// headers
// ----------------------------------------------------------------------------
-#ifdef __GNUG__
- #pragma implementation "regex.h"
-#endif
-
// For compilers that support precompilation, includes "wx.h".
#include "wx/wxprec.h"
#include "wx/intl.h"
#endif //WX_PRECOMP
-// FreeBSD & Watcom require this, it probably doesn't hurt for others
-#if defined(__UNIX__) || defined(__WATCOMC__)
- #include <sys/types.h>
+// FreeBSD, Watcom and DMars require this, CW doesn't have nor need it.
+// Others also don't seem to need it. If you have an error related to
+// (not) including <sys/types.h> please report details to
+// wx-dev@lists.wxwindows.org
+#if defined(__UNIX__) || defined(__WATCOMC__) || defined(__DIGITALMARS__)
+# include <sys/types.h>
#endif
#include <regex.h>
-
#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
+#ifdef __REG_NOFRONT
+# define WXREGEX_USING_BUILTIN
+# define WXREGEX_BUILTIN_ONLY(x) ,x
+#else
+# define WXREGEX_BUILTIN_ONLY(x)
+# if wxUSE_UNICODE
+# define WXREGEX_CONVERT_TO_MB
+# endif
+#endif
+
// ----------------------------------------------------------------------------
// private classes
// ----------------------------------------------------------------------------
+// the character type used by the regular expression engine
+#ifndef WXREGEX_CONVERT_TO_MB
+typedef wxChar wxRegChar;
+#else
+typedef char wxRegChar;
+#endif
+
// the real implementation of wxRegEx
class wxRegExImpl
{
wxRegExImpl();
~wxRegExImpl();
- // return TRUE if Compile() had been called successfully
+ // return true if Compile() had been called successfully
bool IsValid() const { return m_isCompiled; }
// RE operations
bool Compile(const wxString& expr, int flags = 0);
- bool Matches(const wxChar *str, int flags = 0) const;
+ bool Matches(const wxRegChar *str, int flags
+ WXREGEX_BUILTIN_ONLY(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,
size_t maxMatches = 0) const;
private:
// return the string containing the error message for the given err code
- wxString GetErrorMsg(int errorcode) const;
+ wxString GetErrorMsg(int errorcode, bool badconv) const;
+
+ // init the members
+ void Init()
+ {
+ m_isCompiled = false;
+ m_Matches = NULL;
+ m_nMatches = 0;
+ }
// free the RE if compiled
void Free()
if ( IsValid() )
{
regfree(&m_RegEx);
-
- m_isCompiled = FALSE;
}
+
+ delete [] m_Matches;
}
+ // free the RE if any and reinit the members
+ void Reinit()
+ {
+ Free();
+ Init();
+ }
+
+
// compiled RE
regex_t m_RegEx;
regmatch_t *m_Matches;
size_t m_nMatches;
- // TRUE if m_RegEx is valid
+ // true if m_RegEx is valid
bool m_isCompiled;
};
wxRegExImpl::wxRegExImpl()
{
- m_isCompiled = FALSE;
- m_Matches = NULL;
- m_nMatches = 0;
+ Init();
}
wxRegExImpl::~wxRegExImpl()
{
Free();
-
- delete [] m_Matches;
}
-wxString wxRegExImpl::GetErrorMsg(int errorcode) const
+wxString wxRegExImpl::GetErrorMsg(int errorcode, bool badconv) const
{
- wxString msg;
+#ifdef WXREGEX_CONVERT_TO_MB
+ // currently only needed when using system library in Unicode mode
+ if ( badconv )
+ {
+ return _("conversion to 8-bit encoding failed");
+ }
+#else
+ // 'use' badconv to avoid a compiler warning
+ (void)badconv;
+#endif
+
+ wxString szError;
// first get the string length needed
int len = regerror(errorcode, &m_RegEx, NULL, 0);
if ( len > 0 )
{
- len++;
-
-#if wxUSE_UNICODE
- wxCharBuffer buf(len);
+ char* szcmbError = new char[++len];
- (void)regerror(errorcode, &m_RegEx, (char *)buf.data(), len);
+ (void)regerror(errorcode, &m_RegEx, szcmbError, len);
- msg = wxString(buf.data(), wxConvLibc);
-#else // !Unicode
- (void)regerror(errorcode, &m_RegEx, msg.GetWriteBuf(len), len);
-
- msg.UngetWriteBuf();
-#endif // Unicode/!Unicode
+ szError = wxConvertMB2WX(szcmbError);
+ delete [] szcmbError;
}
else // regerror() returned 0
{
- msg = _("unknown error");
+ szError = _("unknown error");
}
- return msg;
+ return szError;
}
bool wxRegExImpl::Compile(const wxString& expr, int flags)
{
- Free();
-
- // translate our flags to regcomp() ones
- wxASSERT_MSG( !(flags &
- ~(wxRE_BASIC | wxRE_ICASE | wxRE_NOSUB | wxRE_NEWLINE)),
+ Reinit();
+
+#ifdef WX_NO_REGEX_ADVANCED
+# define FLAVORS wxRE_BASIC
+#else
+# define FLAVORS (wxRE_ADVANCED | wxRE_BASIC)
+ wxASSERT_MSG( (flags & FLAVORS) != FLAVORS,
+ _T("incompatible flags in wxRegEx::Compile") );
+#endif
+ wxASSERT_MSG( !(flags & ~(FLAVORS | wxRE_ICASE | wxRE_NOSUB | wxRE_NEWLINE)),
_T("unrecognized flags in wxRegEx::Compile") );
+ // translate our flags to regcomp() ones
int flagsRE = 0;
if ( !(flags & wxRE_BASIC) )
- flagsRE |= REG_EXTENDED;
+#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
- int errorcode = regcomp(&m_RegEx, expr.mb_str(), flagsRE);
+#ifdef WXREGEX_USING_BUILTIN
+ bool conv = true;
+ int errorcode = wx_re_comp(&m_RegEx, expr, expr.length(), flagsRE);
+#else
+ const wxWX2MBbuf conv = expr.mbc_str();
+ int errorcode = conv ? regcomp(&m_RegEx, conv, flagsRE) : REG_BADPAT;
+#endif
+
if ( errorcode )
{
wxLogError(_("Invalid regular expression '%s': %s"),
- expr.c_str(), GetErrorMsg(errorcode).c_str());
+ expr.c_str(), GetErrorMsg(errorcode, !conv).c_str());
- m_isCompiled = FALSE;
+ m_isCompiled = false;
}
else // ok
{
m_nMatches = 1;
// and some more for bracketed subexperessions
- const wxChar *cptr = expr.c_str();
- wxChar prev = _T('\0');
- while ( *cptr != _T('\0') )
+ for ( const wxChar *cptr = expr.c_str(); *cptr; cptr++ )
{
- // is this a subexpr start, i.e. "(" for extended regex or
- // "\(" for a basic one?
- if ( *cptr == _T('(') &&
- (flags & wxRE_BASIC ? prev == _T('\\')
- : prev != _T('\\')) )
+ if ( *cptr == _T('\\') )
{
- m_nMatches++;
+ // in basic RE syntax groups are inside \(...\)
+ if ( *++cptr == _T('(') && (flags & wxRE_BASIC) )
+ {
+ m_nMatches++;
+ }
+ }
+ else if ( *cptr == _T('(') && !(flags & wxRE_BASIC) )
+ {
+ // we know that the previous character is not an unquoted
+ // backslash because it would have been eaten above, so we
+ // have a bare '(' and this indicates a group start for the
+ // 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('?') )
+ m_nMatches++;
}
-
- prev = *cptr;
- cptr++;
}
}
- m_isCompiled = TRUE;
+ m_isCompiled = true;
}
return IsValid();
}
-bool wxRegExImpl::Matches(const wxChar *str, int flags) const
+bool wxRegExImpl::Matches(const wxRegChar *str,
+ int flags
+ WXREGEX_BUILTIN_ONLY(size_t len)) const
{
- wxCHECK_MSG( IsValid(), FALSE, _T("must successfully Compile() first") );
+ wxCHECK_MSG( IsValid(), false, _T("must successfully Compile() first") );
// translate our flags to regexec() ones
wxASSERT_MSG( !(flags & ~(wxRE_NOTBOL | wxRE_NOTEOL)),
}
// do match it
- int rc = regexec(&self->m_RegEx, wxConvertWX2MB(str), m_nMatches, m_Matches, flagsRE);
+#ifdef WXREGEX_USING_BUILTIN
+ int rc = wx_re_exec(&self->m_RegEx, str, len, NULL, m_nMatches, m_Matches, flagsRE);
+#else
+ int rc = str ? regexec(&self->m_RegEx, str, m_nMatches, m_Matches, flagsRE) : REG_BADPAT;
+#endif
switch ( rc )
{
case 0:
// matched successfully
- return TRUE;
+ return true;
default:
- // an error occured
- wxLogError(_("Failed to match '%s' in regular expression: %s"),
- str, GetErrorMsg(rc).c_str());
+ // an error occurred
+ wxLogError(_("Failed to find match for regular expression: %s"),
+ GetErrorMsg(rc, !str).c_str());
// fall through
case REG_NOMATCH:
// no match
- return FALSE;
+ return false;
}
}
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_Matches, FALSE, _T("can't use with wxRE_NOSUB") );
- wxCHECK_MSG( index < m_nMatches, FALSE, _T("invalid match index") );
+ 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") );
const regmatch_t& match = m_Matches[index];
+ // we need the casts because rm_so can be a 64 bit quantity
if ( start )
- *start = match.rm_so;
+ *start = wx_truncate_cast(size_t, match.rm_so);
if ( len )
- *len = match.rm_eo - match.rm_so;
+ *len = wx_truncate_cast(size_t, match.rm_eo - match.rm_so);
- return TRUE;
+ 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") );
+
+ return m_nMatches;
}
int wxRegExImpl::Replace(wxString *text,
const wxString& replacement,
size_t maxMatches) const
{
- wxCHECK_MSG( text, -1, _T("NULL text in wxRegEx::Replace") );
- wxCHECK_MSG( IsValid(), -1, _T("must successfully Compile() first") );
+ wxCHECK_MSG( text, wxNOT_FOUND, _T("NULL text in wxRegEx::Replace") );
+ wxCHECK_MSG( IsValid(), wxNOT_FOUND, _T("must successfully Compile() first") );
+
+ // the input string
+#ifndef WXREGEX_CONVERT_TO_MB
+ const wxChar *textstr = text->c_str();
+ size_t textlen = text->length();
+#else
+ const wxWX2MBbuf textstr = wxConvertWX2MB(*text);
+ if (!textstr)
+ {
+ wxLogError(_("Failed to find match for regular expression: %s"),
+ GetErrorMsg(0, true).c_str());
+ return 0;
+ }
+ size_t textlen = strlen(textstr);
+ text->clear();
+#endif
// the replacement text
wxString textNew;
+ // the result, allow 25% extra
+ wxString result;
+ result.reserve(5 * textlen / 4);
+
// attempt at optimization: don't iterate over the string if it doesn't
// contain back references at all
bool mayHaveBackrefs =
}
// the position where we start looking for the match
- //
- // NB: initial version had a nasty bug because it used a wxChar* instead of
- // an index but the problem is that replace() in the loop invalidates
- // all pointers into the string so we have to use indices instead
size_t matchStart = 0;
// number of replacement made: we won't make more than maxMatches of them
// 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(text->c_str() + matchStart, countRepl ? wxRE_NOTBOL : 0) )
+ Matches(textstr + matchStart,
+ countRepl ? wxRE_NOTBOL : 0
+ WXREGEX_BUILTIN_ONLY(textlen - matchStart)) )
{
// the string possibly contains back references: we need to calculate
// the replacement text anew after each match
if ( mayHaveBackrefs )
{
- mayHaveBackrefs = FALSE;
+ mayHaveBackrefs = false;
textNew.clear();
textNew.reserve(replacement.length());
}
else
{
- textNew += wxString(text->c_str() + matchStart + start,
- len);
+ textNew += wxString(textstr + matchStart + start,
+ *wxConvCurrent, len);
- mayHaveBackrefs = TRUE;
+ mayHaveBackrefs = true;
}
}
else // ordinary character
// we did have match as Matches() returned true above!
wxFAIL_MSG( _T("internal logic error in wxRegEx::Replace") );
- return -1;
+ return wxNOT_FOUND;
}
+ // an insurance against implementations that don't grow exponentially
+ // to ensure building the result takes linear time
+ if (result.capacity() < result.length() + start + textNew.length())
+ result.reserve(2 * result.length());
+
+#ifndef WXREGEX_CONVERT_TO_MB
+ result.append(*text, matchStart, start);
+#else
+ result.append(wxString(textstr + matchStart, *wxConvCurrent, start));
+#endif
matchStart += start;
- text->replace(matchStart, len, textNew);
+ result.append(textNew);
countRepl++;
- matchStart += textNew.length();
+ matchStart += len;
}
+#ifndef WXREGEX_CONVERT_TO_MB
+ result.append(*text, matchStart, wxString::npos);
+#else
+ result.append(wxString(textstr + matchStart, *wxConvCurrent));
+#endif
+ *text = result;
+
return countRepl;
}
m_impl = NULL;
}
-
wxRegEx::~wxRegEx()
{
delete m_impl;
delete m_impl;
m_impl = NULL;
- return FALSE;
+ return false;
}
- return TRUE;
+ return true;
+}
+
+bool wxRegEx::Matches(const wxChar *str, int flags, size_t len) 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") );
+ wxCHECK_MSG( IsValid(), false, _T("must successfully Compile() first") );
- return m_impl->Matches(str, flags);
+#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
}
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, _T("must successfully Compile() first") );
return m_impl->GetMatch(start, len, index);
}
return text.Mid(start, len);
}
+size_t wxRegEx::GetMatchCount() const
+{
+ wxCHECK_MSG( IsValid(), 0, _T("must successfully Compile() first") );
+
+ return m_impl->GetMatchCount();
+}
+
int wxRegEx::Replace(wxString *pattern,
const wxString& replacement,
size_t maxMatches) const
{
- wxCHECK_MSG( IsValid(), -1, _T("must successfully Compile() first") );
+ wxCHECK_MSG( IsValid(), wxNOT_FOUND, _T("must successfully Compile() first") );
return m_impl->Replace(pattern, replacement, maxMatches);
}