///////////////////////////////////////////////////////////////////////////////
// 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 <ballueder@gmx.net>
+// Copyright: (c) 2000 Karsten Ballueder <ballueder@gmx.net>
// 2001 Vadim Zeitlin <vadim@wxwindows.org>
// Licence: wxWindows licence
///////////////////////////////////////////////////////////////////////////////
#if wxUSE_REGEX
+#include "wx/regex.h"
+
#ifndef WX_PRECOMP
#include "wx/object.h"
- #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.
#endif
#include <regex.h>
-#include "wx/regex.h"
-// defined when the regex lib uses 'char' but 'wxChar' is wide
-#if wxUSE_UNICODE && !defined(__REG_NOFRONT)
-# define WX_NEED_CONVERT
+// 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_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
+# 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 WX_NEED_CONVERT
+#ifndef WXREGEX_CONVERT_TO_MB
typedef wxChar wxRegChar;
#else
typedef char wxRegChar;
// RE operations
bool Compile(const wxString& expr, int flags = 0);
- bool Matches(const wxRegChar *str, int flags, size_t len) const;
+ bool Matches(const wxRegChar *str, int flags
+ 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,
{
if ( IsValid() )
{
- regfree(&m_RegEx);
+ wx_regfree(&m_RegEx);
}
- delete [] m_Matches;
+ delete m_Matches;
}
// free the RE if any and reinit the members
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
// ============================================================================
wxString wxRegExImpl::GetErrorMsg(int errorcode, bool badconv) const
{
-#ifdef WX_NEED_CONVERT
+#ifdef WXREGEX_CONVERT_TO_MB
// currently only needed when using system library in Unicode mode
if ( badconv )
{
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
#else
# define FLAVORS (wxRE_ADVANCED | wxRE_BASIC)
wxASSERT_MSG( (flags & FLAVORS) != FLAVORS,
- _T("incompatible flags in wxRegEx::Compile") );
+ wxT("incompatible flags in wxRegEx::Compile") );
#endif
wxASSERT_MSG( !(flags & ~(FLAVORS | wxRE_ICASE | wxRE_NOSUB | wxRE_NEWLINE)),
- _T("unrecognized flags in wxRegEx::Compile") );
+ wxT("unrecognized flags in wxRegEx::Compile") );
// 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 )
flagsRE |= REG_NEWLINE;
// compile it
-#ifdef __REG_NOFRONT
+#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
// and some more for bracketed subexperessions
for ( const wxChar *cptr = expr.c_str(); *cptr; cptr++ )
{
- if ( *cptr == _T('\\') )
+ if ( *cptr == wxT('\\') )
{
// in basic RE syntax groups are inside \(...\)
- if ( *++cptr == _T('(') && (flags & wxRE_BASIC) )
+ if ( *++cptr == wxT('(') && (flags & wxRE_BASIC) )
{
m_nMatches++;
}
}
- else if ( *cptr == _T('(') && !(flags & wxRE_BASIC) )
+ else if ( *cptr == wxT('(') && !(flags & wxRE_BASIC) )
{
// we know that the previous character is not an unquoted
// backslash because it would have been eaten above, so we
// extended syntax. '(?' is used for extensions by perl-
// like REs (e.g. advanced), and is not valid for POSIX
// extended, so ignore them always.
- if ( cptr[1] != _T('?') )
+ if ( cptr[1] != wxT('?') )
m_nMatches++;
}
}
return IsValid();
}
-bool wxRegExImpl::Matches(const wxRegChar *str, int flags, size_t len) const
+#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 = 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_IF_NEED_LEN(size_t len)) const
{
- wxCHECK_MSG( IsValid(), false, _T("must successfully Compile() first") );
+ wxCHECK_MSG( IsValid(), false, wxT("must successfully Compile() first") );
// translate our flags to regexec() ones
wxASSERT_MSG( !(flags & ~(wxRE_NOTBOL | wxRE_NOTEOL)),
- _T("unrecognized flags in wxRegEx::Matches") );
+ wxT("unrecognized flags in wxRegEx::Matches") );
int flagsRE = 0;
if ( flags & wxRE_NOTBOL )
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 __REG_NOFRONT
- 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 )
bool wxRegExImpl::GetMatch(size_t *start, size_t *len, size_t index) const
{
- wxCHECK_MSG( IsValid(), false, _T("must successfully Compile() first") );
- wxCHECK_MSG( m_nMatches, false, _T("can't use with wxRE_NOSUB") );
- wxCHECK_MSG( m_Matches, false, _T("must call Matches() first") );
- wxCHECK_MSG( index < m_nMatches, false, _T("invalid match index") );
+ wxCHECK_MSG( IsValid(), false, wxT("must successfully Compile() first") );
+ wxCHECK_MSG( m_nMatches, false, wxT("can't use with wxRE_NOSUB") );
+ wxCHECK_MSG( m_Matches, false, wxT("must call Matches() first") );
+ wxCHECK_MSG( index < m_nMatches, false, wxT("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;
}
size_t wxRegExImpl::GetMatchCount() const
{
- wxCHECK_MSG( IsValid(), 0, _T("must successfully Compile() first") );
- wxCHECK_MSG( m_nMatches, 0, _T("can't use with wxRE_NOSUB") );
+ wxCHECK_MSG( IsValid(), 0, wxT("must successfully Compile() first") );
+ wxCHECK_MSG( m_nMatches, 0, wxT("can't use with wxRE_NOSUB") );
return m_nMatches;
}
const wxString& replacement,
size_t maxMatches) const
{
- wxCHECK_MSG( text, wxNOT_FOUND, _T("NULL text in wxRegEx::Replace") );
- wxCHECK_MSG( IsValid(), wxNOT_FOUND, _T("must successfully Compile() first") );
+ wxCHECK_MSG( text, wxNOT_FOUND, wxT("NULL text in wxRegEx::Replace") );
+ wxCHECK_MSG( IsValid(), wxNOT_FOUND, wxT("must successfully Compile() first") );
// the input string
-#ifndef WX_NEED_CONVERT
+#ifndef WXREGEX_CONVERT_TO_MB
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"),
// attempt at optimization: don't iterate over the string if it doesn't
// contain back references at all
bool mayHaveBackrefs =
- replacement.find_first_of(_T("\\&")) != wxString::npos;
+ replacement.find_first_of(wxT("\\&")) != wxString::npos;
if ( !mayHaveBackrefs )
{
// 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) &&
- Matches(textstr + matchStart,
- countRepl ? wxRE_NOTBOL : 0,
- textlen - matchStart) )
+ Matches(
+#ifndef WXREGEX_CONVERT_TO_MB
+ textstr + matchStart,
+#else
+ textstr.data() + matchStart,
+#endif
+ countRepl ? wxRE_NOTBOL : 0
+ WXREGEX_IF_NEED_LEN(textlen - matchStart)) )
{
// the string possibly contains back references: we need to calculate
// the replacement text anew after each match
{
size_t index = (size_t)-1;
- if ( *p == _T('\\') )
+ if ( *p == wxT('\\') )
{
if ( wxIsdigit(*++p) )
{
}
//else: backslash used as escape character
}
- else if ( *p == _T('&') )
+ else if ( *p == wxT('&') )
{
// treat this as "\0" for compatbility with ed and such
index = 0;
size_t start, len;
if ( !GetMatch(&start, &len, index) )
{
- wxFAIL_MSG( _T("invalid back reference") );
+ wxFAIL_MSG( wxT("invalid back reference") );
// just eat it...
}
else
{
- textNew += wxString(textstr + matchStart + start,
- *wxConvCurrent, len);
+ textNew += wxString(
+#ifndef WXREGEX_CONVERT_TO_MB
+ textstr
+#else
+ textstr.data()
+#endif
+ + matchStart + start,
+ *wxConvCurrent, len);
mayHaveBackrefs = true;
}
if ( !GetMatch(&start, &len) )
{
// we did have match as Matches() returned true above!
- wxFAIL_MSG( _T("internal logic error in wxRegEx::Replace") );
+ wxFAIL_MSG( wxT("internal logic error in wxRegEx::Replace") );
return wxNOT_FOUND;
}
if (result.capacity() < result.length() + start + textNew.length())
result.reserve(2 * result.length());
-#ifndef WX_NEED_CONVERT
+#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);
matchStart += len;
}
-#ifndef WX_NEED_CONVERT
+#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;
return true;
}
-bool wxRegEx::Matches(const wxChar *str, int flags) const
+bool wxRegEx::Matches(const wxString& str, int flags) const
{
- wxCHECK_MSG( IsValid(), false, _T("must successfully Compile() first") );
+ wxCHECK_MSG( IsValid(), false, wxT("must successfully Compile() first") );
-#ifndef WX_NEED_CONVERT
- return m_impl->Matches(str, flags, wxStrlen(str));
-#else
- return m_impl->Matches(wxConvertWX2MB(str), flags, 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
{
- wxCHECK_MSG( IsValid(), false, _T("must successfully Compile() first") );
+ wxCHECK_MSG( IsValid(), false, wxT("must successfully Compile() first") );
return m_impl->GetMatch(start, len, index);
}
size_t wxRegEx::GetMatchCount() const
{
- wxCHECK_MSG( IsValid(), 0, _T("must successfully Compile() first") );
+ wxCHECK_MSG( IsValid(), 0, wxT("must successfully Compile() first") );
return m_impl->GetMatchCount();
}
const wxString& replacement,
size_t maxMatches) const
{
- wxCHECK_MSG( IsValid(), wxNOT_FOUND, _T("must successfully Compile() first") );
+ wxCHECK_MSG( IsValid(), wxNOT_FOUND, wxT("must successfully Compile() first") );
return m_impl->Replace(pattern, replacement, maxMatches);
}