1 //===-- Regex.cpp - Regular Expression matcher implementation -------------===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 // 10 // This file implements a POSIX regular expression matcher. 11 // 12 //===----------------------------------------------------------------------===// 13 #include "llvm/Support/Regex.h" 14 #include "llvm/Support/ErrorHandling.h" 15 #include "llvm/Support/raw_ostream.h" 16 #include "regex_impl.h" 17 #include <string> 18 19 using namespace llvm; 20 Regex::Regex(const StringRef ®ex, unsigned Flags) 21 { 22 unsigned flags = 0; 23 preg = new struct llvm_regex; 24 preg->re_endp = regex.end(); 25 if (Flags & IgnoreCase) 26 flags |= REG_ICASE; 27 if (Flags & NoSub) { 28 flags |= REG_NOSUB; 29 sub = false; 30 } else { 31 sub = true; 32 } 33 if (Flags & Newline) 34 flags |= REG_NEWLINE; 35 error = llvm_regcomp(preg, regex.data(), flags|REG_EXTENDED|REG_PEND); 36 } 37 38 bool Regex::isValid(std::string &Error) 39 { 40 if (!error) 41 return true; 42 43 size_t len = llvm_regerror(error, preg, NULL, 0); 44 char *errbuff = new char[len]; 45 llvm_regerror(error, preg, errbuff, len); 46 Error.assign(errbuff); 47 return false; 48 } 49 50 Regex::~Regex() 51 { 52 llvm_regfree(preg); 53 delete preg; 54 } 55 56 bool Regex::match(const StringRef &String, SmallVectorImpl<StringRef> *Matches) 57 { 58 unsigned nmatch = Matches ? preg->re_nsub+1 : 0; 59 60 if (Matches) { 61 assert(sub && "Substring matching requested but pattern compiled without"); 62 Matches->clear(); 63 } 64 65 // pmatch needs to have at least one element. 66 SmallVector<llvm_regmatch_t, 2> pm; 67 pm.resize(nmatch > 0 ? nmatch : 1); 68 pm[0].rm_so = 0; 69 pm[0].rm_eo = String.size(); 70 71 int rc = llvm_regexec(preg, String.data(), nmatch, pm.data(), REG_STARTEND); 72 73 if (rc == REG_NOMATCH) 74 return false; 75 if (rc != 0) { 76 // regexec can fail due to invalid pattern or running out of memory. 77 error = rc; 78 return false; 79 } 80 81 // There was a match. 82 83 if (Matches) { // match position requested 84 for (unsigned i=0;i<nmatch; i++) { 85 if (pm[i].rm_so == -1) { 86 // this group didn't match 87 Matches->push_back(StringRef()); 88 continue; 89 } 90 assert(pm[i].rm_eo > pm[i].rm_so); 91 Matches->push_back(StringRef(String.data()+pm[i].rm_so, 92 pm[i].rm_eo-pm[i].rm_so)); 93 } 94 } 95 96 return true; 97 } 98