search for: llvm_regmatch_t

...compiled with NEWLINE flag. > + // noteol: The match-end-of-line operator always fails to match. > + // unless regex was compiled with NEWLINE flag. > + > + bool matches(const char *string, bool notbol=false, bool noteol=false); > + bool match_sub(const char *string, llvm_regmatch_t pmatch[], > + unsigned nmatch, bool notbol=false, bool noteol=false); These should be doxymented, and I don't understand the flags. Does this mean that if I *don't* compile with NEWLINE, but *do* want ^/$ support, then notbol/noteol need to be false? Do we care about u...

...NE flag. >> + // noteol: The match-end-of-line operator always fails to match. >> + // unless regex was compiled with NEWLINE flag. >> + >> + bool matches(const char *string, bool notbol=false, bool noteol=false); >> + bool match_sub(const char *string, llvm_regmatch_t pmatch[], >> + unsigned nmatch, bool notbol=false, bool noteol=false); >> > > These should be doxymented, and I don't understand the flags. Does this mean > that if I *don't* compile with NEWLINE, but *do* want ^/$ support, then > notbol/noteol...

...2009, at 1:53 AM, Daniel Dunbar wrote: > Nice! > > This looks good to me but probably Chris or someone else should sign > off on it. This seems ok to me, please commit. One minor comment: +++ b/lib/Support/Regex.cpp + // Allocate pmatch with at least one element. + pmatch = new llvm_regmatch_t[nmatch > 0 ? nmatch : 1]; + pmatch[0].rm_so = 0; + pmatch[0].rm_eo = String.size(); Can this be a smallvector to avoid a heap allocation on every match? -Chris

On 2009-08-24 20:14, Chris Lattner wrote: > On Aug 23, 2009, at 11:59 PM, Török Edwin wrote: >> If LLVM is going to have an integrated regex library I suggest using it >> regardless if the platform has one. >> The LLVM integrated regex library will provide consistent behaviour and >> execution time, the system one will not. > > Hi Edwin, > > Can you propose

...-of-line operator always fails to match. >>>>> +    //    unless regex was compiled with NEWLINE flag. >>>>> + >>>>> +    bool matches(const char *string, bool notbol=false, bool noteol=false); >>>>> +    bool match_sub(const char *string, llvm_regmatch_t pmatch[], >>>>> +                   unsigned nmatch, bool notbol=false, bool noteol=false); >>>>> >>>>> >>>> These should be doxymented, and I don't understand the flags. Does this mean >>>> that if I *don't* compile with...

On Aug 23, 2009, at 11:59 PM, Török Edwin wrote: > If LLVM is going to have an integrated regex library I suggest using > it > regardless if the platform has one. > The LLVM integrated regex library will provide consistent behaviour > and > execution time, the system one will not. Hi Edwin, Can you propose the openbsd implementation as a patch to lib/support? -Chris

...1]: Compiling circular_raw_ostream.cpp for Release build llvm[1]: Compiling raw_os_ostream.cpp for Release build llvm[1]: Compiling raw_ostream.cpp for Release build llvm[1]: Compiling regcomp.c for Release build UNREACHABLE executed! 0 clang 0x0000000100c268b2 llvm::SmallVectorImpl<llvm_regmatch_t>::resize(unsigned int) + 2098 1 clang 0x0000000100c26d89 llvm::SmallVectorImpl<llvm_regmatch_t>::resize(unsigned int) + 3337 2 libSystem.B.dylib 0x00007fff86e7367a _sigtramp + 26 3 libSystem.B.dylib 0x8000006300000001 _sigtramp + 2031667617 4 clang 0x0000000100c...

...; +    // noteol:  The match-end-of-line operator always fails to match. >>> +    //    unless regex was compiled with NEWLINE flag. >>> + >>> +    bool matches(const char *string, bool notbol=false, bool noteol=false); >>> +    bool match_sub(const char *string, llvm_regmatch_t pmatch[], >>> +                   unsigned nmatch, bool notbol=false, bool noteol=false); >>> >> >> These should be doxymented, and I don't understand the flags. Does this mean >> that if I *don't* compile with NEWLINE, but *do* want ^/$ support, then &gt...

...: The match-end-of-line operator always fails to match. >>>> + // unless regex was compiled with NEWLINE flag. >>>> + >>>> + bool matches(const char *string, bool notbol=false, bool noteol=false); >>>> + bool match_sub(const char *string, llvm_regmatch_t pmatch[], >>>> + unsigned nmatch, bool notbol=false, bool noteol=false); >>>> >>>> >>> These should be doxymented, and I don't understand the flags. Does this mean >>> that if I *don't* compile with NEWLINE, bu...

...it, failing here: G:\public\zorg\buildbot\osuosl\slave\clang-i686-xp-msvc9\llvm\include\llvm/Support/IRBuilder.h(133) : error C3861: 'InsertHelper': identifier not found > > +++ b/lib/Support/Regex.cpp > > + // Allocate pmatch with at least one element. > + pmatch = new llvm_regmatch_t[nmatch > 0 ? nmatch : 1]; > + pmatch[0].rm_so = 0; > + pmatch[0].rm_eo = String.size(); > > Can this be a smallvector to avoid a heap allocation on every match? Yes, I changed it to use SmallVector. Best regards, --Edwin