: " << CEscape(regexp_str_); |
| 190 | error_ = true; |
| 191 | return; |
| 192 | } |
| 193 | if (FLAGS_dump_prog) { |
| 194 | LOG(INFO) << "Prog for " |
| 195 | << " regexp " |
| 196 | << CEscape(regexp_str_) |
| 197 | << " (" << FormatKind(kind_) |
| 198 | << ", " << FormatMode(flags_) |
| 199 | << ")\n" |
| 200 | << prog_->Dump(); |
| 201 | } |
| 202 | |
| 203 | // Compile regexp to reversed prog. Only needed for DFA engines. |
| 204 | if (Engines() & ((1<<kEngineDFA)|(1<<kEngineDFA1))) { |
| 205 | rprog_ = regexp_->CompileToReverseProg(0); |
| 206 | if (rprog_ == NULL) { |
| 207 | LOG(INFO) << "Cannot reverse compile: " << CEscape(regexp_str_); |
| 208 | error_ = true; |
| 209 | return; |
| 210 | } |
| 211 | if (FLAGS_dump_rprog) |
| 212 | LOG(INFO) << rprog_->Dump(); |
| 213 | } |
| 214 | |
| 215 | // Create re string that will be used for RE and RE2. |
| 216 | string re = regexp_str.as_string(); |
| 217 | // Accomodate flags. |
| 218 | // Regexp::Latin1 will be accomodated below. |
| 219 | if (!(flags & Regexp::OneLine)) |
| 220 | re = "(?m)" + re; |
| 221 | if (flags & Regexp::NonGreedy) |
| 222 | re = "(?U)" + re; |
| 223 | if (flags & Regexp::DotNL) |
| 224 | re = "(?s)" + re; |
| 225 | |
| 226 | // Compile regexp to RE2. |
| 227 | if (Engines() & ((1<<kEngineRE2)|(1<<kEngineRE2a)|(1<<kEngineRE2b))) { |
| 228 | RE2::Options options; |
| 229 | if (flags & Regexp::Latin1) |
| 230 | options.set_encoding(RE2::Options::EncodingLatin1); |
| 231 | if (kind_ == Prog::kLongestMatch) |
| 232 | options.set_longest_match(true); |
| 233 | re2_ = new RE2(re, options); |
| 234 | if (!re2_->error().empty()) { |
| 235 | LOG(INFO) << "Cannot RE2: " << CEscape(re); |
| 236 | error_ = true; |
| 237 | return; |
| 238 | } |
| 239 | } |
| 240 | |
| 241 | // Compile regexp to RE. |
| 242 | // PCRE as exposed by the RE interface isn't always usable. |
| 243 | // 1. It disagrees about handling of empty-string reptitions |
| 244 | // like matching (a*)* against "b". PCRE treats the (a*) as |
| 245 | // occurring once, while we treat it as occurring not at all. |
| 246 | // 2. It treats $ as this weird thing meaning end of string |
| 247 | // or before the \n at the end of the string. |
| 248 | // 3. It doesn't implement POSIX leftmost-longest matching. |
| 249 | // MimicsPCRE() detects 1 and 2. |
| 250 | if ((Engines() & (1<<kEnginePCRE)) && regexp_->MimicsPCRE() && |
| 251 | kind_ != Prog::kLongestMatch) { |
| 252 | PCRE_Options o; |
| 253 | o.set_option(PCRE::UTF8); |
| 254 | if (flags & Regexp::Latin1) |
| 255 | o.set_option(PCRE::None); |
| 256 | // PCRE has interface bug keeping us from finding $0, so |
| 257 | // add one more layer of parens. |
| 258 | re_ = new PCRE("("+re+")", o); |
| 259 | if (!re_->error().empty()) { |
| 260 | LOG(INFO) << "Cannot PCRE: " << CEscape(re); |
| 261 | error_ = true; |
| 262 | return; |
| 263 | } |
| 264 | } |
| 265 | } |
| 266 | |
| 267 | TestInstance::~TestInstance() { |
| 268 | if (regexp_) |
| 269 | regexp_->Decref(); |
| 270 | delete prog_; |
| 271 | delete rprog_; |
| 272 | delete re_; |
| 273 | delete re2_; |
| 274 | } |
| 275 | |
| 276 | // Runs a single search using the named engine type. |
| 277 | // This interface hides all the irregularities of the various |
| 278 | // engine interfaces from the rest of this file. |
| 279 | void TestInstance::RunSearch(Engine type, |
| 280 | const StringPiece& orig_text, |
| 281 | const StringPiece& orig_context, |
| 282 | Prog::Anchor anchor, |
| 283 | Result *result) { |
| 284 | memset(result, 0, sizeof *result); |
| 285 | if (regexp_ == NULL) { |
| 286 | result->skipped = true; |
| 287 | return; |
| 288 | } |
| 289 | int nsubmatch = 1 + num_captures_; // NumCaptures doesn't count $0 |
| 290 | if (nsubmatch > kMaxSubmatch) |
| 291 | nsubmatch = kMaxSubmatch; |
| 292 | |
| 293 | StringPiece text = orig_text; |
| 294 | StringPiece context = orig_context; |
| 295 | |
| 296 | switch (type) { |
| 297 | default: |
| 298 | LOG(FATAL) << "Bad RunSearch type: " << (int)type; |
| 299 | |
| 300 | case kEngineBacktrack: |
| 301 | if (prog_ == NULL) { |
| 302 | result->skipped = true; |
| 303 | break; |
| 304 | } |
| 305 | result->matched = |
| 306 | prog_->UnsafeSearchBacktrack(text, context, anchor, kind_, |
| 307 | result->submatch, nsubmatch); |
| 308 | result->have_submatch = true; |
| 309 | break; |
| 310 | |
| 311 | case kEngineNFA: |
| 312 | if (prog_ == NULL) { |
| 313 | result->skipped = true; |
| 314 | break; |
| 315 | } |
| 316 | result->matched = |
| 317 | prog_->SearchNFA(text, context, anchor, kind_, |
| 318 | result->submatch, nsubmatch); |
| 319 | result->have_submatch = true; |
| 320 | break; |
| 321 | |
| 322 | case kEngineDFA: |
| 323 | if (prog_ == NULL) { |
| 324 | result->skipped = true; |
| 325 | break; |
| 326 | } |
| 327 | result->matched = prog_->SearchDFA(text, context, anchor, kind_, NULL, |
| 328 | &result->skipped, NULL); |
| 329 | break; |
| 330 | |
| 331 | case kEngineDFA1: |
| 332 | if (prog_ == NULL || rprog_ == NULL) { |
| 333 | result->skipped = true; |
| 334 | break; |
| 335 | } |
| 336 | result->matched = |
| 337 | prog_->SearchDFA(text, context, anchor, kind_, result->submatch, |
| 338 | &result->skipped, NULL); |
| 339 | // If anchored, no need for second run, |
| 340 | // but do it anyway to find more bugs. |
| 341 | if (result->matched) { |
| 342 | if (!rprog_->SearchDFA(result->submatch[0], context, |
| 343 | Prog::kAnchored, Prog::kLongestMatch, |
| 344 | result->submatch, |
| 345 | &result->skipped, NULL)) { |
| 346 | LOG(ERROR) << "Reverse DFA inconsistency: " << CEscape(regexp_str_) |
| 347 | << " on " << CEscape(text); |
| 348 | result->matched = false; |
| 349 | } |
| 350 | } |
| 351 | result->have_submatch0 = true; |
| 352 | break; |
| 353 | |
| 354 | case kEngineOnePass: |
| 355 | if (prog_ == NULL || |
| 356 | anchor == Prog::kUnanchored || |
| 357 | !prog_->IsOnePass() || |
| 358 | nsubmatch > Prog::kMaxOnePassCapture) { |
| 359 | result->skipped = true; |
| 360 | break; |
| 361 | } |
| 362 | result->matched = prog_->SearchOnePass(text, context, anchor, kind_, |
| 363 | result->submatch, nsubmatch); |
| 364 | result->have_submatch = true; |
| 365 | break; |
| 366 | |
| 367 | case kEngineBitState: |
| 368 | if (prog_ == NULL) { |
| 369 | result->skipped = true; |
| 370 | break; |
| 371 | } |
| 372 | result->matched = prog_->SearchBitState(text, context, anchor, kind_, |
| 373 | result->submatch, nsubmatch); |
| 374 | result->have_submatch = true; |
| 375 | break; |
| 376 | |
| 377 | case kEngineRE2: |
| 378 | case kEngineRE2a: |
| 379 | case kEngineRE2b: { |
| 380 | if (!re2_ || text.end() != context.end()) { |
| 381 | result->skipped = true; |
| 382 | break; |
| 383 | } |
| 384 | |
| 385 | RE2::Anchor re_anchor; |
| 386 | if (anchor == Prog::kAnchored) |
| 387 | re_anchor = RE2::ANCHOR_START; |
| 388 | else |
| 389 | re_anchor = RE2::UNANCHORED; |
| 390 | if (kind_ == Prog::kFullMatch) |
| 391 | re_anchor = RE2::ANCHOR_BOTH; |
| 392 | |
| 393 | result->matched = re2_->Match(context, |
| 394 | text.begin() - context.begin(), |
| 395 | text.end() - context.begin(), |
| 396 | re_anchor, result->submatch, nsubmatch); |
| 397 | result->have_submatch = nsubmatch > 0; |
| 398 | break; |
| 399 | } |
| 400 | |
| 401 | case kEnginePCRE: { |
| 402 | if (!re_ || text.begin() != context.begin() || |
| 403 | text.end() != context.end()) { |
| 404 | result->skipped = true; |
| 405 | break; |
| 406 | } |
| 407 | |
| 408 | const PCRE::Arg **argptr = new const PCRE::Arg*[nsubmatch]; |
| 409 | PCRE::Arg *a = new PCRE::Arg[nsubmatch]; |
| 410 | for (int i = 0; i < nsubmatch; i++) { |
| 411 | a[i] = PCRE::Arg(&result->submatch[i]); |
| 412 | argptr[i] = &a[i]; |
| 413 | } |
| 414 | int consumed; |
| 415 | PCRE::Anchor pcre_anchor; |
| 416 | if (anchor == Prog::kAnchored) |
| 417 | pcre_anchor = PCRE::ANCHOR_START; |
| 418 | else |
| 419 | pcre_anchor = PCRE::UNANCHORED; |
| 420 | if (kind_ == Prog::kFullMatch) |
| 421 | pcre_anchor = PCRE::ANCHOR_BOTH; |
| 422 | re_->ClearHitLimit(); |
| 423 | result->matched = |
| 424 | re_->DoMatch(text, |
| 425 | pcre_anchor, |
| 426 | &consumed, |
| 427 | argptr, nsubmatch); |
| 428 | if (re_->HitLimit()) { |
| 429 | result->untrusted = true; |
| 430 | delete[] argptr; |
| 431 | delete[] a; |
| 432 | break; |
| 433 | } |
| 434 | result->have_submatch = true; |
| 435 | |
| 436 | // Work around RE interface bug: PCRE returns -1 as the |
| 437 | // offsets for an unmatched subexpression, and RE should |
| 438 | // turn that into StringPiece(NULL) but in fact it uses |
| 439 | // StringPiece(text.begin() - 1, 0). Oops. |
| 440 | for (int i = 0; i < nsubmatch; i++) |
| 441 | if (result->submatch[i].begin() == text.begin() - 1) |
| 442 | result->submatch[i] = NULL; |
| 443 | delete[] argptr; |
| 444 | delete[] a; |
| 445 | break; |
| 446 | } |
| 447 | } |
| 448 | |
| 449 | if (!result->matched) |
| 450 | memset(result->submatch, 0, sizeof result->submatch); |
| 451 | } |
| 452 | |
| 453 | // Checks whether r is okay given that correct is the right answer. |
| 454 | // Specifically, r's answers have to match (but it doesn't have to |
| 455 | // claim to have all the answers). |
| 456 | static bool ResultOkay(const Result& r, const Result& correct) { |
| 457 | if (r.skipped) |
| 458 | return true; |
| 459 | if (r.matched != correct.matched) |
| 460 | return false; |
| 461 | if (r.have_submatch || r.have_submatch0) { |
| 462 | for (int i = 0; i < kMaxSubmatch; i++) { |
| 463 | if (correct.submatch[i].begin() != r.submatch[i].begin() || |
| 464 | correct.submatch[i].size() != r.submatch[i].size()) |
| 465 | return false; |
| 466 | if (!r.have_submatch) |
| 467 | break; |
| 468 | } |
| 469 | } |
| 470 | return true; |
| 471 | } |
| 472 | |
| 473 | // Runs a single test. |
| 474 | bool TestInstance::RunCase(const StringPiece& text, const StringPiece& context, |
| 475 | Prog::Anchor anchor) { |
| 476 | // Backtracking is the gold standard. |
| 477 | Result correct; |
| 478 | RunSearch(kEngineBacktrack, text, context, anchor, &correct); |
| 479 | if (correct.skipped) { |
| 480 | if (regexp_ == NULL) |
| 481 | return true; |
| 482 | LOG(ERROR) << "Skipped backtracking! " << CEscape(regexp_str_) |
| 483 | << " " << FormatMode(flags_); |
| 484 | return false; |
| 485 | } |
| 486 | VLOG(1) << "Try: regexp " << CEscape(regexp_str_) |
| 487 | << " text " << CEscape(text) |
| 488 | << " (" << FormatKind(kind_) |
| 489 | << ", " << FormatAnchor(anchor) |
| 490 | << ", " << FormatMode(flags_) |
| 491 | << ")"; |
| 492 | |
| 493 | // Compare the others. |
| 494 | bool all_okay = true; |
| 495 | for (Engine i = kEngineBacktrack+1; i < kEngineMax; i++) { |
| 496 | if (!(Engines() & (1<<i))) |
| 497 | continue; |
| 498 | |
| 499 | Result r; |
| 500 | RunSearch(i, text, context, anchor, &r); |
| 501 | if (ResultOkay(r, correct)) { |
| 502 | if (FLAGS_log_okay) |
| 503 | LogMatch(r.skipped ? "Skipped: " : "Okay: ", i, text, context, anchor); |
| 504 | continue; |
| 505 | } |
| 506 | |
| 507 | // We disagree with PCRE on the meaning of some Unicode matches. |
| 508 | // In particular, we treat all non-ASCII UTF-8 as word characters. |
| 509 | // We also treat "empty" character sets like [^\w\W] as being |
| 510 | // impossible to match, while PCRE apparently excludes some code |
| 511 | // points (e.g., 0x0080) from both \w and \W. |
| 512 | if (i == kEnginePCRE && NonASCII(text)) |
| 513 | continue; |
| 514 | |
| 515 | if (!r.untrusted) |
| 516 | all_okay = false; |
| 517 | |
| 518 | LogMatch(r.untrusted ? "(Untrusted) Mismatch: " : "Mismatch: ", i, text, |
| 519 | context, anchor); |
| 520 | if (r.matched != correct.matched) { |
| 521 | if (r.matched) { |
| 522 | LOG(INFO) << " Should not match (but does)."; |
| 523 | } else { |
| 524 | LOG(INFO) << " Should match (but does not)."; |
| 525 | continue; |
| 526 | } |
| 527 | } |
| 528 | for (int i = 0; i < 1+num_captures_; i++) { |
| 529 | if (r.submatch[i].begin() != correct.submatch[i].begin() || |
| 530 | r.submatch[i].end() != correct.submatch[i].end()) { |
| 531 | LOG(INFO) << |
| 532 | StringPrintf(" $%d: should be %s is %s", |
| 533 | i, |
| 534 | FormatCapture(text, correct.submatch[i]).c_str(), |
| 535 | FormatCapture(text, r.submatch[i]).c_str()); |
| 536 | } else { |
| 537 | LOG(INFO) << |
| 538 | StringPrintf(" $%d: %s ok", i, |
| 539 | FormatCapture(text, r.submatch[i]).c_str()); |
| 540 | } |
| 541 | } |
| 542 | } |
| 543 | |
| 544 | if (!all_okay) { |
| 545 | if (FLAGS_max_regexp_failures > 0 && --FLAGS_max_regexp_failures == 0) |
| 546 | LOG(QFATAL) << "Too many regexp failures."; |
| 547 | } |
| 548 | |
| 549 | return all_okay; |
| 550 | } |
| 551 | |
| 552 | void TestInstance::LogMatch(const char* prefix, Engine e, |
| 553 | const StringPiece& text, const StringPiece& context, |
| 554 | Prog::Anchor anchor) { |
| 555 | LOG(INFO) << prefix |
| 556 | << EngineString(e) |
| 557 | << " regexp " |
| 558 | << CEscape(regexp_str_) |
| 559 | << " " |
| 560 | << CEscape(regexp_->ToString()) |
| 561 | << " text " |
| 562 | << CEscape(text) |
| 563 | << " (" |
| 564 | << text.begin() - context.begin() |
| 565 | << "," |
| 566 | << text.end() - context.begin() |
| 567 | << ") of context " |
| 568 | << CEscape(context) |
| 569 | << " (" << FormatKind(kind_) |
| 570 | << ", " << FormatAnchor(anchor) |
| 571 | << ", " << FormatMode(flags_) |
| 572 | << ")"; |
| 573 | } |
| 574 | |
| 575 | static Prog::MatchKind kinds[] = { |
| 576 | Prog::kFirstMatch, |
| 577 | Prog::kLongestMatch, |
| 578 | Prog::kFullMatch, |
| 579 | }; |
| 580 | |
| 581 | // Test all possible match kinds and parse modes. |
| 582 | Tester::Tester(const StringPiece& regexp) { |
| 583 | error_ = false; |
| 584 | for (int i = 0; i < arraysize(kinds); i++) { |
| 585 | for (int j = 0; j < arraysize(parse_modes); j++) { |
| 586 | TestInstance* t = new TestInstance(regexp, kinds[i], |
| 587 | parse_modes[j].parse_flags); |
| 588 | error_ |= t->error(); |
| 589 | v_.push_back(t); |
| 590 | } |
| 591 | } |
| 592 | } |
| 593 | |
| 594 | Tester::~Tester() { |
| 595 | for (int i = 0; i < v_.size(); i++) |
| 596 | delete v_[i]; |
| 597 | } |
| 598 | |
| 599 | bool Tester::TestCase(const StringPiece& text, const StringPiece& context, |
| 600 | Prog::Anchor anchor) { |
| 601 | bool okay = true; |
| 602 | for (int i = 0; i < v_.size(); i++) |
| 603 | okay &= (!v_[i]->error() && v_[i]->RunCase(text, context, anchor)); |
| 604 | return okay; |
| 605 | } |
| 606 | |
| 607 | static Prog::Anchor anchors[] = { |
| 608 | Prog::kAnchored, |
| 609 | Prog::kUnanchored |
| 610 | }; |
| 611 | |
| 612 | bool Tester::TestInput(const StringPiece& text) { |
| 613 | bool okay = TestInputInContext(text, text); |
| 614 | if (text.size() > 0) { |
| 615 | StringPiece sp; |
| 616 | sp = text; |
| 617 | sp.remove_prefix(1); |
| 618 | okay &= TestInputInContext(sp, text); |
| 619 | sp = text; |
| 620 | sp.remove_suffix(1); |
| 621 | okay &= TestInputInContext(sp, text); |
| 622 | } |
| 623 | return okay; |
| 624 | } |
| 625 | |
| 626 | bool Tester::TestInputInContext(const StringPiece& text, |
| 627 | const StringPiece& context) { |
| 628 | bool okay = true; |
| 629 | for (int i = 0; i < arraysize(anchors); i++) |
| 630 | okay &= TestCase(text, context, anchors[i]); |
| 631 | return okay; |
| 632 | } |
| 633 | |
| 634 | bool TestRegexpOnText(const StringPiece& regexp, |
| 635 | const StringPiece& text) { |
| 636 | Tester t(regexp); |
| 637 | return t.TestInput(text); |
| 638 | } |
| 639 | |
| 640 | } // namespace re2 |