X-Git-Url: https://git.saurik.com/wxWidgets.git/blobdiff_plain/c9eee7f035ebc6b5c044475d508b018ae9fb5bc2..7f6cbcead99e018aa640fca5ceb30054f2513243:/src/common/regex.cpp diff --git a/src/common/regex.cpp b/src/common/regex.cpp index b3c78820df..f6ca878291 100644 --- a/src/common/regex.cpp +++ b/src/common/regex.cpp @@ -1,11 +1,11 @@ /////////////////////////////////////////////////////////////////////////////// // Name: src/common/regex.cpp // Purpose: regular expression matching -// Author: Karsten Ballüder and Vadim Zeitlin +// Author: Karsten Ballueder and Vadim Zeitlin // Modified by: // Created: 13.07.01 // RCS-ID: $Id$ -// Copyright: (c) 2000 Karsten Ballüder +// Copyright: (c) 2000 Karsten Ballueder // 2001 Vadim Zeitlin // Licence: wxWindows licence /////////////////////////////////////////////////////////////////////////////// @@ -32,6 +32,7 @@ #include "wx/string.h" #include "wx/log.h" #include "wx/intl.h" + #include "wx/crt.h" #endif //WX_PRECOMP // FreeBSD, Watcom and DMars require this, CW doesn't have nor need it. @@ -45,24 +46,104 @@ #include #include "wx/regex.h" -// WXREGEX_USING_BUILTIN defined when using the built-in regex lib -// WXREGEX_BUILTIN_ONLY() wrap a parameter only used with the built-in regex -// WXREGEX_CONVERT_TO_MB indicates when the regex lib is using chars and -// wxChar is wide, so conversion must be done +// WXREGEX_USING_BUILTIN defined when using the built-in regex lib +// WXREGEX_USING_RE_SEARCH defined when using re_search in the GNU regex lib +// WXREGEX_IF_NEED_LEN() wrap the len parameter only used with the built-in +// or GNU regex +// WXREGEX_CONVERT_TO_MB defined when the regex lib is using chars and +// wxChar is wide, so conversion must be done +// WXREGEX_CHAR(x) Convert wxChar to wxRegChar +// #ifdef __REG_NOFRONT # define WXREGEX_USING_BUILTIN -# define WXREGEX_BUILTIN_ONLY(x) ,x +# define WXREGEX_IF_NEED_LEN(x) ,x +# if wxUSE_UNICODE +# define WXREGEX_CHAR(x) (x).wc_str() +# else +# define WXREGEX_CHAR(x) (x).mb_str() +# endif #else -# define WXREGEX_BUILTIN_ONLY(x) +# ifdef HAVE_RE_SEARCH +# define WXREGEX_IF_NEED_LEN(x) ,x +# define WXREGEX_USING_RE_SEARCH +# else +# define WXREGEX_IF_NEED_LEN(x) +# endif # if wxUSE_UNICODE # define WXREGEX_CONVERT_TO_MB # endif +# define WXREGEX_CHAR(x) (x).mb_str() +# define wx_regfree regfree +# define wx_regerror regerror #endif // ---------------------------------------------------------------------------- // private classes // ---------------------------------------------------------------------------- +#ifndef WXREGEX_USING_RE_SEARCH + +// the array of offsets for the matches, the usual POSIX regmatch_t array. +class wxRegExMatches +{ +public: + typedef regmatch_t *match_type; + + wxRegExMatches(size_t n) { m_matches = new regmatch_t[n]; } + ~wxRegExMatches() { delete [] m_matches; } + + // we just use casts here because the fields of regmatch_t struct may be 64 + // bit but we're limited to size_t in our public API and are not going to + // change it because operating on strings longer than 4GB using it is + // absolutely impractical anyhow + size_t Start(size_t n) const + { + return wx_truncate_cast(size_t, m_matches[n].rm_so); + } + + size_t End(size_t n) const + { + return wx_truncate_cast(size_t, m_matches[n].rm_eo); + } + + regmatch_t *get() const { return m_matches; } + +private: + regmatch_t *m_matches; +}; + +#else // WXREGEX_USING_RE_SEARCH + +// the array of offsets for the matches, the struct used by the GNU lib +class wxRegExMatches +{ +public: + typedef re_registers *match_type; + + wxRegExMatches(size_t n) + { + m_matches.num_regs = n; + m_matches.start = new regoff_t[n]; + m_matches.end = new regoff_t[n]; + } + + ~wxRegExMatches() + { + delete [] m_matches.start; + delete [] m_matches.end; + } + + size_t Start(size_t n) const { return m_matches.start[n]; } + size_t End(size_t n) const { return m_matches.end[n]; } + + re_registers *get() { return &m_matches; } + +private: + re_registers m_matches; +}; + +#endif // WXREGEX_USING_RE_SEARCH + // the character type used by the regular expression engine #ifndef WXREGEX_CONVERT_TO_MB typedef wxChar wxRegChar; @@ -84,7 +165,7 @@ public: // RE operations bool Compile(const wxString& expr, int flags = 0); bool Matches(const wxRegChar *str, int flags - WXREGEX_BUILTIN_ONLY(size_t len)) const; + WXREGEX_IF_NEED_LEN(size_t len)) const; bool GetMatch(size_t *start, size_t *len, size_t index = 0) const; size_t GetMatchCount() const; int Replace(wxString *pattern, const wxString& replacement, @@ -107,10 +188,10 @@ private: { if ( IsValid() ) { - regfree(&m_RegEx); + wx_regfree(&m_RegEx); } - delete [] m_Matches; + delete m_Matches; } // free the RE if any and reinit the members @@ -120,18 +201,18 @@ private: Init(); } - // compiled RE - regex_t m_RegEx; + regex_t m_RegEx; // the subexpressions data - regmatch_t *m_Matches; - size_t m_nMatches; + wxRegExMatches *m_Matches; + size_t m_nMatches; // true if m_RegEx is valid - bool m_isCompiled; + bool m_isCompiled; }; + // ============================================================================ // implementation // ============================================================================ @@ -166,14 +247,14 @@ wxString wxRegExImpl::GetErrorMsg(int errorcode, bool badconv) const wxString szError; // first get the string length needed - int len = regerror(errorcode, &m_RegEx, NULL, 0); + int len = wx_regerror(errorcode, &m_RegEx, NULL, 0); if ( len > 0 ) { char* szcmbError = new char[++len]; - (void)regerror(errorcode, &m_RegEx, szcmbError, len); + (void)wx_regerror(errorcode, &m_RegEx, szcmbError, len); - szError = wxConvertMB2WX(szcmbError); + szError = wxConvLibc.cMB2WX(szcmbError); delete [] szcmbError; } else // regerror() returned 0 @@ -201,12 +282,14 @@ bool wxRegExImpl::Compile(const wxString& expr, int flags) // translate our flags to regcomp() ones int flagsRE = 0; if ( !(flags & wxRE_BASIC) ) + { #ifndef WX_NO_REGEX_ADVANCED if (flags & wxRE_ADVANCED) flagsRE |= REG_ADVANCED; else #endif flagsRE |= REG_EXTENDED; + } if ( flags & wxRE_ICASE ) flagsRE |= REG_ICASE; if ( flags & wxRE_NOSUB ) @@ -217,8 +300,11 @@ bool wxRegExImpl::Compile(const wxString& expr, int flags) // compile it #ifdef WXREGEX_USING_BUILTIN bool conv = true; - int errorcode = wx_re_comp(&m_RegEx, expr, expr.length(), flagsRE); + // FIXME-UTF8: use wc_str() after removing ANSI build + int errorcode = wx_re_comp(&m_RegEx, expr.c_str(), expr.length(), flagsRE); #else + // FIXME-UTF8: this is potentially broken, we shouldn't even try it + // and should always use builtin regex library (or PCRE?) const wxWX2MBbuf conv = expr.mbc_str(); int errorcode = conv ? regcomp(&m_RegEx, conv, flagsRE) : REG_BADPAT; #endif @@ -277,9 +363,36 @@ bool wxRegExImpl::Compile(const wxString& expr, int flags) return IsValid(); } +#ifdef WXREGEX_USING_RE_SEARCH + +// On GNU, regexec is implemented as a wrapper around re_search. re_search +// requires a length parameter which the POSIX regexec does not have, +// therefore regexec must do a strlen on the search text each time it is +// called. This can drastically affect performance when matching is done in +// a loop along a string, such as during a search and replace. Therefore if +// re_search is detected by configure, it is used directly. +// +static int ReSearch(const regex_t *preg, + const char *text, + size_t len, + re_registers *matches, + int eflags) +{ + regex_t *pattern = wx_const_cast(regex_t*, preg); + + pattern->not_bol = (eflags & REG_NOTBOL) != 0; + pattern->not_eol = (eflags & REG_NOTEOL) != 0; + pattern->regs_allocated = REGS_FIXED; + + int ret = re_search(pattern, text, len, 0, len, matches); + return ret >= 0 ? 0 : REG_NOMATCH; +} + +#endif // WXREGEX_USING_RE_SEARCH + bool wxRegExImpl::Matches(const wxRegChar *str, int flags - WXREGEX_BUILTIN_ONLY(size_t len)) const + WXREGEX_IF_NEED_LEN(size_t len)) const { wxCHECK_MSG( IsValid(), false, _T("must successfully Compile() first") ); @@ -297,14 +410,18 @@ bool wxRegExImpl::Matches(const wxRegChar *str, wxRegExImpl *self = wxConstCast(this, wxRegExImpl); if ( !m_Matches && m_nMatches ) { - self->m_Matches = new regmatch_t[m_nMatches]; + self->m_Matches = new wxRegExMatches(m_nMatches); } + wxRegExMatches::match_type matches = m_Matches ? m_Matches->get() : NULL; + // do match it -#ifdef WXREGEX_USING_BUILTIN - int rc = wx_re_exec(&self->m_RegEx, str, len, NULL, m_nMatches, m_Matches, flagsRE); +#if defined WXREGEX_USING_BUILTIN + int rc = wx_re_exec(&self->m_RegEx, str, len, NULL, m_nMatches, matches, flagsRE); +#elif defined WXREGEX_USING_RE_SEARCH + int rc = str ? ReSearch(&self->m_RegEx, str, len, matches, flagsRE) : REG_BADPAT; #else - int rc = str ? regexec(&self->m_RegEx, str, m_nMatches, m_Matches, flagsRE) : REG_BADPAT; + int rc = str ? regexec(&self->m_RegEx, str, m_nMatches, matches, flagsRE) : REG_BADPAT; #endif switch ( rc ) @@ -332,13 +449,10 @@ bool wxRegExImpl::GetMatch(size_t *start, size_t *len, size_t index) const wxCHECK_MSG( m_Matches, false, _T("must call Matches() first") ); wxCHECK_MSG( index < m_nMatches, false, _T("invalid match index") ); - const regmatch_t& match = m_Matches[index]; - - // we need the casts because rm_so can be a 64 bit quantity if ( start ) - *start = wx_truncate_cast(size_t, match.rm_so); + *start = m_Matches->Start(index); if ( len ) - *len = wx_truncate_cast(size_t, match.rm_eo - match.rm_so); + *len = m_Matches->End(index) - m_Matches->Start(index); return true; } @@ -363,7 +477,7 @@ int wxRegExImpl::Replace(wxString *text, const wxChar *textstr = text->c_str(); size_t textlen = text->length(); #else - const wxWX2MBbuf textstr = wxConvertWX2MB(*text); + const wxWX2MBbuf textstr = WXREGEX_CHAR(*text); if (!textstr) { wxLogError(_("Failed to find match for regular expression: %s"), @@ -401,9 +515,13 @@ int wxRegExImpl::Replace(wxString *text, // note that "^" shouldn't match after the first call to Matches() so we // use wxRE_NOTBOL to prevent it from happening while ( (!maxMatches || countRepl < maxMatches) && +#ifndef WXREGEX_CONVERT_TO_MB Matches(textstr + matchStart, +#else + Matches(textstr.data() + matchStart, +#endif countRepl ? wxRE_NOTBOL : 0 - WXREGEX_BUILTIN_ONLY(textlen - matchStart)) ) + WXREGEX_IF_NEED_LEN(textlen - matchStart)) ) { // the string possibly contains back references: we need to calculate // the replacement text anew after each match @@ -447,7 +565,12 @@ int wxRegExImpl::Replace(wxString *text, } else { +#ifndef WXREGEX_CONVERT_TO_MB textNew += wxString(textstr + matchStart + start, +#else + textNew += wxString(textstr.data() + matchStart + + start, +#endif *wxConvCurrent, len); mayHaveBackrefs = true; @@ -477,7 +600,8 @@ int wxRegExImpl::Replace(wxString *text, #ifndef WXREGEX_CONVERT_TO_MB result.append(*text, matchStart, start); #else - result.append(wxString(textstr + matchStart, *wxConvCurrent, start)); + result.append(wxString(textstr.data() + matchStart, *wxConvCurrent, + start)); #endif matchStart += start; result.append(textNew); @@ -490,7 +614,7 @@ int wxRegExImpl::Replace(wxString *text, #ifndef WXREGEX_CONVERT_TO_MB result.append(*text, matchStart, wxString::npos); #else - result.append(wxString(textstr + matchStart, *wxConvCurrent)); + result.append(wxString(textstr.data() + matchStart, *wxConvCurrent)); #endif *text = result; @@ -530,27 +654,12 @@ bool wxRegEx::Compile(const wxString& expr, int flags) return true; } -bool wxRegEx::Matches(const wxChar *str, int flags, size_t len) const +bool wxRegEx::Matches(const wxString& str, int flags) const { wxCHECK_MSG( IsValid(), false, _T("must successfully Compile() first") ); - (void)len; -#ifdef WXREGEX_CONVERT_TO_MB - return m_impl->Matches(wxConvertWX2MB(str), flags); -#else - return m_impl->Matches(str, flags WXREGEX_BUILTIN_ONLY(len)); -#endif -} - -bool wxRegEx::Matches(const wxChar *str, int flags) const -{ - wxCHECK_MSG( IsValid(), false, _T("must successfully Compile() first") ); - -#ifdef WXREGEX_CONVERT_TO_MB - return m_impl->Matches(wxConvertWX2MB(str), flags); -#else - return m_impl->Matches(str, flags WXREGEX_BUILTIN_ONLY(wxStrlen(str))); -#endif + return m_impl->Matches(WXREGEX_CHAR(str), flags + WXREGEX_IF_NEED_LEN(str.length())); } bool wxRegEx::GetMatch(size_t *start, size_t *len, size_t index) const