///////////////////////////////////////////////////////////////////////////////
// Name: wx/string.h
-// Purpose: wxString and wxArrayString classes
+// Purpose: wxString class
// Author: Vadim Zeitlin
// Modified by:
// Created: 29/01/98
missing from string.h.
*/
-#ifndef _WX_WXSTRINGH__
-#define _WX_WXSTRINGH__
+#ifndef _WX_WXSTRING_H__
+#define _WX_WXSTRING_H__
// ----------------------------------------------------------------------------
// headers
#include <StringMgr.h>
#endif
-#include "wx/wxchar.h" // for wxChar, wxStrlen() etc.
-#include "wx/unichar.h"
+#include "wx/wxcrtbase.h" // for wxChar, wxStrlen() etc.
#include "wx/strvararg.h"
#include "wx/buffer.h" // for wxCharBuffer
#include "wx/strconv.h" // for wxConvertXXX() macros and wxMBConv classes
+#include "wx/stringimpl.h"
+#include "wx/stringops.h"
+#include "wx/unichar.h"
-class WXDLLIMPEXP_BASE wxString;
+class WXDLLIMPEXP_FWD_BASE wxString;
+
+// unless this symbol is predefined to disable the compatibility functions, do
+// use them
+#ifndef WXWIN_COMPATIBILITY_STRING_PTR_AS_ITER
+ #define WXWIN_COMPATIBILITY_STRING_PTR_AS_ITER 1
+#endif
// ---------------------------------------------------------------------------
// macros
#define wxMBSTRINGCAST (char *)(const char *)
#define wxWCSTRINGCAST (wchar_t *)(const wchar_t *)
-// implementation only
-#define wxASSERT_VALID_INDEX(i) \
- wxASSERT_MSG( (size_t)(i) <= length(), _T("invalid index in wxString") )
+// like _T(), but for literals in wxString's internal representation, i.e.
+// char* in UTF-8 build and wxChar* otherwise:
+#if wxUSE_UNICODE_UTF8
+ #define wxSTRING_TEXT(str) str
+#else
+ #define wxSTRING_TEXT(str) _T(str)
+#endif
// ----------------------------------------------------------------------------
// constants
// deprecated in favour of wxString::npos, don't use in new code
//
// maximum possible length for a string means "take all string" everywhere
-#define wxSTRING_MAXLEN wxStringBase::npos
+#define wxSTRING_MAXLEN wxString::npos
#endif // WXWIN_COMPATIBILITY_2_6
-// ----------------------------------------------------------------------------
-// global data
-// ----------------------------------------------------------------------------
-
-// global pointer to empty string
-extern WXDLLIMPEXP_DATA_BASE(const wxChar*) wxEmptyString;
-
// ---------------------------------------------------------------------------
// global functions complementing standard C string library replacements for
// strlen() and portable strcasecmp()
#endif // WXWIN_COMPATIBILITY_2_8
-// ----------------------------------------------------------------------------
-// deal with STL/non-STL/non-STL-but-wxUSE_STD_STRING
-// ----------------------------------------------------------------------------
-
-// FIXME-UTF8: using std::string as wxString base class is currently broken,
-// so we use the standard wxString with std::string conversion
-// enabled, this is API-compatible.
-#define wxUSE_STL_BASED_WXSTRING 0
-#if wxUSE_STL
- #undef wxUSE_STD_STRING
- #define wxUSE_STD_STRING 1
-#endif
-//#define wxUSE_STL_BASED_WXSTRING wxUSE_STL
-
-// in both cases we need to define wxStdString
-#if wxUSE_STL_BASED_WXSTRING || wxUSE_STD_STRING
-
-#include "wx/beforestd.h"
-#include <string>
-#include "wx/afterstd.h"
-
-#if wxUSE_UNICODE
- #ifdef HAVE_STD_WSTRING
- typedef std::wstring wxStdString;
- #else
- typedef std::basic_string<wxChar> wxStdString;
- #endif
-#else
- typedef std::string wxStdString;
-#endif
-
-#endif // need <string>
-
-#if wxUSE_STL_BASED_WXSTRING
-
- // we don't need an extra ctor from std::string when copy ctor already does
- // the work
- #undef wxUSE_STD_STRING
- #define wxUSE_STD_STRING 0
-
- #if (defined(__GNUG__) && (__GNUG__ < 3)) || \
- (defined(_MSC_VER) && (_MSC_VER <= 1200))
- #define wxSTRING_BASE_HASNT_CLEAR
- #endif
-
- typedef wxStdString wxStringBase;
-#else // if !wxUSE_STL_BASED_WXSTRING
-
-#if !defined(HAVE_STD_STRING_COMPARE) && \
- (!defined(__WX_SETUP_H__) || wxUSE_STL_BASED_WXSTRING == 0)
- #define HAVE_STD_STRING_COMPARE
-#endif
-
-// ---------------------------------------------------------------------------
-// string data prepended with some housekeeping info (used by wxString class),
-// is never used directly (but had to be put here to allow inlining)
-// ---------------------------------------------------------------------------
-
-struct WXDLLIMPEXP_BASE wxStringData
-{
- int nRefs; // reference count
- size_t nDataLength, // actual string length
- nAllocLength; // allocated memory size
-
- // mimics declaration 'wxChar data[nAllocLength]'
- wxChar* data() const { return (wxChar*)(this + 1); }
-
- // empty string has a special ref count so it's never deleted
- bool IsEmpty() const { return (nRefs == -1); }
- bool IsShared() const { return (nRefs > 1); }
-
- // lock/unlock
- void Lock() { if ( !IsEmpty() ) nRefs++; }
-
- // VC++ will refuse to inline Unlock but profiling shows that it is wrong
-#if defined(__VISUALC__) && (__VISUALC__ >= 1200)
- __forceinline
-#endif
- // VC++ free must take place in same DLL as allocation when using non dll
- // run-time library (e.g. Multithreaded instead of Multithreaded DLL)
-#if defined(__VISUALC__) && defined(_MT) && !defined(_DLL)
- void Unlock() { if ( !IsEmpty() && --nRefs == 0) Free(); }
- // we must not inline deallocation since allocation is not inlined
- void Free();
-#else
- void Unlock() { if ( !IsEmpty() && --nRefs == 0) free(this); }
-#endif
-
- // if we had taken control over string memory (GetWriteBuf), it's
- // intentionally put in invalid state
- void Validate(bool b) { nRefs = (b ? 1 : 0); }
- bool IsValid() const { return (nRefs != 0); }
-};
-
-class WXDLLIMPEXP_BASE wxStringBase
-{
-public :
- // an 'invalid' value for string index, moved to this place due to a CW bug
- static const size_t npos;
-protected:
- // points to data preceded by wxStringData structure with ref count info
- wxChar *m_pchData;
-
- // accessor to string data
- wxStringData* GetStringData() const { return (wxStringData*)m_pchData - 1; }
-
- // string (re)initialization functions
- // initializes the string to the empty value (must be called only from
- // ctors, use Reinit() otherwise)
- void Init() { m_pchData = (wxChar *)wxEmptyString; }
- // initializes the string with (a part of) C-string
- void InitWith(const wxChar *psz, size_t nPos = 0, size_t nLen = npos);
- // as Init, but also frees old data
- void Reinit() { GetStringData()->Unlock(); Init(); }
-
- // memory allocation
- // allocates memory for string of length nLen
- bool AllocBuffer(size_t nLen);
- // copies data to another string
- bool AllocCopy(wxString&, int, int) const;
- // effectively copies data to string
- bool AssignCopy(size_t, const wxChar *);
-
- // append a (sub)string
- bool ConcatSelf(size_t nLen, const wxChar *src, size_t nMaxLen);
- bool ConcatSelf(size_t nLen, const wxChar *src)
- { return ConcatSelf(nLen, src, nLen); }
-
- // functions called before writing to the string: they copy it if there
- // are other references to our data (should be the only owner when writing)
- bool CopyBeforeWrite();
- bool AllocBeforeWrite(size_t);
-
- // compatibility with wxString
- bool Alloc(size_t nLen);
-public:
- // standard types
- typedef wxUniChar value_type;
- typedef wxUniChar char_type;
- typedef wxUniCharRef reference;
- typedef wxChar* pointer;
- typedef const wxChar* const_pointer;
-
- typedef size_t size_type;
- typedef wxUniChar const_reference;
-
- #define WX_STR_ITERATOR_IMPL(iterator_name, pointer_type, \
- reference_type, reference_ctor) \
- public: \
- typedef wxUniChar value_type; \
- typedef reference_type reference; \
- typedef pointer_type pointer; \
- \
- iterator_name(const iterator_name& i) : m_cur(i.m_cur) {} \
- \
- reference operator*() const { return reference_ctor; } \
- \
- iterator_name& operator++() \
- { ++m_cur; return *this; } \
- iterator_name operator++(int) \
- { iterator_name tmp = *this; ++m_cur; return tmp; } \
- iterator_name& operator--() \
- { --m_cur; return *this; } \
- iterator_name operator--(int) \
- { iterator_name tmp = *this; --m_cur; return tmp; } \
- \
- iterator_name operator+(int n) const \
- { return iterator_name(m_cur + n); } \
- iterator_name operator+(size_t n) const \
- { return iterator_name(m_cur + n); } \
- iterator_name operator-(int n) const \
- { return iterator_name(m_cur - n); } \
- iterator_name operator-(size_t n) const \
- { return iterator_name(m_cur - n); } \
- iterator_name operator+=(int n) \
- { m_cur += n; return *this; } \
- iterator_name operator+=(size_t n) \
- { m_cur += n; return *this; } \
- iterator_name operator-=(int n) \
- { m_cur -= n; return *this; } \
- iterator_name operator-=(size_t n) \
- { m_cur -= n; return *this; } \
- \
- unsigned operator-(const iterator_name& i) const \
- { return m_cur - i.m_cur; } \
- \
- bool operator==(const iterator_name&i) const \
- { return m_cur == i.m_cur; } \
- bool operator!=(const iterator_name& i) const \
- { return m_cur != i.m_cur; } \
- \
- bool operator<(const iterator_name& i) const \
- { return m_cur < i.m_cur; } \
- bool operator>(const iterator_name& i) const \
- { return m_cur > i.m_cur; } \
- bool operator<=(const iterator_name& i) const \
- { return m_cur <= i.m_cur; } \
- bool operator>=(const iterator_name& i) const \
- { return m_cur >= i.m_cur; } \
- \
- protected: \
- /* for internal wxString use only: */ \
- iterator_name(pointer ptr) : m_cur(ptr) {} \
- operator pointer() const { return m_cur; } \
- \
- friend class WXDLLIMPEXP_BASE wxString; \
- friend class WXDLLIMPEXP_BASE wxStringBase; \
- friend class WXDLLIMPEXP_BASE wxCStrData; \
- \
- protected: \
- pointer m_cur;
-
- class const_iterator;
-
- class iterator
- {
- WX_STR_ITERATOR_IMPL(iterator, wxChar*, wxUniCharRef,
- wxUniCharRef::CreateForString(m_cur))
-
- friend class const_iterator;
- };
-
- class const_iterator
- {
- // NB: reference_type is intentionally value, not reference, the character
- // may be encoded differently in wxString data:
- WX_STR_ITERATOR_IMPL(const_iterator, const wxChar*, wxUniChar,
- wxUniChar(*m_cur))
-
- public:
- const_iterator(const iterator& i) : m_cur(i.m_cur) {}
- };
-
- #undef WX_STR_ITERATOR
-
- template <typename T>
- class reverse_iterator_impl
- {
- public:
- typedef T iterator_type;
- typedef typename T::value_type value_type;
- typedef typename T::reference reference;
- typedef typename T::pointer *pointer;
-
- reverse_iterator_impl(iterator_type i) : m_cur(i) {}
- reverse_iterator_impl(const reverse_iterator_impl& ri)
- : m_cur(ri.m_cur) {}
-
- iterator_type base() const { return m_cur; }
-
- reference operator*() const { return *(m_cur-1); }
-
- reverse_iterator_impl& operator++()
- { --m_cur; return *this; }
- reverse_iterator_impl operator++(int)
- { reverse_iterator_impl tmp = *this; --m_cur; return tmp; }
- reverse_iterator_impl& operator--()
- { ++m_cur; return *this; }
- reverse_iterator_impl operator--(int)
- { reverse_iterator_impl tmp = *this; ++m_cur; return tmp; }
-
- bool operator==(const reverse_iterator_impl& ri) const
- { return m_cur == ri.m_cur; }
- bool operator!=(const reverse_iterator_impl& ri) const
- { return !(*this == ri); }
-
- private:
- iterator_type m_cur;
- };
-
- typedef reverse_iterator_impl<iterator> reverse_iterator;
- typedef reverse_iterator_impl<const_iterator> const_reverse_iterator;
-
-
- // constructors and destructor
- // ctor for an empty string
- wxStringBase() { Init(); }
- // copy ctor
- wxStringBase(const wxStringBase& stringSrc)
- {
- wxASSERT_MSG( stringSrc.GetStringData()->IsValid(),
- _T("did you forget to call UngetWriteBuf()?") );
-
- if ( stringSrc.empty() ) {
- // nothing to do for an empty string
- Init();
- }
- else {
- m_pchData = stringSrc.m_pchData; // share same data
- GetStringData()->Lock(); // => one more copy
- }
- }
- // string containing nRepeat copies of ch
- wxStringBase(size_type nRepeat, wxUniChar ch);
- // ctor takes first nLength characters from C string
- // (default value of npos means take all the string)
- wxStringBase(const wxChar *psz)
- { InitWith(psz, 0, npos); }
- wxStringBase(const wxChar *psz, size_t nLength)
- { InitWith(psz, 0, nLength); }
- wxStringBase(const wxChar *psz,
- const wxMBConv& WXUNUSED(conv),
- size_t nLength = npos)
- { InitWith(psz, 0, nLength); }
- // take nLen chars starting at nPos
- wxStringBase(const wxStringBase& str, size_t nPos, size_t nLen)
- {
- wxASSERT_MSG( str.GetStringData()->IsValid(),
- _T("did you forget to call UngetWriteBuf()?") );
- Init();
- size_t strLen = str.length() - nPos; nLen = strLen < nLen ? strLen : nLen;
- InitWith(str.c_str(), nPos, nLen);
- }
- // take all characters from pStart to pEnd
- wxStringBase(const void *pStart, const void *pEnd);
-
- // dtor is not virtual, this class must not be inherited from!
- ~wxStringBase()
- {
-#if defined(__VISUALC__) && (__VISUALC__ >= 1200)
- //RN - according to the above VC++ does indeed inline this,
- //even though it spits out two warnings
- #pragma warning (disable:4714)
-#endif
-
- GetStringData()->Unlock();
- }
-
-#if defined(__VISUALC__) && (__VISUALC__ >= 1200)
- //re-enable inlining warning
- #pragma warning (default:4714)
-#endif
- // overloaded assignment
- // from another wxString
- wxStringBase& operator=(const wxStringBase& stringSrc);
- // from a character
- wxStringBase& operator=(wxUniChar ch);
- // from a C string
- wxStringBase& operator=(const wxChar *psz);
-
- // return the length of the string
- size_type length() const { return GetStringData()->nDataLength; }
- // return the length of the string
- size_type size() const { return length(); }
- // return the maximum size of the string
- size_type max_size() const { return npos; }
- // resize the string, filling the space with c if c != 0
- void resize(size_t nSize, wxUniChar ch = wxT('\0'));
- // delete the contents of the string
- void clear() { erase(0, npos); }
- // returns true if the string is empty
- bool empty() const { return length() == 0; }
- // inform string about planned change in size
- void reserve(size_t sz) { Alloc(sz); }
- size_type capacity() const { return GetStringData()->nAllocLength; }
-
- // lib.string.access
- // return the character at position n
- value_type at(size_type n) const
- { wxASSERT_VALID_INDEX( n ); return m_pchData[n]; }
- // returns the writable character at position n
- reference at(size_type n)
- {
- wxASSERT_VALID_INDEX( n );
- CopyBeforeWrite();
- return wxUniCharRef::CreateForString(&m_pchData[n]);
- }
-
- // lib.string.modifiers
- // append elements str[pos], ..., str[pos+n]
- wxStringBase& append(const wxStringBase& str, size_t pos, size_t n)
- {
- wxASSERT(pos <= str.length());
- ConcatSelf(n, str.c_str() + pos, str.length() - pos);
- return *this;
- }
- // append a string
- wxStringBase& append(const wxStringBase& str)
- { ConcatSelf(str.length(), str.c_str()); return *this; }
- // append first n (or all if n == npos) characters of sz
- wxStringBase& append(const wxChar *sz)
- { ConcatSelf(wxStrlen(sz), sz); return *this; }
- wxStringBase& append(const wxChar *sz, size_t n)
- { ConcatSelf(n, sz); return *this; }
- // append n copies of ch
- wxStringBase& append(size_t n, wxUniChar ch);
- // append from first to last
- wxStringBase& append(const_iterator first, const_iterator last)
- { ConcatSelf(last - first, first); return *this; }
-
- // same as `this_string = str'
- wxStringBase& assign(const wxStringBase& str)
- { return *this = str; }
- // same as ` = str[pos..pos + n]
- wxStringBase& assign(const wxStringBase& str, size_t pos, size_t n)
- { clear(); return append(str, pos, n); }
- // same as `= first n (or all if n == npos) characters of sz'
- wxStringBase& assign(const wxChar *sz)
- { clear(); return append(sz, wxStrlen(sz)); }
- wxStringBase& assign(const wxChar *sz, size_t n)
- { clear(); return append(sz, n); }
- // same as `= n copies of ch'
- wxStringBase& assign(size_t n, wxUniChar ch)
- { clear(); return append(n, ch); }
- // assign from first to last
- wxStringBase& assign(const_iterator first, const_iterator last)
- { clear(); return append(first, last); }
-
- // first valid index position
- const_iterator begin() const { return m_pchData; }
- iterator begin();
- // position one after the last valid one
- const_iterator end() const { return m_pchData + length(); }
- iterator end();
-
- // first element of the reversed string
- const_reverse_iterator rbegin() const { return const_reverse_iterator(end()); }
- reverse_iterator rbegin() { return reverse_iterator(end()); }
- // one beyond the end of the reversed string
- const_reverse_iterator rend() const { return const_reverse_iterator(begin()); }
- reverse_iterator rend() { return reverse_iterator(begin()); }
-
- // insert another string
- wxStringBase& insert(size_t nPos, const wxStringBase& str)
- {
- wxASSERT( str.GetStringData()->IsValid() );
- return insert(nPos, str.c_str(), str.length());
- }
- // insert n chars of str starting at nStart (in str)
- wxStringBase& insert(size_t nPos, const wxStringBase& str, size_t nStart, size_t n)
- {
- wxASSERT( str.GetStringData()->IsValid() );
- wxASSERT( nStart < str.length() );
- size_t strLen = str.length() - nStart;
- n = strLen < n ? strLen : n;
- return insert(nPos, str.c_str() + nStart, n);
- }
- // insert first n (or all if n == npos) characters of sz
- wxStringBase& insert(size_t nPos, const wxChar *sz, size_t n = npos);
- // insert n copies of ch
- wxStringBase& insert(size_t nPos, size_t n, wxUniChar ch)
- { return insert(nPos, wxStringBase(n, ch)); }
- iterator insert(iterator it, wxUniChar ch)
- { size_t idx = it - begin(); insert(idx, 1, ch); return begin() + idx; }
- void insert(iterator it, const_iterator first, const_iterator last)
- { insert(it - begin(), first, last - first); }
- void insert(iterator it, size_type n, wxUniChar ch)
- { insert(it - begin(), n, ch); }
-
- // delete characters from nStart to nStart + nLen
- wxStringBase& erase(size_type pos = 0, size_type n = npos);
- iterator erase(iterator first, iterator last)
- {
- size_t idx = first - begin();
- erase(idx, last - first);
- return begin() + idx;
- }
- iterator erase(iterator first);
-
- // explicit conversion to C string (use this with printf()!)
- const wxChar* c_str() const { return m_pchData; }
- const wxChar* data() const { return m_pchData; }
-
- // replaces the substring of length nLen starting at nStart
- wxStringBase& replace(size_t nStart, size_t nLen, const wxChar* sz);
- // replaces the substring of length nLen starting at nStart
- wxStringBase& replace(size_t nStart, size_t nLen, const wxStringBase& str)
- { return replace(nStart, nLen, str.c_str()); }
- // replaces the substring with nCount copies of ch
- wxStringBase& replace(size_t nStart, size_t nLen, size_t nCount, wxUniChar ch);
- // replaces a substring with another substring
- wxStringBase& replace(size_t nStart, size_t nLen,
- const wxStringBase& str, size_t nStart2, size_t nLen2);
- // replaces the substring with first nCount chars of sz
- wxStringBase& replace(size_t nStart, size_t nLen,
- const wxChar* sz, size_t nCount);
- wxStringBase& replace(iterator first, iterator last, const_pointer s)
- { return replace(first - begin(), last - first, s); }
- wxStringBase& replace(iterator first, iterator last, const_pointer s,
- size_type n)
- { return replace(first - begin(), last - first, s, n); }
- wxStringBase& replace(iterator first, iterator last, const wxStringBase& s)
- { return replace(first - begin(), last - first, s); }
- wxStringBase& replace(iterator first, iterator last, size_type n, wxUniChar c)
- { return replace(first - begin(), last - first, n, c); }
- wxStringBase& replace(iterator first, iterator last,
- const_iterator first1, const_iterator last1)
- { return replace(first - begin(), last - first, first1, last1 - first1); }
-
- // swap two strings
- void swap(wxStringBase& str);
-
- // All find() functions take the nStart argument which specifies the
- // position to start the search on, the default value is 0. All functions
- // return npos if there were no match.
-
- // find a substring
- size_t find(const wxStringBase& str, size_t nStart = 0) const;
-
- // find first n characters of sz
- size_t find(const wxChar* sz, size_t nStart = 0, size_t n = npos) const;
-
- // find the first occurence of character ch after nStart
- size_t find(wxUniChar ch, size_t nStart = 0) const;
-
- // rfind() family is exactly like find() but works right to left
-
- // as find, but from the end
- size_t rfind(const wxStringBase& str, size_t nStart = npos) const;
-
- // as find, but from the end
- size_t rfind(const wxChar* sz, size_t nStart = npos,
- size_t n = npos) const;
- // as find, but from the end
- size_t rfind(wxUniChar ch, size_t nStart = npos) const;
-
- // find first/last occurence of any character in the set
-
- // as strpbrk() but starts at nStart, returns npos if not found
- size_t find_first_of(const wxStringBase& str, size_t nStart = 0) const
- { return find_first_of(str.c_str(), nStart); }
- // same as above
- size_t find_first_of(const wxChar* sz, size_t nStart = 0) const;
- size_t find_first_of(const wxChar* sz, size_t nStart, size_t n) const;
- // same as find(char, size_t)
- size_t find_first_of(wxUniChar c, size_t nStart = 0) const
- { return find(c, nStart); }
- // find the last (starting from nStart) char from str in this string
- size_t find_last_of (const wxStringBase& str, size_t nStart = npos) const
- { return find_last_of(str.c_str(), nStart); }
- // same as above
- size_t find_last_of (const wxChar* sz, size_t nStart = npos) const;
- size_t find_last_of(const wxChar* sz, size_t nStart, size_t n) const;
- // same as above
- size_t find_last_of(wxUniChar c, size_t nStart = npos) const
- { return rfind(c, nStart); }
-
- // find first/last occurence of any character not in the set
-
- // as strspn() (starting from nStart), returns npos on failure
- size_t find_first_not_of(const wxStringBase& str, size_t nStart = 0) const
- { return find_first_not_of(str.c_str(), nStart); }
- // same as above
- size_t find_first_not_of(const wxChar* sz, size_t nStart = 0) const;
- size_t find_first_not_of(const wxChar* sz, size_t nStart, size_t n) const;
- // same as above
- size_t find_first_not_of(wxUniChar ch, size_t nStart = 0) const;
- // as strcspn()
- size_t find_last_not_of(const wxStringBase& str, size_t nStart = npos) const
- { return find_last_not_of(str.c_str(), nStart); }
- // same as above
- size_t find_last_not_of(const wxChar* sz, size_t nStart = npos) const;
- size_t find_last_not_of(const wxChar* sz, size_t nStart, size_t n) const;
- // same as above
- size_t find_last_not_of(wxUniChar ch, size_t nStart = npos) const;
-
- // All compare functions return -1, 0 or 1 if the [sub]string is less,
- // equal or greater than the compare() argument.
-
- // comparison with another string
- int compare(const wxStringBase& str) const;
- // comparison with a substring
- int compare(size_t nStart, size_t nLen, const wxStringBase& str) const;
- // comparison of 2 substrings
- int compare(size_t nStart, size_t nLen,
- const wxStringBase& str, size_t nStart2, size_t nLen2) const;
- // comparison with a c string
- int compare(const wxChar* sz) const;
- // substring comparison with first nCount characters of sz
- int compare(size_t nStart, size_t nLen,
- const wxChar* sz, size_t nCount = npos) const;
-
- size_type copy(wxChar* s, size_type n, size_type pos = 0);
-
- // substring extraction
- wxStringBase substr(size_t nStart = 0, size_t nLen = npos) const;
-
- // string += string
- wxStringBase& operator+=(const wxStringBase& s) { return append(s); }
- // string += C string
- wxStringBase& operator+=(const wxChar *psz) { return append(psz); }
- // string += char
- wxStringBase& operator+=(wxUniChar ch) { return append(1, ch); }
- wxStringBase& operator+=(wxUniCharRef ch) { return append(1, ch); }
- wxStringBase& operator+=(char ch) { return append(1, ch); }
- wxStringBase& operator+=(wchar_t ch) { return append(1, ch); }
-};
-
-#endif // !wxUSE_STL_BASED_WXSTRING
-
// ----------------------------------------------------------------------------
// wxCStrData
// ----------------------------------------------------------------------------
public:
// Ctor constructs the object from char literal; they are needed to make
// operator?: compile and they intentionally take char*, not const char*
- wxCStrData(char *buf);
- wxCStrData(wchar_t *buf);
+ inline wxCStrData(char *buf);
+ inline wxCStrData(wchar_t *buf);
+ inline wxCStrData(const wxCStrData& data);
- ~wxCStrData();
+ inline ~wxCStrData();
- // FIXME: we'll need convertors for both char* and wchar_t* and NONE
- // for wxChar*, but that's after completing the transition to
- // "smart" wxUniChar class. For now, just have conversion to
- // char* in ANSI build and wchar_t in Unicode build.
-#if wxUSE_UNICODE
+ // methods defined inline below must be declared inline or mingw32 3.4.5
+ // warns about "<symbol> defined locally after being referenced with
+ // dllimport linkage"
+#if wxUSE_UNICODE_WCHAR
+ inline
+#endif
const wchar_t* AsWChar() const;
operator const wchar_t*() const { return AsWChar(); }
-#else
+
+#if !wxUSE_UNICODE || wxUSE_UTF8_LOCALE_ONLY
+ inline
+#endif
const char* AsChar() const;
+ const unsigned char* AsUnsignedChar() const
+ { return (const unsigned char *) AsChar(); }
operator const char*() const { return AsChar(); }
-#endif
+ operator const unsigned char*() const { return AsUnsignedChar(); }
+
+ operator const void*() const { return AsChar(); }
- wxString AsString() const;
- operator wxString() const;
+ inline const wxCharBuffer AsCharBuf() const;
+ inline const wxWCharBuffer AsWCharBuf() const;
+
+ inline wxString AsString() const;
+
+ // returns the value as C string in internal representation (equivalent
+ // to AsString().wx_str(), but more efficient)
+ const wxStringCharType *AsInternal() const;
// allow expressions like "c_str()[0]":
+ inline wxUniChar operator[](size_t n) const;
wxUniChar operator[](int n) const { return operator[](size_t(n)); }
- wxUniChar operator[](size_t n) const;
wxUniChar operator[](long n) const { return operator[](size_t(n)); }
#ifndef wxSIZE_T_IS_UINT
wxUniChar operator[](unsigned int n) const { return operator[](size_t(n)); }
wxCStrData operator+(size_t n) const
{ return wxCStrData(m_str, m_offset + n, m_owned); }
- // this operator is need to make expressions like "*c_str()" or
+ // and these for "str.c_str() + n - 2":
+ wxCStrData operator-(int n) const
+ {
+ wxASSERT_MSG( n <= (int)m_offset,
+ _T("attempt to construct address before the beginning of the string") );
+ return wxCStrData(m_str, m_offset - n, m_owned);
+ }
+ wxCStrData operator-(long n) const
+ {
+ wxASSERT_MSG( n <= (int)m_offset,
+ _T("attempt to construct address before the beginning of the string") );
+ return wxCStrData(m_str, m_offset - n, m_owned);
+ }
+ wxCStrData operator-(size_t n) const
+ {
+ wxASSERT_MSG( n <= m_offset,
+ _T("attempt to construct address before the beginning of the string") );
+ return wxCStrData(m_str, m_offset - n, m_owned);
+ }
+
+ // this operator is needed to make expressions like "*c_str()" or
// "*(c_str() + 2)" work
- wxUniChar operator*() const;
+ inline wxUniChar operator*() const;
private:
const wxString *m_str;
size_t m_offset;
bool m_owned;
- friend class WXDLLIMPEXP_BASE wxString;
+ friend class WXDLLIMPEXP_FWD_BASE wxString;
};
// ----------------------------------------------------------------------------
protected:
wxStringPrintfMixinBase() {}
- int DoPrintf(const wxChar *format, ...) ATTRIBUTE_PRINTF_2;
- static wxString DoFormat(const wxChar *format, ...) ATTRIBUTE_PRINTF_1;
+#if !wxUSE_UTF8_LOCALE_ONLY
+ int DoPrintfWchar(const wxChar *format, ...);
+ static wxString DoFormatWchar(const wxChar *format, ...);
+#endif
+#if wxUSE_UNICODE_UTF8
+ int DoPrintfUtf8(const char *format, ...);
+ static wxString DoFormatUtf8(const char *format, ...);
+#endif
};
// this class contains template wrappers for wxString's vararg methods, it's
// these are duplicated wxString methods, they're also declared below
// if !wxNEEDS_WXSTRING_PRINTF_MIXIN:
- // int Printf(const wxChar *pszFormat, ...);
- WX_DEFINE_VARARG_FUNC(int, Printf, DoPrintf)
- // static wxString Format(const wxChar *pszFormat, ...) ATTRIBUTE_PRINTF_1;
- WX_DEFINE_VARARG_FUNC(static typename StringReturnType<T1>::type,
- Format, DoFormat)
- // int sprintf(const wxChar *pszFormat, ...) ATTRIBUTE_PRINTF_2;
- WX_DEFINE_VARARG_FUNC(int, sprintf, DoPrintf)
+ // static wxString Format(const wString& format, ...) ATTRIBUTE_PRINTF_1;
+ WX_DEFINE_VARARG_FUNC_SANS_N0(static typename StringReturnType<T1>::type,
+ Format, 1, (const wxFormatString&),
+ DoFormatWchar, DoFormatUtf8)
+ // We have to implement the version without template arguments manually
+ // because of the StringReturnType<> hack, although WX_DEFINE_VARARG_FUNC
+ // normally does it itself. It has to be a template so that we can use
+ // the hack, even though there's no real template parameter. We can't move
+ // it to wxStrig, because it would shadow these versions of Format() then.
+ template<typename T>
+ inline static typename StringReturnType<T>::type
+ Format(const T& fmt)
+ {
+ // NB: this doesn't compile if T is not (some form of) a string;
+ // this makes Format's prototype equivalent to
+ // Format(const wxFormatString& fmt)
+ return DoFormatWchar(wxFormatString(fmt));
+ }
+
+ // int Printf(const wxString& format, ...);
+ WX_DEFINE_VARARG_FUNC(int, Printf, 1, (const wxFormatString&),
+ DoPrintfWchar, DoPrintfUtf8)
+ // int sprintf(const wxString& format, ...) ATTRIBUTE_PRINTF_2;
+ WX_DEFINE_VARARG_FUNC(int, sprintf, 1, (const wxFormatString&),
+ DoPrintfWchar, DoPrintfUtf8)
protected:
wxStringPrintfMixin() : wxStringPrintfMixinBase() {}
#pragma warning (disable:4275)
#endif
-class WXDLLIMPEXP_BASE wxString : public wxStringBase
+#if wxUSE_UNICODE_UTF8
+// see the comment near wxString::iterator for why we need this
+class WXDLLIMPEXP_BASE wxStringIteratorNode
+{
+public:
+ wxStringIteratorNode()
+ : m_str(NULL), m_citer(NULL), m_iter(NULL), m_prev(NULL), m_next(NULL) {}
+ wxStringIteratorNode(const wxString *str,
+ wxStringImpl::const_iterator *citer)
+ { DoSet(str, citer, NULL); }
+ wxStringIteratorNode(const wxString *str, wxStringImpl::iterator *iter)
+ { DoSet(str, NULL, iter); }
+ ~wxStringIteratorNode()
+ { clear(); }
+
+ inline void set(const wxString *str, wxStringImpl::const_iterator *citer)
+ { clear(); DoSet(str, citer, NULL); }
+ inline void set(const wxString *str, wxStringImpl::iterator *iter)
+ { clear(); DoSet(str, NULL, iter); }
+
+ const wxString *m_str;
+ wxStringImpl::const_iterator *m_citer;
+ wxStringImpl::iterator *m_iter;
+ wxStringIteratorNode *m_prev, *m_next;
+
+private:
+ inline void clear();
+ inline void DoSet(const wxString *str,
+ wxStringImpl::const_iterator *citer,
+ wxStringImpl::iterator *iter);
+
+ // the node belongs to a particular iterator instance, it's not copied
+ // when a copy of the iterator is made
+ DECLARE_NO_COPY_CLASS(wxStringIteratorNode)
+};
+#endif // wxUSE_UNICODE_UTF8
+
+class WXDLLIMPEXP_BASE wxString
#ifdef wxNEEDS_WXSTRING_PRINTF_MIXIN
- ,public wxStringPrintfMixin
+ : public wxStringPrintfMixin
#endif
{
// NB: special care was taken in arranging the member functions in such order
// that all inline functions can be effectively inlined, verify that all
// performance critical functions are still inlined if you change order!
+public:
+ // an 'invalid' value for string index, moved to this place due to a CW bug
+ static const size_t npos;
+
private:
// if we hadn't made these operators private, it would be possible to
// compile "wxString s; s = 17;" without any warnings as 17 is implicitly
// try `s << i' or `s.Printf("%d", i)' instead
wxString(int);
+
+ // buffer for holding temporary substring when using any of the methods
+ // that take (char*,size_t) or (wchar_t*,size_t) arguments:
+ template<typename T>
+ struct SubstrBufFromType
+ {
+ T data;
+ size_t len;
+
+ SubstrBufFromType(const T& data_, size_t len_)
+ : data(data_), len(len_) {}
+ };
+
+#if wxUSE_UNICODE_UTF8
+ // even char* -> char* needs conversion, from locale charset to UTF-8
+ typedef SubstrBufFromType<wxCharBuffer> SubstrBufFromWC;
+ typedef SubstrBufFromType<wxCharBuffer> SubstrBufFromMB;
+#elif wxUSE_UNICODE_WCHAR
+ typedef SubstrBufFromType<const wchar_t*> SubstrBufFromWC;
+ typedef SubstrBufFromType<wxWCharBuffer> SubstrBufFromMB;
+#else
+ typedef SubstrBufFromType<const char*> SubstrBufFromMB;
+ typedef SubstrBufFromType<wxCharBuffer> SubstrBufFromWC;
+#endif
+
+
+ // Functions implementing primitive operations on string data; wxString
+ // methods and iterators are implemented in terms of it. The differences
+ // between UTF-8 and wchar_t* representations of the string are mostly
+ // contained here.
+
+#if wxUSE_UNICODE_UTF8
+ static SubstrBufFromMB ConvertStr(const char *psz, size_t nLength,
+ const wxMBConv& conv);
+ static SubstrBufFromWC ConvertStr(const wchar_t *pwz, size_t nLength,
+ const wxMBConv& conv);
+#elif wxUSE_UNICODE_WCHAR
+ static SubstrBufFromMB ConvertStr(const char *psz, size_t nLength,
+ const wxMBConv& conv);
+#else
+ static SubstrBufFromWC ConvertStr(const wchar_t *pwz, size_t nLength,
+ const wxMBConv& conv);
+#endif
+
+#if !wxUSE_UNICODE_UTF8 // wxUSE_UNICODE_WCHAR or !wxUSE_UNICODE
+ // returns C string encoded as the implementation expects:
+ #if wxUSE_UNICODE
+ static const wchar_t* ImplStr(const wchar_t* str)
+ { return str ? str : wxT(""); }
+ static const SubstrBufFromWC ImplStr(const wchar_t* str, size_t n)
+ { return SubstrBufFromWC(str, (str && n == npos) ? wxWcslen(str) : n); }
+ static wxWCharBuffer ImplStr(const char* str,
+ const wxMBConv& conv = wxConvLibc)
+ { return ConvertStr(str, npos, conv).data; }
+ static SubstrBufFromMB ImplStr(const char* str, size_t n,
+ const wxMBConv& conv = wxConvLibc)
+ { return ConvertStr(str, n, conv); }
+ #else
+ static const char* ImplStr(const char* str,
+ const wxMBConv& WXUNUSED(conv) = wxConvLibc)
+ { return str ? str : ""; }
+ static const SubstrBufFromMB ImplStr(const char* str, size_t n,
+ const wxMBConv& WXUNUSED(conv) = wxConvLibc)
+ { return SubstrBufFromMB(str, (str && n == npos) ? wxStrlen(str) : n); }
+ static wxCharBuffer ImplStr(const wchar_t* str)
+ { return ConvertStr(str, npos, wxConvLibc).data; }
+ static SubstrBufFromWC ImplStr(const wchar_t* str, size_t n)
+ { return ConvertStr(str, n, wxConvLibc); }
+ #endif
+
+ // translates position index in wxString to/from index in underlying
+ // wxStringImpl:
+ static size_t PosToImpl(size_t pos) { return pos; }
+ static void PosLenToImpl(size_t pos, size_t len,
+ size_t *implPos, size_t *implLen)
+ { *implPos = pos; *implLen = len; }
+ static size_t LenToImpl(size_t len) { return len; }
+ static size_t PosFromImpl(size_t pos) { return pos; }
+
+#else // wxUSE_UNICODE_UTF8
+
+ static wxCharBuffer ImplStr(const char* str,
+ const wxMBConv& conv = wxConvLibc)
+ { return ConvertStr(str, npos, conv).data; }
+ static SubstrBufFromMB ImplStr(const char* str, size_t n,
+ const wxMBConv& conv = wxConvLibc)
+ { return ConvertStr(str, n, conv); }
+
+ static wxCharBuffer ImplStr(const wchar_t* str)
+ { return ConvertStr(str, npos, wxMBConvUTF8()).data; }
+ static SubstrBufFromWC ImplStr(const wchar_t* str, size_t n)
+ { return ConvertStr(str, n, wxMBConvUTF8()); }
+
+ size_t PosToImpl(size_t pos) const
+ {
+ if ( pos == 0 || pos == npos )
+ return pos;
+ else
+ return (begin() + pos).impl() - m_impl.begin();
+ }
+
+ void PosLenToImpl(size_t pos, size_t len, size_t *implPos, size_t *implLen) const;
+
+ size_t LenToImpl(size_t len) const
+ {
+ size_t pos, len2;
+ PosLenToImpl(0, len, &pos, &len2);
+ return len2;
+ }
+
+ size_t PosFromImpl(size_t pos) const
+ {
+ if ( pos == 0 || pos == npos )
+ return pos;
+ else
+ return const_iterator(this, m_impl.begin() + pos) - begin();
+ }
+#endif // !wxUSE_UNICODE_UTF8/wxUSE_UNICODE_UTF8
+
+public:
+ // standard types
+ typedef wxUniChar value_type;
+ typedef wxUniChar char_type;
+ typedef wxUniCharRef reference;
+ typedef wxChar* pointer;
+ typedef const wxChar* const_pointer;
+
+ typedef size_t size_type;
+ typedef wxUniChar const_reference;
+
+#if wxUSE_STL
+ #if wxUSE_UNICODE_UTF8
+ // random access is not O(1), as required by Random Access Iterator
+ #define WX_STR_ITERATOR_TAG std::bidirectional_iterator_tag
+ #else
+ #define WX_STR_ITERATOR_TAG std::random_access_iterator_tag
+ #endif
+#else
+ #define WX_STR_ITERATOR_TAG void /* dummy type */
+#endif
+
+ #define WX_STR_ITERATOR_IMPL(iterator_name, pointer_type, reference_type) \
+ private: \
+ typedef wxStringImpl::iterator_name underlying_iterator; \
+ public: \
+ typedef WX_STR_ITERATOR_TAG iterator_category; \
+ typedef wxUniChar value_type; \
+ typedef int difference_type; \
+ typedef reference_type reference; \
+ typedef pointer_type pointer; \
+ \
+ reference operator[](size_t n) const { return *(*this + n); } \
+ \
+ iterator_name& operator++() \
+ { wxStringOperations::IncIter(m_cur); return *this; } \
+ iterator_name& operator--() \
+ { wxStringOperations::DecIter(m_cur); return *this; } \
+ iterator_name operator++(int) \
+ { \
+ iterator_name tmp = *this; \
+ wxStringOperations::IncIter(m_cur); \
+ return tmp; \
+ } \
+ iterator_name operator--(int) \
+ { \
+ iterator_name tmp = *this; \
+ wxStringOperations::DecIter(m_cur); \
+ return tmp; \
+ } \
+ \
+ iterator_name& operator+=(int n) \
+ { \
+ m_cur = wxStringOperations::AddToIter(m_cur, n); \
+ return *this; \
+ } \
+ iterator_name& operator+=(size_t n) \
+ { \
+ m_cur = wxStringOperations::AddToIter(m_cur, (int)n); \
+ return *this; \
+ } \
+ iterator_name& operator-=(int n) \
+ { \
+ m_cur = wxStringOperations::AddToIter(m_cur, -n); \
+ return *this; \
+ } \
+ iterator_name& operator-=(size_t n) \
+ { \
+ m_cur = wxStringOperations::AddToIter(m_cur, -(int)n); \
+ return *this; \
+ } \
+ \
+ difference_type operator-(const iterator_name& i) const \
+ { return wxStringOperations::DiffIters(m_cur, i.m_cur); } \
+ \
+ bool operator==(const iterator_name& i) const \
+ { return m_cur == i.m_cur; } \
+ bool operator!=(const iterator_name& i) const \
+ { return m_cur != i.m_cur; } \
+ \
+ bool operator<(const iterator_name& i) const \
+ { return m_cur < i.m_cur; } \
+ bool operator>(const iterator_name& i) const \
+ { return m_cur > i.m_cur; } \
+ bool operator<=(const iterator_name& i) const \
+ { return m_cur <= i.m_cur; } \
+ bool operator>=(const iterator_name& i) const \
+ { return m_cur >= i.m_cur; } \
+ \
+ private: \
+ /* for internal wxString use only: */ \
+ underlying_iterator impl() const { return m_cur; } \
+ \
+ friend class wxString; \
+ friend class wxCStrData; \
+ \
+ private: \
+ underlying_iterator m_cur
+
+ class WXDLLIMPEXP_FWD_BASE const_iterator;
+
+#if wxUSE_UNICODE_UTF8
+ // NB: In UTF-8 build, (non-const) iterator needs to keep reference
+ // to the underlying wxStringImpl, because UTF-8 is variable-length
+ // encoding and changing the value pointer to by an iterator (using
+ // its operator*) requires calling wxStringImpl::replace() if the old
+ // and new values differ in their encoding's length.
+ //
+ // Furthermore, the replace() call may invalid all iterators for the
+ // string, so we have to keep track of outstanding iterators and update
+ // them if replace() happens.
+ //
+ // This is implemented by maintaining linked list of iterators for every
+ // string and traversing it in wxUniCharRef::operator=(). Head of the
+ // list is stored in wxString. (FIXME-UTF8)
+
+ class WXDLLIMPEXP_BASE iterator
+ {
+ WX_STR_ITERATOR_IMPL(iterator, wxChar*, wxUniCharRef);
+
+ public:
+ iterator() {}
+ iterator(const iterator& i)
+ : m_cur(i.m_cur), m_node(i.str(), &m_cur) {}
+ iterator& operator=(const iterator& i)
+ { m_cur = i.m_cur; m_node.set(i.str(), &m_cur); return *this; }
+
+ reference operator*()
+ { return wxUniCharRef::CreateForString(m_node, m_cur); }
+
+ iterator operator+(int n) const
+ { return iterator(str(), wxStringOperations::AddToIter(m_cur, n)); }
+ iterator operator+(size_t n) const
+ { return iterator(str(), wxStringOperations::AddToIter(m_cur, (int)n)); }
+ iterator operator-(int n) const
+ { return iterator(str(), wxStringOperations::AddToIter(m_cur, -n)); }
+ iterator operator-(size_t n) const
+ { return iterator(str(), wxStringOperations::AddToIter(m_cur, -(int)n)); }
+
+ private:
+ iterator(wxString *str, underlying_iterator ptr)
+ : m_cur(ptr), m_node(str, &m_cur) {}
+
+ wxString* str() const { return wx_const_cast(wxString*, m_node.m_str); }
+
+ wxStringIteratorNode m_node;
+
+ friend class const_iterator;
+ };
+
+ class WXDLLIMPEXP_BASE const_iterator
+ {
+ // NB: reference_type is intentionally value, not reference, the character
+ // may be encoded differently in wxString data:
+ WX_STR_ITERATOR_IMPL(const_iterator, const wxChar*, wxUniChar);
+
+ public:
+ const_iterator() {}
+ const_iterator(const const_iterator& i)
+ : m_cur(i.m_cur), m_node(i.str(), &m_cur) {}
+ const_iterator(const iterator& i)
+ : m_cur(i.m_cur), m_node(i.str(), &m_cur) {}
+
+ const_iterator& operator=(const const_iterator& i)
+ { m_cur = i.m_cur; m_node.set(i.str(), &m_cur); return *this; }
+ const_iterator& operator=(const iterator& i)
+ { m_cur = i.m_cur; m_node.set(i.str(), &m_cur); return *this; }
+
+ reference operator*() const
+ { return wxStringOperations::DecodeChar(m_cur); }
+
+ const_iterator operator+(int n) const
+ { return const_iterator(str(), wxStringOperations::AddToIter(m_cur, n)); }
+ const_iterator operator+(size_t n) const
+ { return const_iterator(str(), wxStringOperations::AddToIter(m_cur, (int)n)); }
+ const_iterator operator-(int n) const
+ { return const_iterator(str(), wxStringOperations::AddToIter(m_cur, -n)); }
+ const_iterator operator-(size_t n) const
+ { return const_iterator(str(), wxStringOperations::AddToIter(m_cur, -(int)n)); }
+
+ private:
+ // for internal wxString use only:
+ const_iterator(const wxString *str, underlying_iterator ptr)
+ : m_cur(ptr), m_node(str, &m_cur) {}
+
+ const wxString* str() const { return m_node.m_str; }
+
+ wxStringIteratorNode m_node;
+ };
+
+ size_t IterToImplPos(wxString::iterator i) const
+ { return wxStringImpl::const_iterator(i.impl()) - m_impl.begin(); }
+
+#else // !wxUSE_UNICODE_UTF8
+
+ class WXDLLIMPEXP_BASE iterator
+ {
+ WX_STR_ITERATOR_IMPL(iterator, wxChar*, wxUniCharRef);
+
+ public:
+ iterator() {}
+ iterator(const iterator& i) : m_cur(i.m_cur) {}
+
+ reference operator*()
+ { return wxUniCharRef::CreateForString(m_cur); }
+
+ iterator operator+(int n) const
+ { return iterator(wxStringOperations::AddToIter(m_cur, n)); }
+ iterator operator+(size_t n) const
+ { return iterator(wxStringOperations::AddToIter(m_cur, (int)n)); }
+ iterator operator-(int n) const
+ { return iterator(wxStringOperations::AddToIter(m_cur, -n)); }
+ iterator operator-(size_t n) const
+ { return iterator(wxStringOperations::AddToIter(m_cur, -(int)n)); }
+
+ private:
+ // for internal wxString use only:
+ iterator(underlying_iterator ptr) : m_cur(ptr) {}
+ iterator(wxString *WXUNUSED(str), underlying_iterator ptr) : m_cur(ptr) {}
+
+ friend class const_iterator;
+ };
+
+ class WXDLLIMPEXP_BASE const_iterator
+ {
+ // NB: reference_type is intentionally value, not reference, the character
+ // may be encoded differently in wxString data:
+ WX_STR_ITERATOR_IMPL(const_iterator, const wxChar*, wxUniChar);
+
+ public:
+ const_iterator() {}
+ const_iterator(const const_iterator& i) : m_cur(i.m_cur) {}
+ const_iterator(const iterator& i) : m_cur(i.m_cur) {}
+
+ reference operator*() const
+ { return wxStringOperations::DecodeChar(m_cur); }
+
+ const_iterator operator+(int n) const
+ { return const_iterator(wxStringOperations::AddToIter(m_cur, n)); }
+ const_iterator operator+(size_t n) const
+ { return const_iterator(wxStringOperations::AddToIter(m_cur, (int)n)); }
+ const_iterator operator-(int n) const
+ { return const_iterator(wxStringOperations::AddToIter(m_cur, -n)); }
+ const_iterator operator-(size_t n) const
+ { return const_iterator(wxStringOperations::AddToIter(m_cur, -(int)n)); }
+
+ private:
+ // for internal wxString use only:
+ const_iterator(underlying_iterator ptr) : m_cur(ptr) {}
+ const_iterator(const wxString *WXUNUSED(str), underlying_iterator ptr)
+ : m_cur(ptr) {}
+ };
+#endif // wxUSE_UNICODE_UTF8/!wxUSE_UNICODE_UTF8
+
+ #undef WX_STR_ITERATOR_TAG
+ #undef WX_STR_ITERATOR_IMPL
+
+ friend class iterator;
+ friend class const_iterator;
+
+ template <typename T>
+ class reverse_iterator_impl
+ {
+ public:
+ typedef T iterator_type;
+
+ typedef typename T::iterator_category iterator_category;
+ typedef typename T::value_type value_type;
+ typedef typename T::difference_type difference_type;
+ typedef typename T::reference reference;
+ typedef typename T::pointer *pointer;
+
+ reverse_iterator_impl() {}
+ reverse_iterator_impl(iterator_type i) : m_cur(i) {}
+ reverse_iterator_impl(const reverse_iterator_impl& ri)
+ : m_cur(ri.m_cur) {}
+
+ iterator_type base() const { return m_cur; }
+
+ reference operator*() const { return *(m_cur-1); }
+ reference operator[](size_t n) const { return *(*this + n); }
+
+ reverse_iterator_impl& operator++()
+ { --m_cur; return *this; }
+ reverse_iterator_impl operator++(int)
+ { reverse_iterator_impl tmp = *this; --m_cur; return tmp; }
+ reverse_iterator_impl& operator--()
+ { ++m_cur; return *this; }
+ reverse_iterator_impl operator--(int)
+ { reverse_iterator_impl tmp = *this; ++m_cur; return tmp; }
+
+ // NB: explicit <T> in the functions below is to keep BCC 5.5 happy
+ reverse_iterator_impl operator+(int n) const
+ { return reverse_iterator_impl<T>(m_cur - n); }
+ reverse_iterator_impl operator+(size_t n) const
+ { return reverse_iterator_impl<T>(m_cur - n); }
+ reverse_iterator_impl operator-(int n) const
+ { return reverse_iterator_impl<T>(m_cur + n); }
+ reverse_iterator_impl operator-(size_t n) const
+ { return reverse_iterator_impl<T>(m_cur + n); }
+ reverse_iterator_impl operator+=(int n)
+ { m_cur -= n; return *this; }
+ reverse_iterator_impl operator+=(size_t n)
+ { m_cur -= n; return *this; }
+ reverse_iterator_impl operator-=(int n)
+ { m_cur += n; return *this; }
+ reverse_iterator_impl operator-=(size_t n)
+ { m_cur += n; return *this; }
+
+ unsigned operator-(const reverse_iterator_impl& i) const
+ { return i.m_cur - m_cur; }
+
+ bool operator==(const reverse_iterator_impl& ri) const
+ { return m_cur == ri.m_cur; }
+ bool operator!=(const reverse_iterator_impl& ri) const
+ { return !(*this == ri); }
+
+ bool operator<(const reverse_iterator_impl& i) const
+ { return m_cur > i.m_cur; }
+ bool operator>(const reverse_iterator_impl& i) const
+ { return m_cur < i.m_cur; }
+ bool operator<=(const reverse_iterator_impl& i) const
+ { return m_cur >= i.m_cur; }
+ bool operator>=(const reverse_iterator_impl& i) const
+ { return m_cur <= i.m_cur; }
+
+ private:
+ iterator_type m_cur;
+ };
+
+ typedef reverse_iterator_impl<iterator> reverse_iterator;
+ typedef reverse_iterator_impl<const_iterator> const_reverse_iterator;
+
+private:
+ // used to transform an expression built using c_str() (and hence of type
+ // wxCStrData) to an iterator into the string
+ static const_iterator CreateConstIterator(const wxCStrData& data)
+ {
+ return const_iterator(data.m_str,
+ (data.m_str->begin() + data.m_offset).impl());
+ }
+
+ // in UTF-8 STL build, creation from std::string requires conversion under
+ // non-UTF8 locales, so we can't have and use wxString(wxStringImpl) ctor;
+ // instead we define dummy type that lets us have wxString ctor for creation
+ // from wxStringImpl that couldn't be used by user code (in all other builds,
+ // "standard" ctors can be used):
+#if wxUSE_UNICODE_UTF8 && wxUSE_STL_BASED_WXSTRING
+ struct CtorFromStringImplTag {};
+
+ wxString(CtorFromStringImplTag* WXUNUSED(dummy), const wxStringImpl& src)
+ : m_impl(src) {}
+
+ static wxString FromImpl(const wxStringImpl& src)
+ { return wxString((CtorFromStringImplTag*)NULL, src); }
+#else
+ #if !wxUSE_STL_BASED_WXSTRING
+ wxString(const wxStringImpl& src) : m_impl(src) { }
+ // else: already defined as wxString(wxStdString) below
+ #endif
+ static wxString FromImpl(const wxStringImpl& src) { return wxString(src); }
+#endif
+
public:
// constructors and destructor
// ctor for an empty string
- wxString() : wxStringBase() { }
+ wxString() {}
+
// copy ctor
- wxString(const wxStringBase& stringSrc) : wxStringBase(stringSrc) { }
- wxString(const wxString& stringSrc) : wxStringBase(stringSrc) { }
+ wxString(const wxString& stringSrc) : m_impl(stringSrc.m_impl) { }
+
// string containing nRepeat copies of ch
wxString(wxUniChar ch, size_t nRepeat = 1)
- : wxStringBase(nRepeat, ch) { }
+ { assign(nRepeat, ch); }
wxString(size_t nRepeat, wxUniChar ch)
- : wxStringBase(nRepeat, ch) { }
+ { assign(nRepeat, ch); }
wxString(wxUniCharRef ch, size_t nRepeat = 1)
- : wxStringBase(nRepeat, ch) { }
+ { assign(nRepeat, ch); }
wxString(size_t nRepeat, wxUniCharRef ch)
- : wxStringBase(nRepeat, ch) { }
+ { assign(nRepeat, ch); }
wxString(char ch, size_t nRepeat = 1)
- : wxStringBase(nRepeat, ch) { }
+ { assign(nRepeat, ch); }
wxString(size_t nRepeat, char ch)
- : wxStringBase(nRepeat, ch) { }
+ { assign(nRepeat, ch); }
wxString(wchar_t ch, size_t nRepeat = 1)
- : wxStringBase(nRepeat, ch) { }
+ { assign(nRepeat, ch); }
wxString(size_t nRepeat, wchar_t ch)
- : wxStringBase(nRepeat, ch) { }
- // ctor takes first nLength characters from C string
- // (default value of npos means take all the string)
- wxString(const wxChar *psz)
- : wxStringBase(psz ? psz : wxT("")) { }
- wxString(const wxChar *psz, size_t nLength)
- : wxStringBase(psz, nLength) { }
- wxString(const wxChar *psz,
- const wxMBConv& WXUNUSED(conv),
- size_t nLength = npos)
- : wxStringBase(psz, nLength == npos ? wxStrlen(psz) : nLength) { }
+ { assign(nRepeat, ch); }
+
+ // ctors from char* strings:
+ wxString(const char *psz)
+ : m_impl(ImplStr(psz)) {}
+ wxString(const char *psz, const wxMBConv& conv)
+ : m_impl(ImplStr(psz, conv)) {}
+ wxString(const char *psz, size_t nLength)
+ { assign(psz, nLength); }
+ wxString(const char *psz, const wxMBConv& conv, size_t nLength)
+ {
+ SubstrBufFromMB str(ImplStr(psz, nLength, conv));
+ m_impl.assign(str.data, str.len);
+ }
+
+ // and unsigned char*:
+ wxString(const unsigned char *psz)
+ : m_impl(ImplStr((const char*)psz)) {}
+ wxString(const unsigned char *psz, const wxMBConv& conv)
+ : m_impl(ImplStr((const char*)psz, conv)) {}
+ wxString(const unsigned char *psz, size_t nLength)
+ { assign((const char*)psz, nLength); }
+ wxString(const unsigned char *psz, const wxMBConv& conv, size_t nLength)
+ {
+ SubstrBufFromMB str(ImplStr((const char*)psz, nLength, conv));
+ m_impl.assign(str.data, str.len);
+ }
+
+ // ctors from wchar_t* strings:
+ wxString(const wchar_t *pwz)
+ : m_impl(ImplStr(pwz)) {}
+ wxString(const wchar_t *pwz, const wxMBConv& WXUNUSED(conv))
+ : m_impl(ImplStr(pwz)) {}
+ wxString(const wchar_t *pwz, size_t nLength)
+ { assign(pwz, nLength); }
+ wxString(const wchar_t *pwz, const wxMBConv& WXUNUSED(conv), size_t nLength)
+ { assign(pwz, nLength); }
+
+ wxString(const wxCharBuffer& buf)
+ { assign(buf.data()); } // FIXME-UTF8: fix for embedded NUL and buffer length
+ wxString(const wxWCharBuffer& buf)
+ { assign(buf.data()); } // FIXME-UTF8: fix for embedded NUL and buffer length
+
+ wxString(const wxCStrData& cstr)
+ : m_impl(cstr.AsString().m_impl) { }
+
+ // as we provide both ctors with this signature for both char and unsigned
+ // char string, we need to provide one for wxCStrData to resolve ambiguity
+ wxString(const wxCStrData& cstr, size_t nLength)
+ : m_impl(cstr.AsString().Mid(0, nLength).m_impl) {}
+
+ // and because wxString is convertible to wxCStrData and const wxChar *
+ // we also need to provide this one
+ wxString(const wxString& str, size_t nLength)
+ { assign(str, nLength); }
// even if we're not built with wxUSE_STL == 1 it is very convenient to allow
- // implicit conversions from std::string to wxString as this allows to use
- // the same strings in non-GUI and GUI code, however we don't want to
- // unconditionally add this ctor as it would make wx lib dependent on
+ // implicit conversions from std::string to wxString and vice verse as this
+ // allows to use the same strings in non-GUI and GUI code, however we don't
+ // want to unconditionally add this ctor as it would make wx lib dependent on
// libstdc++ on some Linux versions which is bad, so instead we ask the
// client code to define this wxUSE_STD_STRING symbol if they need it
#if wxUSE_STD_STRING
- wxString(const wxStdString& s)
- : wxStringBase(s.c_str()) { }
+ #if wxUSE_UNICODE_WCHAR
+ wxString(const wxStdWideString& str) : m_impl(str) {}
+ #else // UTF-8 or ANSI
+ wxString(const wxStdWideString& str)
+ { assign(str.c_str(), str.length()); }
+ #endif
+
+ #if !wxUSE_UNICODE // ANSI build
+ // FIXME-UTF8: do this in UTF8 build #if wxUSE_UTF8_LOCALE_ONLY, too
+ wxString(const std::string& str) : m_impl(str) {}
+ #else // Unicode
+ wxString(const std::string& str)
+ { assign(str.c_str(), str.length()); }
+ #endif
#endif // wxUSE_STD_STRING
-#if wxUSE_UNICODE
- // from multibyte string
- wxString(const char *psz,
- const wxMBConv& conv = wxConvLibc,
- size_t nLength = npos);
- // from wxWCharBuffer (i.e. return from wxGetString)
- wxString(const wxWCharBuffer& psz) : wxStringBase(psz.data()) { }
-#else // ANSI
- // from C string (for compilers using unsigned char)
- wxString(const unsigned char* psz)
- : wxStringBase((const char*)psz) { }
- // from part of C string (for compilers using unsigned char)
- wxString(const unsigned char* psz, size_t nLength)
- : wxStringBase((const char*)psz, nLength) { }
+ // Unlike ctor from std::string, we provide conversion to std::string only
+ // if wxUSE_STL and not merely wxUSE_STD_STRING (which is on by default),
+ // because it conflicts with operator const char/wchar_t*:
+#if wxUSE_STL
+ #if wxUSE_UNICODE_WCHAR && wxUSE_STL_BASED_WXSTRING
+ // wxStringImpl is std::string in the encoding we want
+ operator const wxStdWideString&() const { return m_impl; }
+ #else
+ // wxStringImpl is either not std::string or needs conversion
+ operator wxStdWideString() const
+ // FIXME-UTF8: broken for embedded NULs
+ { return wxStdWideString(wc_str()); }
+ #endif
+
+ #if (!wxUSE_UNICODE || wxUSE_UTF8_LOCALE_ONLY) && wxUSE_STL_BASED_WXSTRING
+ // wxStringImpl is std::string in the encoding we want
+ operator const std::string&() const { return m_impl; }
+ #else
+ // wxStringImpl is either not std::string or needs conversion
+ operator std::string() const
+ // FIXME-UTF8: broken for embedded NULs
+ { return std::string(mb_str()); }
+ #endif
+#endif // wxUSE_STL
+
+ // first valid index position
+ const_iterator begin() const { return const_iterator(this, m_impl.begin()); }
+ iterator begin() { return iterator(this, m_impl.begin()); }
+ // position one after the last valid one
+ const_iterator end() const { return const_iterator(this, m_impl.end()); }
+ iterator end() { return iterator(this, m_impl.end()); }
+
+ // first element of the reversed string
+ const_reverse_iterator rbegin() const
+ { return const_reverse_iterator(end()); }
+ reverse_iterator rbegin()
+ { return reverse_iterator(end()); }
+ // one beyond the end of the reversed string
+ const_reverse_iterator rend() const
+ { return const_reverse_iterator(begin()); }
+ reverse_iterator rend()
+ { return reverse_iterator(begin()); }
+
+ // std::string methods:
+#if wxUSE_UNICODE_UTF8
+ size_t length() const { return end() - begin(); } // FIXME-UTF8: optimize!
+#else
+ size_t length() const { return m_impl.length(); }
+#endif
+
+ size_type size() const { return length(); }
+ size_type max_size() const { return npos; }
+
+ bool empty() const { return m_impl.empty(); }
-#if wxUSE_WCHAR_T
- // from wide (Unicode) string
- wxString(const wchar_t *pwz,
- const wxMBConv& conv = wxConvLibc,
- size_t nLength = npos);
-#endif // !wxUSE_WCHAR_T
+ size_type capacity() const { return m_impl.capacity(); } // FIXME-UTF8
+ void reserve(size_t sz) { m_impl.reserve(sz); } // FIXME-UTF8
- // from wxCharBuffer
- wxString(const wxCharBuffer& psz)
- : wxStringBase(psz) { }
-#endif // Unicode/ANSI
+ void resize(size_t nSize, wxUniChar ch = wxT('\0'))
+ {
+ const size_t len = length();
+ if ( nSize == len)
+ return;
+
+#if wxUSE_UNICODE_UTF8
+ if ( nSize < len )
+ {
+ // we can't use wxStringImpl::resize() for truncating the string as it
+ // counts in bytes, not characters
+ erase(nSize);
+ return;
+ }
+
+ // we also can't use (presumably more efficient) resize() if we have to
+ // append characters taking more than one byte
+ if ( !ch.IsAscii() )
+ append(nSize - len, ch);
+ else
+#endif // wxUSE_UNICODE_UTF8
+ m_impl.resize(nSize, (wxStringCharType)ch);
+ }
+
+ wxString substr(size_t nStart = 0, size_t nLen = npos) const
+ {
+ size_t pos, len;
+ PosLenToImpl(nStart, nLen, &pos, &len);
+ return FromImpl(m_impl.substr(pos, len));
+ }
// generic attributes & operations
// as standard strlen()
// data access (all indexes are 0 based)
// read access
+ wxUniChar at(size_t n) const
+ { return *(begin() + n); } // FIXME-UTF8: optimize?
wxUniChar GetChar(size_t n) const
{ return at(n); }
// read/write access
+ wxUniCharRef at(size_t n)
+ { return *(begin() + n); } // FIXME-UTF8: optimize?
wxUniCharRef GetWritableChar(size_t n)
{ return at(n); }
// write access
{ at(n) = ch; }
// get last character
- wxUniChar Last() const
- {
- wxASSERT_MSG( !empty(), _T("wxString: index out of bounds") );
-
- return at(length() - 1);
- }
+ wxUniChar Last() const
+ {
+ wxASSERT_MSG( !empty(), _T("wxString: index out of bounds") );
+ return *rbegin();
+ }
// get writable last character
wxUniCharRef Last()
- {
- wxASSERT_MSG( !empty(), _T("wxString: index out of bounds") );
- return at(length() - 1);
- }
+ {
+ wxASSERT_MSG( !empty(), _T("wxString: index out of bounds") );
+ return *rbegin();
+ }
/*
Note that we we must define all of the overloads below to avoid
ambiguity when using str[0].
*/
wxUniChar operator[](int n) const
- { return wxStringBase::at(n); }
+ { return at(n); }
wxUniChar operator[](long n) const
- { return wxStringBase::at(n); }
+ { return at(n); }
wxUniChar operator[](size_t n) const
- { return wxStringBase::at(n); }
+ { return at(n); }
#ifndef wxSIZE_T_IS_UINT
wxUniChar operator[](unsigned int n) const
- { return wxStringBase::at(n); }
+ { return at(n); }
#endif // size_t != unsigned int
// operator versions of GetWriteableChar()
wxUniCharRef operator[](int n)
- { return wxStringBase::at(n); }
+ { return at(n); }
wxUniCharRef operator[](long n)
- { return wxStringBase::at(n); }
+ { return at(n); }
wxUniCharRef operator[](size_t n)
- { return wxStringBase::at(n); }
+ { return at(n); }
#ifndef wxSIZE_T_IS_UINT
wxUniCharRef operator[](unsigned int n)
- { return wxStringBase::at(n); }
+ { return at(n); }
#endif // size_t != unsigned int
// explicit conversion to C string (use this with printf()!)
wxCStrData c_str() const { return wxCStrData(this); }
+ wxCStrData data() const { return c_str(); }
// implicit conversion to C string
operator wxCStrData() const { return c_str(); }
- operator const wxChar*() const { return c_str(); }
+
+ // the first two operators conflict with operators for conversion to
+ // std::string and they must be disabled in STL build; the next one only
+ // makes sense if conversions to char* are also defined and not defining it
+ // in STL build also helps us to get more clear error messages for the code
+ // which relies on implicit conversion to char* in STL build
+#if !wxUSE_STL
+ operator const char*() const { return c_str(); }
+ operator const wchar_t*() const { return c_str(); }
+
+ // implicit conversion to untyped pointer for compatibility with previous
+ // wxWidgets versions: this is the same as conversion to const char * so it
+ // may fail!
+ operator const void*() const { return c_str(); }
+#endif // wxUSE_STL
// identical to c_str(), for MFC compatibility
const wxCStrData GetData() const { return c_str(); }
// explicit conversion to C string in internal representation (char*,
// wchar_t*, UTF-8-encoded char*, depending on the build):
- const_pointer wx_str() const { return data(); }
+ const wxStringCharType *wx_str() const { return m_impl.c_str(); }
+
+ // conversion to *non-const* multibyte or widestring buffer; modifying
+ // returned buffer won't affect the string, these methods are only useful
+ // for passing values to const-incorrect functions
+ wxWritableCharBuffer char_str(const wxMBConv& conv = wxConvLibc) const
+ { return mb_str(conv); }
+ wxWritableWCharBuffer wchar_str() const { return wc_str(); }
// conversion to/from plain (i.e. 7 bit) ASCII: this is useful for
// converting numbers or strings which are certain not to contain special
// the behaviour of these functions with the strings containing anything
// else than 7 bit ASCII characters is undefined, use at your own risk.
#if wxUSE_UNICODE
- static wxString FromAscii(const char *ascii); // string
- static wxString FromAscii(const char ascii); // char
+ static wxString FromAscii(const char *ascii, size_t len);
+ static wxString FromAscii(const char *ascii);
+ static wxString FromAscii(char ascii);
const wxCharBuffer ToAscii() const;
#else // ANSI
static wxString FromAscii(const char *ascii) { return wxString( ascii ); }
- static wxString FromAscii(const char ascii) { return wxString( ascii ); }
+ static wxString FromAscii(const char *ascii, size_t len)
+ { return wxString( ascii, len ); }
+ static wxString FromAscii(char ascii) { return wxString( ascii ); }
const char *ToAscii() const { return c_str(); }
#endif // Unicode/!Unicode
+ // also provide unsigned char overloads as signed/unsigned doesn't matter
+ // for 7 bit ASCII characters
+ static wxString FromAscii(const unsigned char *ascii)
+ { return FromAscii((const char *)ascii); }
+ static wxString FromAscii(const unsigned char *ascii, size_t len)
+ { return FromAscii((const char *)ascii, len); }
+
+ // conversion to/from UTF-8:
+#if wxUSE_UNICODE_UTF8
+ static wxString FromUTF8(const char *utf8)
+ {
+ if ( !utf8 )
+ return wxEmptyString;
+
+ wxASSERT( wxStringOperations::IsValidUtf8String(utf8) );
+ return FromImpl(wxStringImpl(utf8));
+ }
+ static wxString FromUTF8(const char *utf8, size_t len)
+ {
+ if ( !utf8 )
+ return wxEmptyString;
+ if ( len == npos )
+ return FromUTF8(utf8);
+
+ wxASSERT( wxStringOperations::IsValidUtf8String(utf8, len) );
+ return FromImpl(wxStringImpl(utf8, len));
+ }
+ const char* utf8_str() const { return wx_str(); }
+ const char* ToUTF8() const { return wx_str(); }
+#elif wxUSE_UNICODE_WCHAR
+ static wxString FromUTF8(const char *utf8)
+ { return wxString(utf8, wxMBConvUTF8()); }
+ static wxString FromUTF8(const char *utf8, size_t len)
+ { return wxString(utf8, wxMBConvUTF8(), len); }
+ const wxCharBuffer utf8_str() const { return mb_str(wxMBConvUTF8()); }
+ const wxCharBuffer ToUTF8() const { return utf8_str(); }
+#else // ANSI
+ static wxString FromUTF8(const char *utf8)
+ { return wxString(wxMBConvUTF8().cMB2WC(utf8)); }
+ static wxString FromUTF8(const char *utf8, size_t len)
+ {
+ size_t wlen;
+ wxWCharBuffer buf(wxMBConvUTF8().cMB2WC(utf8, len == npos ? wxNO_LEN : len, &wlen));
+ return wxString(buf.data(), wlen);
+ }
+ const wxCharBuffer utf8_str() const
+ { return wxMBConvUTF8().cWC2MB(wc_str()); }
+ const wxCharBuffer ToUTF8() const { return utf8_str(); }
+#endif
+
+ // functions for storing binary data in wxString:
+#if wxUSE_UNICODE
+ static wxString From8BitData(const char *data, size_t len)
+ { return wxString(data, wxConvISO8859_1, len); }
+ // version for NUL-terminated data:
+ static wxString From8BitData(const char *data)
+ { return wxString(data, wxConvISO8859_1); }
+ const wxCharBuffer To8BitData() const { return mb_str(wxConvISO8859_1); }
+#else // ANSI
+ static wxString From8BitData(const char *data, size_t len)
+ { return wxString(data, len); }
+ // version for NUL-terminated data:
+ static wxString From8BitData(const char *data)
+ { return wxString(data); }
+ const char *To8BitData() const { return c_str(); }
+#endif // Unicode/ANSI
+
// conversions with (possible) format conversions: have to return a
// buffer with temporary data
//
// type differs because a function may either return pointer to the buffer
// directly or have to use intermediate buffer for translation.
#if wxUSE_UNICODE
+
+#if wxUSE_UTF8_LOCALE_ONLY
+ const char* mb_str() const { return wx_str(); }
+ const wxCharBuffer mb_str(const wxMBConv& conv) const;
+#else
const wxCharBuffer mb_str(const wxMBConv& conv = wxConvLibc) const;
+#endif
const wxWX2MBbuf mbc_str() const { return mb_str(*wxConvCurrent); }
- const wxChar* wc_str() const { return c_str(); }
-
+#if wxUSE_UNICODE_WCHAR
+ const wxChar* wc_str() const { return wx_str(); }
+#elif wxUSE_UNICODE_UTF8
+ const wxWCharBuffer wc_str() const;
+#endif
// for compatibility with !wxUSE_UNICODE version
- const wxChar* wc_str(const wxMBConv& WXUNUSED(conv)) const { return c_str(); }
+ const wxWX2WCbuf wc_str(const wxMBConv& WXUNUSED(conv)) const
+ { return wc_str(); }
#if wxMBFILES
const wxCharBuffer fn_str() const { return mb_str(wxConvFile); }
#else // !wxMBFILES
- const wxChar* fn_str() const { return c_str(); }
+ const wxWX2WCbuf fn_str() const { return wc_str(); }
#endif // wxMBFILES/!wxMBFILES
+
#else // ANSI
- const wxChar* mb_str() const { return c_str(); }
+ const wxChar* mb_str() const { return wx_str(); }
// for compatibility with wxUSE_UNICODE version
- const wxChar* mb_str(const wxMBConv& WXUNUSED(conv)) const { return c_str(); }
+ const wxChar* mb_str(const wxMBConv& WXUNUSED(conv)) const { return wx_str(); }
const wxWX2MBbuf mbc_str() const { return mb_str(); }
#if wxUSE_WCHAR_T
- const wxWCharBuffer wc_str(const wxMBConv& conv) const;
+ const wxWCharBuffer wc_str(const wxMBConv& conv = wxConvLibc) const;
#endif // wxUSE_WCHAR_T
#ifdef __WXOSX__
const wxCharBuffer fn_str() const { return wxConvFile.cWC2WX( wc_str( wxConvLocal ) ); }
// overloaded assignment
// from another wxString
- wxString& operator=(const wxStringBase& stringSrc)
- { return (wxString&)wxStringBase::operator=(stringSrc); }
- wxString& operator=(const wxCStrData& cstr);
+ wxString& operator=(const wxString& stringSrc)
+ { m_impl = stringSrc.m_impl; return *this; }
+ wxString& operator=(const wxCStrData& cstr)
+ { return *this = cstr.AsString(); }
// from a character
wxString& operator=(wxUniChar ch)
- { return (wxString&)wxStringBase::operator=(ch); }
+ { m_impl = wxStringOperations::EncodeChar(ch); return *this; }
wxString& operator=(wxUniCharRef ch)
- { return (wxString&)wxStringBase::operator=((wxUniChar)ch); }
+ { return operator=((wxUniChar)ch); }
wxString& operator=(char ch)
- { return (wxString&)wxStringBase::operator=(wxUniChar(ch)); }
+ { return operator=(wxUniChar(ch)); }
+ wxString& operator=(unsigned char ch)
+ { return operator=(wxUniChar(ch)); }
wxString& operator=(wchar_t ch)
- { return (wxString&)wxStringBase::operator=(wxUniChar(ch)); }
+ { return operator=(wxUniChar(ch)); }
// from a C string - STL probably will crash on NULL,
// so we need to compensate in that case
#if wxUSE_STL_BASED_WXSTRING
- wxString& operator=(const wxChar *psz)
- { if(psz) wxStringBase::operator=(psz); else Clear(); return *this; }
+ wxString& operator=(const char *psz)
+ { if (psz) m_impl = ImplStr(psz); else Clear(); return *this; }
+ wxString& operator=(const wchar_t *pwz)
+ { if (pwz) m_impl = ImplStr(pwz); else Clear(); return *this; }
#else
- wxString& operator=(const wxChar *psz)
- { return (wxString&)wxStringBase::operator=(psz); }
+ wxString& operator=(const char *psz)
+ { m_impl = ImplStr(psz); return *this; }
+ wxString& operator=(const wchar_t *pwz)
+ { m_impl = ImplStr(pwz); return *this; }
#endif
+ wxString& operator=(const unsigned char *psz)
+ { return operator=((const char*)psz); }
-#if wxUSE_UNICODE
// from wxWCharBuffer
- wxString& operator=(const wxWCharBuffer& psz)
- { (void) operator=((const wchar_t *)psz); return *this; }
- // from C string
- wxString& operator=(const char* psz)
- { return operator=(wxString(psz)); }
-#else // ANSI
- // from another kind of C string
- wxString& operator=(const unsigned char* psz);
-#if wxUSE_WCHAR_T
- // from a wide string
- wxString& operator=(const wchar_t *pwz);
-#endif
+ wxString& operator=(const wxWCharBuffer& s)
+ { return operator=(s.data()); } // FIXME-UTF8: fix for embedded NULs
// from wxCharBuffer
- wxString& operator=(const wxCharBuffer& psz)
- { (void) operator=((const char *)psz); return *this; }
-#endif // Unicode/ANSI
+ wxString& operator=(const wxCharBuffer& s)
+ { return operator=(s.data()); } // FIXME-UTF8: fix for embedded NULs
// string concatenation
// in place concatenation
// string += string
wxString& operator<<(const wxString& s)
{
-#if WXWIN_COMPATIBILITY_2_8 && !wxUSE_STL
- wxASSERT_MSG( s.GetStringData()->IsValid(),
+#if WXWIN_COMPATIBILITY_2_8 && !wxUSE_STL_BASED_WXSTRING && !wxUSE_UNICODE_UTF8
+ wxASSERT_MSG( s.IsValid(),
_T("did you forget to call UngetWriteBuf()?") );
#endif
return *this;
}
// string += C string
- wxString& operator<<(const wxChar *psz)
+ wxString& operator<<(const char *psz)
{ append(psz); return *this; }
+ wxString& operator<<(const wchar_t *pwz)
+ { append(pwz); return *this; }
wxString& operator<<(const wxCStrData& psz)
- { append(psz); return *this; }
+ { append(psz.AsString()); return *this; }
// string += char
wxString& operator<<(wxUniChar ch) { append(1, ch); return *this; }
wxString& operator<<(wxUniCharRef ch) { append(1, ch); return *this; }
wxString& operator<<(char ch) { append(1, ch); return *this; }
+ wxString& operator<<(unsigned char ch) { append(1, ch); return *this; }
wxString& operator<<(wchar_t ch) { append(1, ch); return *this; }
// string += buffer (i.e. from wxGetString)
-#if wxUSE_UNICODE
wxString& operator<<(const wxWCharBuffer& s)
{ return operator<<((const wchar_t *)s); }
- wxString& operator+=(const wxWCharBuffer& s)
- { return operator<<((const wchar_t *)s); }
-#else // !wxUSE_UNICODE
wxString& operator<<(const wxCharBuffer& s)
{ return operator<<((const char *)s); }
- wxString& operator+=(const wxCharBuffer& s)
- { return operator<<((const char *)s); }
-#endif // wxUSE_UNICODE/!wxUSE_UNICODE
-
-#if wxUSE_UNICODE
- // string += C string in Unicode build (with conversion)
- wxString& operator<<(const char *s)
- { return operator<<(wxString(s)); }
- wxString& operator+=(const char *s)
- { return operator+=(wxString(s)); }
-#endif // wxUSE_UNICODE
// string += C string
wxString& Append(const wxString& s)
append(s);
return *this;
}
+ wxString& Append(const char* psz)
+ { append(psz); return *this; }
+ wxString& Append(const wchar_t* pwz)
+ { append(pwz); return *this; }
wxString& Append(const wxCStrData& psz)
{ append(psz); return *this; }
- wxString& Append(const wxChar* psz)
+ wxString& Append(const wxCharBuffer& psz)
+ { append(psz); return *this; }
+ wxString& Append(const wxWCharBuffer& psz)
{ append(psz); return *this; }
// append count copies of given character
wxString& Append(wxUniChar ch, size_t count = 1u)
{ append(count, ch); return *this; }
wxString& Append(char ch, size_t count = 1u)
{ append(count, ch); return *this; }
+ wxString& Append(unsigned char ch, size_t count = 1u)
+ { append(count, ch); return *this; }
wxString& Append(wchar_t ch, size_t count = 1u)
{ append(count, ch); return *this; }
- wxString& Append(const wxChar* psz, size_t nLen)
+ wxString& Append(const char* psz, size_t nLen)
{ append(psz, nLen); return *this; }
+ wxString& Append(const wchar_t* pwz, size_t nLen)
+ { append(pwz, nLen); return *this; }
// prepend a string, return the string itself
wxString& Prepend(const wxString& str)
friend wxString WXDLLIMPEXP_BASE operator+(wxUniChar ch, const wxString& string);
// string with C string
friend wxString WXDLLIMPEXP_BASE operator+(const wxString& string,
- const wxChar *psz);
+ const char *psz);
+ friend wxString WXDLLIMPEXP_BASE operator+(const wxString& string,
+ const wchar_t *pwz);
// C string with string
- friend wxString WXDLLIMPEXP_BASE operator+(const wxChar *psz,
+ friend wxString WXDLLIMPEXP_BASE operator+(const char *psz,
+ const wxString& string);
+ friend wxString WXDLLIMPEXP_BASE operator+(const wchar_t *pwz,
const wxString& string);
// stream-like functions
// string comparison
// case-sensitive comparison (returns a value < 0, = 0 or > 0)
- int Cmp(const wxChar *psz) const;
- int Cmp(const wxString& s) const;
+ int Cmp(const char *psz) const
+ { return compare(psz); }
+ int Cmp(const wchar_t *pwz) const
+ { return compare(pwz); }
+ int Cmp(const wxString& s) const
+ { return compare(s); }
+ int Cmp(const wxCStrData& s) const
+ { return compare(s); }
+ int Cmp(const wxCharBuffer& s) const
+ { return compare(s); }
+ int Cmp(const wxWCharBuffer& s) const
+ { return compare(s); }
// same as Cmp() but not case-sensitive
- int CmpNoCase(const wxChar *psz) const;
int CmpNoCase(const wxString& s) const;
// test for the string equality, either considering case or not
// (if compareWithCase then the case matters)
- bool IsSameAs(const wxChar *psz, bool compareWithCase = true) const
- { return (compareWithCase ? Cmp(psz) : CmpNoCase(psz)) == 0; }
+ bool IsSameAs(const wxString& str, bool compareWithCase = true) const
+ { return (compareWithCase ? Cmp(str) : CmpNoCase(str)) == 0; }
+ bool IsSameAs(const char *str, bool compareWithCase = true) const
+ { return (compareWithCase ? Cmp(str) : CmpNoCase(str)) == 0; }
+ bool IsSameAs(const wchar_t *str, bool compareWithCase = true) const
+ { return (compareWithCase ? Cmp(str) : CmpNoCase(str)) == 0; }
+ bool IsSameAs(const wxCStrData& str, bool compareWithCase = true) const
+ { return IsSameAs(str.AsString(), compareWithCase); }
+ bool IsSameAs(const wxCharBuffer& str, bool compareWithCase = true) const
+ { return IsSameAs(str.data(), compareWithCase); }
+ bool IsSameAs(const wxWCharBuffer& str, bool compareWithCase = true) const
+ { return IsSameAs(str.data(), compareWithCase); }
// comparison with a single character: returns true if equal
- bool IsSameAs(wxUniChar c, bool compareWithCase = true) const
- {
- return (length() == 1) && (compareWithCase ? GetChar(0u) == c
- : wxToupper(GetChar(0u)) == wxToupper(c));
- }
+ bool IsSameAs(wxUniChar c, bool compareWithCase = true) const;
+ // FIXME-UTF8: remove these overloads
+ bool IsSameAs(wxUniCharRef c, bool compareWithCase = true) const
+ { return IsSameAs(wxUniChar(c), compareWithCase); }
+ bool IsSameAs(char c, bool compareWithCase = true) const
+ { return IsSameAs(wxUniChar(c), compareWithCase); }
+ bool IsSameAs(unsigned char c, bool compareWithCase = true) const
+ { return IsSameAs(wxUniChar(c), compareWithCase); }
+ bool IsSameAs(wchar_t c, bool compareWithCase = true) const
+ { return IsSameAs(wxUniChar(c), compareWithCase); }
+ bool IsSameAs(int c, bool compareWithCase = true) const
+ { return IsSameAs(wxUniChar(c), compareWithCase); }
// simple sub-string extraction
// return substring starting at nFirst of length nCount (or till the end
// check if the string starts with the given prefix and return the rest
// of the string in the provided pointer if it is not NULL; otherwise
// return false
- bool StartsWith(const wxChar *prefix, wxString *rest = NULL) const;
+ bool StartsWith(const wxString& prefix, wxString *rest = NULL) const;
// check if the string ends with the given suffix and return the
// beginning of the string before the suffix in the provided pointer if
// it is not NULL; otherwise return false
- bool EndsWith(const wxChar *suffix, wxString *rest = NULL) const;
+ bool EndsWith(const wxString& suffix, wxString *rest = NULL) const;
// get first nCount characters
wxString Left(size_t nCount) const;
// searching and replacing
// searching (return starting index, or -1 if not found)
int Find(wxUniChar ch, bool bFromEnd = false) const; // like strchr/strrchr
+ int Find(wxUniCharRef ch, bool bFromEnd = false) const
+ { return Find(wxUniChar(ch), bFromEnd); }
+ int Find(char ch, bool bFromEnd = false) const
+ { return Find(wxUniChar(ch), bFromEnd); }
+ int Find(unsigned char ch, bool bFromEnd = false) const
+ { return Find(wxUniChar(ch), bFromEnd); }
+ int Find(wchar_t ch, bool bFromEnd = false) const
+ { return Find(wxUniChar(ch), bFromEnd); }
// searching (return starting index, or -1 if not found)
- int Find(const wxChar *pszSub) const; // like strstr
+ int Find(const wxString& sub) const // like strstr
+ {
+ size_type idx = find(sub);
+ return (idx == npos) ? wxNOT_FOUND : (int)idx;
+ }
+ int Find(const char *sub) const // like strstr
+ {
+ size_type idx = find(sub);
+ return (idx == npos) ? wxNOT_FOUND : (int)idx;
+ }
+ int Find(const wchar_t *sub) const // like strstr
+ {
+ size_type idx = find(sub);
+ return (idx == npos) ? wxNOT_FOUND : (int)idx;
+ }
+
+ int Find(const wxCStrData& sub) const
+ { return Find(sub.AsString()); }
+ int Find(const wxCharBuffer& sub) const
+ { return Find(sub.data()); }
+ int Find(const wxWCharBuffer& sub) const
+ { return Find(sub.data()); }
+
// replace first (or all of bReplaceAll) occurences of substring with
// another string, returns the number of replacements made
- size_t Replace(const wxChar *szOld,
- const wxChar *szNew,
+ size_t Replace(const wxString& strOld,
+ const wxString& strNew,
bool bReplaceAll = true);
// check if the string contents matches a mask containing '*' and '?'
- bool Matches(const wxChar *szMask) const;
+ bool Matches(const wxString& mask) const;
// conversion to numbers: all functions return true only if the whole
// string is a number and put the value of this number into the pointer
// formatted input/output
// as sprintf(), returns the number of characters written or < 0 on error
// (take 'this' into account in attribute parameter count)
- // int Printf(const wxChar *pszFormat, ...);
- WX_DEFINE_VARARG_FUNC(int, Printf, DoPrintf)
+ // int Printf(const wxString& format, ...);
+ WX_DEFINE_VARARG_FUNC(int, Printf, 1, (const wxFormatString&),
+ DoPrintfWchar, DoPrintfUtf8)
+#ifdef __WATCOMC__
+ // workaround for http://bugzilla.openwatcom.org/show_bug.cgi?id=351
+ WX_VARARG_WATCOM_WORKAROUND(int, Printf, 1, (const wxString&),
+ (wxFormatString(f1)));
+ WX_VARARG_WATCOM_WORKAROUND(int, Printf, 1, (const wxCStrData&),
+ (wxFormatString(f1)));
+ WX_VARARG_WATCOM_WORKAROUND(int, Printf, 1, (const char*),
+ (wxFormatString(f1)));
+ WX_VARARG_WATCOM_WORKAROUND(int, Printf, 1, (const wchar_t*),
+ (wxFormatString(f1)));
+#endif
#endif // !wxNEEDS_WXSTRING_PRINTF_MIXIN
// as vprintf(), returns the number of characters written or < 0 on error
int PrintfV(const wxString& format, va_list argptr);
#ifndef wxNEEDS_WXSTRING_PRINTF_MIXIN
// returns the string containing the result of Printf() to it
- // static wxString Format(const wxChar *pszFormat, ...) ATTRIBUTE_PRINTF_1;
- WX_DEFINE_VARARG_FUNC(static wxString, Format, DoFormat)
+ // static wxString Format(const wxString& format, ...) ATTRIBUTE_PRINTF_1;
+ WX_DEFINE_VARARG_FUNC(static wxString, Format, 1, (const wxFormatString&),
+ DoFormatWchar, DoFormatUtf8)
+#ifdef __WATCOMC__
+ // workaround for http://bugzilla.openwatcom.org/show_bug.cgi?id=351
+ WX_VARARG_WATCOM_WORKAROUND(static wxString, Format, 1, (const wxString&),
+ (wxFormatString(f1)));
+ WX_VARARG_WATCOM_WORKAROUND(static wxString, Format, 1, (const wxCStrData&),
+ (wxFormatString(f1)));
+ WX_VARARG_WATCOM_WORKAROUND(static wxString, Format, 1, (const char*),
+ (wxFormatString(f1)));
+ WX_VARARG_WATCOM_WORKAROUND(static wxString, Format, 1, (const wchar_t*),
+ (wxFormatString(f1)));
+#endif
#endif
// the same as above, but takes a va_list
static wxString FormatV(const wxString& format, va_list argptr);
// minimize the string's memory
// only works if the data of this string is not shared
bool Shrink();
-#if WXWIN_COMPATIBILITY_2_8 && !wxUSE_STL
+#if WXWIN_COMPATIBILITY_2_8 && !wxUSE_STL_BASED_WXSTRING && !wxUSE_UNICODE_UTF8
// These are deprecated, use wxStringBuffer or wxStringBufferLength instead
//
// get writable buffer of at least nLen bytes. Unget() *must* be called
// a.s.a.p. to put string back in a reasonable state!
- wxDEPRECATED( wxChar *GetWriteBuf(size_t nLen) );
+ wxDEPRECATED( wxStringCharType *GetWriteBuf(size_t nLen) );
// call this immediately after GetWriteBuf() has been used
wxDEPRECATED( void UngetWriteBuf() );
wxDEPRECATED( void UngetWriteBuf(size_t nLen) );
-#endif // WXWIN_COMPATIBILITY_2_8 && !wxUSE_STL
+#endif // WXWIN_COMPATIBILITY_2_8 && !wxUSE_STL_BASED_WXSTRING && wxUSE_UNICODE_UTF8
// wxWidgets version 1 compatibility functions
#ifndef wxNEEDS_WXSTRING_PRINTF_MIXIN
// use Printf()
// (take 'this' into account in attribute parameter count)
- // int sprintf(const wxChar *pszFormat, ...) ATTRIBUTE_PRINTF_2;
- WX_DEFINE_VARARG_FUNC(int, sprintf, DoPrintf)
+ // int sprintf(const wxString& format, ...) ATTRIBUTE_PRINTF_2;
+ WX_DEFINE_VARARG_FUNC(int, sprintf, 1, (const wxFormatString&),
+ DoPrintfWchar, DoPrintfUtf8)
+#ifdef __WATCOMC__
+ // workaround for http://bugzilla.openwatcom.org/show_bug.cgi?id=351
+ WX_VARARG_WATCOM_WORKAROUND(int, sprintf, 1, (const wxString&),
+ (wxFormatString(f1)));
+ WX_VARARG_WATCOM_WORKAROUND(int, sprintf, 1, (const wxCStrData&),
+ (wxFormatString(f1)));
+ WX_VARARG_WATCOM_WORKAROUND(int, sprintf, 1, (const char*),
+ (wxFormatString(f1)));
+ WX_VARARG_WATCOM_WORKAROUND(int, sprintf, 1, (const wchar_t*),
+ (wxFormatString(f1)));
+#endif
#endif // wxNEEDS_WXSTRING_PRINTF_MIXIN
// use Cmp()
inline int CompareTo(const wxChar* psz, caseCompare cmp = exact) const
{ return cmp == exact ? Cmp(psz) : CmpNoCase(psz); }
- // use Len
+ // use length()
size_t Length() const { return length(); }
// Count the number of characters
int Freq(wxUniChar ch) const;
{ return (wxString&)erase( nStart, nLen ); }
// use Find()
- int First( const wxUniChar ch ) const { return Find(ch); }
+ int First( wxUniChar ch ) const { return Find(ch); }
+ int First( wxUniCharRef ch ) const { return Find(ch); }
int First( char ch ) const { return Find(ch); }
+ int First( unsigned char ch ) const { return Find(ch); }
int First( wchar_t ch ) const { return Find(ch); }
- int First( const wxChar* psz ) const { return Find(psz); }
- int First( const wxString &str ) const { return Find(str); }
- int Last( const wxUniChar ch ) const { return Find(ch, true); }
+ int First( const wxString& str ) const { return Find(str); }
+ int Last( wxUniChar ch ) const { return Find(ch, true); }
bool Contains(const wxString& str) const { return Find(str) != wxNOT_FOUND; }
// use empty()
// take nLen chars starting at nPos
wxString(const wxString& str, size_t nPos, size_t nLen)
- : wxStringBase(str, nPos, nLen) { }
- // take all characters from pStart to pEnd
- wxString(const void *pStart, const void *pEnd)
- : wxStringBase((const wxChar*)pStart, (const wxChar*)pEnd) { }
+ { assign(str, nPos, nLen); }
+ // take all characters from first to last
wxString(const_iterator first, const_iterator last)
- : wxStringBase(first, last) { }
- wxString(iterator first, iterator last)
- : wxStringBase(first, last) { }
+ : m_impl(first.impl(), last.impl()) { }
+#if WXWIN_COMPATIBILITY_STRING_PTR_AS_ITER
+ // the 2 overloads below are for compatibility with the existing code using
+ // pointers instead of iterators
+ wxString(const char *first, const char *last)
+ {
+ SubstrBufFromMB str(ImplStr(first, last - first));
+ m_impl.assign(str.data, str.len);
+ }
+ wxString(const wchar_t *first, const wchar_t *last)
+ {
+ SubstrBufFromWC str(ImplStr(first, last - first));
+ m_impl.assign(str.data, str.len);
+ }
+ // and this one is needed to compile code adding offsets to c_str() result
+ wxString(const wxCStrData& first, const wxCStrData& last)
+ : m_impl(CreateConstIterator(first).impl(),
+ CreateConstIterator(last).impl())
+ {
+ wxASSERT_MSG( first.m_str == last.m_str,
+ _T("pointers must be into the same string") );
+ }
+#endif // WXWIN_COMPATIBILITY_STRING_PTR_AS_ITER
// lib.string.modifiers
// append elements str[pos], ..., str[pos+n]
wxString& append(const wxString& str, size_t pos, size_t n)
- { return (wxString&)wxStringBase::append(str, pos, n); }
+ {
+ size_t from, len;
+ str.PosLenToImpl(pos, n, &from, &len);
+ m_impl.append(str.m_impl, from, len);
+ return *this;
+ }
// append a string
wxString& append(const wxString& str)
- { return (wxString&)wxStringBase::append(str); }
- wxString& append(const wxCStrData& str)
- { return (wxString&)wxStringBase::append(str.AsString()); }
+ { m_impl.append(str.m_impl); return *this; }
// append first n (or all if n == npos) characters of sz
- wxString& append(const wxChar *sz)
- { return (wxString&)wxStringBase::append(sz); }
- wxString& append(const wxChar *sz, size_t n)
- { return (wxString&)wxStringBase::append(sz, n); }
+ wxString& append(const char *sz)
+ { m_impl.append(ImplStr(sz)); return *this; }
+ wxString& append(const wchar_t *sz)
+ { m_impl.append(ImplStr(sz)); return *this; }
+ wxString& append(const char *sz, size_t n)
+ {
+ SubstrBufFromMB str(ImplStr(sz, n));
+ m_impl.append(str.data, str.len);
+ return *this;
+ }
+ wxString& append(const wchar_t *sz, size_t n)
+ {
+ SubstrBufFromWC str(ImplStr(sz, n));
+ m_impl.append(str.data, str.len);
+ return *this;
+ }
+
+ wxString& append(const wxCStrData& str)
+ { return append(str.AsString()); }
+ wxString& append(const wxCharBuffer& str)
+ { return append(str.data()); }
+ wxString& append(const wxWCharBuffer& str)
+ { return append(str.data()); }
+
// append n copies of ch
wxString& append(size_t n, wxUniChar ch)
- { return (wxString&)wxStringBase::append(n, ch); }
+ {
+#if wxUSE_UNICODE_UTF8
+ if ( !ch.IsAscii() )
+ m_impl.append(wxStringOperations::EncodeNChars(n, ch));
+ else
+#endif
+ m_impl.append(n, (wxStringCharType)ch);
+ return *this;
+ }
// append from first to last
wxString& append(const_iterator first, const_iterator last)
- { return (wxString&)wxStringBase::append(first, last); }
+ { m_impl.append(first.impl(), last.impl()); return *this; }
+#if WXWIN_COMPATIBILITY_STRING_PTR_AS_ITER
+ wxString& append(const char *first, const char *last)
+ { return append(first, last - first); }
+ wxString& append(const wchar_t *first, const wchar_t *last)
+ { return append(first, last - first); }
+ wxString& append(const wxCStrData& first, const wxCStrData& last)
+ { return append(CreateConstIterator(first), CreateConstIterator(last)); }
+#endif // WXWIN_COMPATIBILITY_STRING_PTR_AS_ITER
// same as `this_string = str'
wxString& assign(const wxString& str)
- { return (wxString&)wxStringBase::assign(str); }
+ { m_impl = str.m_impl; return *this; }
+ wxString& assign(const wxString& str, size_t len)
+ {
+ m_impl.assign(str.m_impl, 0, str.LenToImpl(len));
+ return *this;
+ }
// same as ` = str[pos..pos + n]
wxString& assign(const wxString& str, size_t pos, size_t n)
- { return (wxString&)wxStringBase::assign(str, pos, n); }
+ {
+ size_t from, len;
+ str.PosLenToImpl(pos, n, &from, &len);
+ m_impl.assign(str.m_impl, from, len);
+ return *this;
+ }
// same as `= first n (or all if n == npos) characters of sz'
- wxString& assign(const wxChar *sz)
- { return (wxString&)wxStringBase::assign(sz); }
- wxString& assign(const wxChar *sz, size_t n)
- { return (wxString&)wxStringBase::assign(sz, n); }
+ wxString& assign(const char *sz)
+ { m_impl.assign(ImplStr(sz)); return *this; }
+ wxString& assign(const wchar_t *sz)
+ { m_impl.assign(ImplStr(sz)); return *this; }
+ wxString& assign(const char *sz, size_t n)
+ {
+ SubstrBufFromMB str(ImplStr(sz, n));
+ m_impl.assign(str.data, str.len);
+ return *this;
+ }
+ wxString& assign(const wchar_t *sz, size_t n)
+ {
+ SubstrBufFromWC str(ImplStr(sz, n));
+ m_impl.assign(str.data, str.len);
+ return *this;
+ }
+
+ wxString& assign(const wxCStrData& str)
+ { return assign(str.AsString()); }
+ wxString& assign(const wxCharBuffer& str)
+ { return assign(str.data()); }
+ wxString& assign(const wxWCharBuffer& str)
+ { return assign(str.data()); }
+ wxString& assign(const wxCStrData& str, size_t len)
+ { return assign(str.AsString(), len); }
+ wxString& assign(const wxCharBuffer& str, size_t len)
+ { return assign(str.data(), len); }
+ wxString& assign(const wxWCharBuffer& str, size_t len)
+ { return assign(str.data(), len); }
+
// same as `= n copies of ch'
wxString& assign(size_t n, wxUniChar ch)
- { return (wxString&)wxStringBase::assign(n, ch); }
+ {
+#if wxUSE_UNICODE_UTF8
+ if ( !ch.IsAscii() )
+ m_impl.assign(wxStringOperations::EncodeNChars(n, ch));
+ else
+#endif
+ m_impl.assign(n, (wxStringCharType)ch);
+ return *this;
+ }
+
+ wxString& assign(size_t n, wxUniCharRef ch)
+ { return assign(n, wxUniChar(ch)); }
+ wxString& assign(size_t n, char ch)
+ { return assign(n, wxUniChar(ch)); }
+ wxString& assign(size_t n, unsigned char ch)
+ { return assign(n, wxUniChar(ch)); }
+ wxString& assign(size_t n, wchar_t ch)
+ { return assign(n, wxUniChar(ch)); }
+
// assign from first to last
wxString& assign(const_iterator first, const_iterator last)
- { return (wxString&)wxStringBase::assign(first, last); }
+ { m_impl.assign(first.impl(), last.impl()); return *this; }
+#if WXWIN_COMPATIBILITY_STRING_PTR_AS_ITER
+ wxString& assign(const char *first, const char *last)
+ { return assign(first, last - first); }
+ wxString& assign(const wchar_t *first, const wchar_t *last)
+ { return assign(first, last - first); }
+ wxString& assign(const wxCStrData& first, const wxCStrData& last)
+ { return assign(CreateConstIterator(first), CreateConstIterator(last)); }
+#endif // WXWIN_COMPATIBILITY_STRING_PTR_AS_ITER
// string comparison
-#if !defined(HAVE_STD_STRING_COMPARE)
- int compare(const wxStringBase& str) const;
+ int compare(const wxString& str) const;
+ int compare(const char* sz) const;
+ int compare(const wchar_t* sz) const;
+ int compare(const wxCStrData& str) const
+ { return compare(str.AsString()); }
+ int compare(const wxCharBuffer& str) const
+ { return compare(str.data()); }
+ int compare(const wxWCharBuffer& str) const
+ { return compare(str.data()); }
// comparison with a substring
- int compare(size_t nStart, size_t nLen, const wxStringBase& str) const;
+ int compare(size_t nStart, size_t nLen, const wxString& str) const;
// comparison of 2 substrings
int compare(size_t nStart, size_t nLen,
- const wxStringBase& str, size_t nStart2, size_t nLen2) const;
- // just like strcmp()
- int compare(const wxChar* sz) const;
+ const wxString& str, size_t nStart2, size_t nLen2) const;
// substring comparison with first nCount characters of sz
int compare(size_t nStart, size_t nLen,
- const wxChar* sz, size_t nCount = npos) const;
-#endif // !defined HAVE_STD_STRING_COMPARE
+ const char* sz, size_t nCount = npos) const;
+ int compare(size_t nStart, size_t nLen,
+ const wchar_t* sz, size_t nCount = npos) const;
// insert another string
wxString& insert(size_t nPos, const wxString& str)
- { return (wxString&)wxStringBase::insert(nPos, str); }
+ { insert(begin() + nPos, str.begin(), str.end()); return *this; }
// insert n chars of str starting at nStart (in str)
wxString& insert(size_t nPos, const wxString& str, size_t nStart, size_t n)
- { return (wxString&)wxStringBase::insert(nPos, str, nStart, n); }
+ {
+ size_t from, len;
+ str.PosLenToImpl(nStart, n, &from, &len);
+ m_impl.insert(PosToImpl(nPos), str.m_impl, from, len);
+ return *this;
+ }
// insert first n (or all if n == npos) characters of sz
- wxString& insert(size_t nPos, const wxChar *sz)
- { return (wxString&)wxStringBase::insert(nPos, sz); }
- wxString& insert(size_t nPos, const wxChar *sz, size_t n)
- { return (wxString&)wxStringBase::insert(nPos, sz, n); }
+ wxString& insert(size_t nPos, const char *sz)
+ { m_impl.insert(PosToImpl(nPos), ImplStr(sz)); return *this; }
+ wxString& insert(size_t nPos, const wchar_t *sz)
+ { m_impl.insert(PosToImpl(nPos), ImplStr(sz)); return *this; }
+ wxString& insert(size_t nPos, const char *sz, size_t n)
+ {
+ SubstrBufFromMB str(ImplStr(sz, n));
+ m_impl.insert(PosToImpl(nPos), str.data, str.len);
+ return *this;
+ }
+ wxString& insert(size_t nPos, const wchar_t *sz, size_t n)
+ {
+ SubstrBufFromWC str(ImplStr(sz, n));
+ m_impl.insert(PosToImpl(nPos), str.data, str.len);
+ return *this;
+ }
// insert n copies of ch
wxString& insert(size_t nPos, size_t n, wxUniChar ch)
- { return (wxString&)wxStringBase::insert(nPos, n, ch); }
+ {
+#if wxUSE_UNICODE_UTF8
+ if ( !ch.IsAscii() )
+ m_impl.insert(PosToImpl(nPos), wxStringOperations::EncodeNChars(n, ch));
+ else
+#endif
+ m_impl.insert(PosToImpl(nPos), n, (wxStringCharType)ch);
+ return *this;
+ }
iterator insert(iterator it, wxUniChar ch)
- { return wxStringBase::insert(it, ch); }
+ {
+#if wxUSE_UNICODE_UTF8
+ if ( !ch.IsAscii() )
+ {
+ size_t pos = IterToImplPos(it);
+ m_impl.insert(pos, wxStringOperations::EncodeChar(ch));
+ return iterator(this, m_impl.begin() + pos);
+ }
+ else
+#endif
+ return iterator(this, m_impl.insert(it.impl(), (wxStringCharType)ch));
+ }
void insert(iterator it, const_iterator first, const_iterator last)
- { wxStringBase::insert(it, first, last); }
+ { m_impl.insert(it.impl(), first.impl(), last.impl()); }
+#if WXWIN_COMPATIBILITY_STRING_PTR_AS_ITER
+ void insert(iterator it, const char *first, const char *last)
+ { insert(it - begin(), first, last - first); }
+ void insert(iterator it, const wchar_t *first, const wchar_t *last)
+ { insert(it - begin(), first, last - first); }
+ void insert(iterator it, const wxCStrData& first, const wxCStrData& last)
+ { insert(it, CreateConstIterator(first), CreateConstIterator(last)); }
+#endif // WXWIN_COMPATIBILITY_STRING_PTR_AS_ITER
+
void insert(iterator it, size_type n, wxUniChar ch)
- { wxStringBase::insert(it, n, ch); }
+ {
+#if wxUSE_UNICODE_UTF8
+ if ( !ch.IsAscii() )
+ m_impl.insert(IterToImplPos(it), wxStringOperations::EncodeNChars(n, ch));
+ else
+#endif
+ m_impl.insert(it.impl(), n, (wxStringCharType)ch);
+ }
+
+ // delete characters from nStart to nStart + nLen
+ wxString& erase(size_type pos = 0, size_type n = npos)
+ {
+ size_t from, len;
+ PosLenToImpl(pos, n, &from, &len);
+ m_impl.erase(from, len);
+ return *this;
+ }
+ // delete characters from first up to last
+ iterator erase(iterator first, iterator last)
+ { return iterator(this, m_impl.erase(first.impl(), last.impl())); }
+ iterator erase(iterator first)
+ { return iterator(this, m_impl.erase(first.impl())); }
+
+#ifdef wxSTRING_BASE_HASNT_CLEAR
+ void clear() { erase(); }
+#else
+ void clear() { m_impl.clear(); }
+#endif
+
+ // replaces the substring of length nLen starting at nStart
+ wxString& replace(size_t nStart, size_t nLen, const char* sz)
+ {
+ size_t from, len;
+ PosLenToImpl(nStart, nLen, &from, &len);
+ m_impl.replace(from, len, ImplStr(sz));
+ return *this;
+ }
+ wxString& replace(size_t nStart, size_t nLen, const wchar_t* sz)
+ {
+ size_t from, len;
+ PosLenToImpl(nStart, nLen, &from, &len);
+ m_impl.replace(from, len, ImplStr(sz));
+ return *this;
+ }
+ // replaces the substring of length nLen starting at nStart
+ wxString& replace(size_t nStart, size_t nLen, const wxString& str)
+ {
+ size_t from, len;
+ PosLenToImpl(nStart, nLen, &from, &len);
+ m_impl.replace(from, len, str.m_impl);
+ return *this;
+ }
+ // replaces the substring with nCount copies of ch
+ wxString& replace(size_t nStart, size_t nLen, size_t nCount, wxUniChar ch)
+ {
+ size_t from, len;
+ PosLenToImpl(nStart, nLen, &from, &len);
+#if wxUSE_UNICODE_UTF8
+ if ( !ch.IsAscii() )
+ m_impl.replace(from, len, wxStringOperations::EncodeNChars(nCount, ch));
+ else
+#endif
+ m_impl.replace(from, len, nCount, (wxStringCharType)ch);
+ return *this;
+ }
+ // replaces a substring with another substring
+ wxString& replace(size_t nStart, size_t nLen,
+ const wxString& str, size_t nStart2, size_t nLen2)
+ {
+ size_t from, len;
+ PosLenToImpl(nStart, nLen, &from, &len);
+
+ size_t from2, len2;
+ str.PosLenToImpl(nStart2, nLen2, &from2, &len2);
+
+ m_impl.replace(from, len, str.m_impl, from2, len2);
+ return *this;
+ }
+ // replaces the substring with first nCount chars of sz
+ wxString& replace(size_t nStart, size_t nLen,
+ const char* sz, size_t nCount)
+ {
+ size_t from, len;
+ PosLenToImpl(nStart, nLen, &from, &len);
+
+ SubstrBufFromMB str(ImplStr(sz, nCount));
+
+ m_impl.replace(from, len, str.data, str.len);
+ return *this;
+ }
+ wxString& replace(size_t nStart, size_t nLen,
+ const wchar_t* sz, size_t nCount)
+ {
+ size_t from, len;
+ PosLenToImpl(nStart, nLen, &from, &len);
+
+ SubstrBufFromWC str(ImplStr(sz, nCount));
+
+ m_impl.replace(from, len, str.data, str.len);
+ return *this;
+ }
+ wxString& replace(size_t nStart, size_t nLen,
+ const wxString& s, size_t nCount)
+ {
+ size_t from, len;
+ PosLenToImpl(nStart, nLen, &from, &len);
+ m_impl.replace(from, len, s.m_impl.c_str(), s.LenToImpl(nCount));
+ return *this;
+ }
+
+ wxString& replace(iterator first, iterator last, const char* s)
+ { m_impl.replace(first.impl(), last.impl(), ImplStr(s)); return *this; }
+ wxString& replace(iterator first, iterator last, const wchar_t* s)
+ { m_impl.replace(first.impl(), last.impl(), ImplStr(s)); return *this; }
+ wxString& replace(iterator first, iterator last, const char* s, size_type n)
+ {
+ SubstrBufFromMB str(ImplStr(s, n));
+ m_impl.replace(first.impl(), last.impl(), str.data, str.len);
+ return *this;
+ }
+ wxString& replace(iterator first, iterator last, const wchar_t* s, size_type n)
+ {
+ SubstrBufFromWC str(ImplStr(s, n));
+ m_impl.replace(first.impl(), last.impl(), str.data, str.len);
+ return *this;
+ }
+ wxString& replace(iterator first, iterator last, const wxString& s)
+ { m_impl.replace(first.impl(), last.impl(), s.m_impl); return *this; }
+ wxString& replace(iterator first, iterator last, size_type n, wxUniChar ch)
+ {
+#if wxUSE_UNICODE_UTF8
+ if ( !ch.IsAscii() )
+ m_impl.replace(first.impl(), last.impl(),
+ wxStringOperations::EncodeNChars(n, ch));
+ else
+#endif
+ m_impl.replace(first.impl(), last.impl(), n, (wxStringCharType)ch);
+ return *this;
+ }
+ wxString& replace(iterator first, iterator last,
+ const_iterator first1, const_iterator last1)
+ {
+ m_impl.replace(first.impl(), last.impl(), first1.impl(), last1.impl());
+ return *this;
+ }
+ wxString& replace(iterator first, iterator last,
+ const char *first1, const char *last1)
+ { replace(first, last, first1, last1 - first1); return *this; }
+ wxString& replace(iterator first, iterator last,
+ const wchar_t *first1, const wchar_t *last1)
+ { replace(first, last, first1, last1 - first1); return *this; }
+
+ // swap two strings
+ void swap(wxString& str)
+ { m_impl.swap(str.m_impl); }
+
+ // find a substring
+ size_t find(const wxString& str, size_t nStart = 0) const
+ { return PosFromImpl(m_impl.find(str.m_impl, PosToImpl(nStart))); }
+
+ // find first n characters of sz
+ size_t find(const char* sz, size_t nStart = 0, size_t n = npos) const
+ {
+ SubstrBufFromMB str(ImplStr(sz, n));
+ return PosFromImpl(m_impl.find(str.data, PosToImpl(nStart), str.len));
+ }
+ size_t find(const wchar_t* sz, size_t nStart = 0, size_t n = npos) const
+ {
+ SubstrBufFromWC str(ImplStr(sz, n));
+ return PosFromImpl(m_impl.find(str.data, PosToImpl(nStart), str.len));
+ }
+ size_t find(const wxCharBuffer& s, size_t nStart = 0, size_t n = npos) const
+ { return find(s.data(), nStart, n); }
+ size_t find(const wxWCharBuffer& s, size_t nStart = 0, size_t n = npos) const
+ { return find(s.data(), nStart, n); }
+ size_t find(const wxCStrData& s, size_t nStart = 0, size_t n = npos) const
+ { return find(s.AsWChar(), nStart, n); }
+
+ // find the first occurence of character ch after nStart
+ size_t find(wxUniChar ch, size_t nStart = 0) const
+ {
+ return PosFromImpl(m_impl.find(wxStringOperations::EncodeChar(ch),
+ PosToImpl(nStart)));
+ }
+ size_t find(wxUniCharRef ch, size_t nStart = 0) const
+ { return find(wxUniChar(ch), nStart); }
+ size_t find(char ch, size_t nStart = 0) const
+ { return find(wxUniChar(ch), nStart); }
+ size_t find(unsigned char ch, size_t nStart = 0) const
+ { return find(wxUniChar(ch), nStart); }
+ size_t find(wchar_t ch, size_t nStart = 0) const
+ { return find(wxUniChar(ch), nStart); }
+
+ // rfind() family is exactly like find() but works right to left
- // delete characters from nStart to nStart + nLen
- wxString& erase(size_type pos = 0, size_type n = npos)
- { return (wxString&)wxStringBase::erase(pos, n); }
- iterator erase(iterator first, iterator last)
- { return wxStringBase::erase(first, last); }
- iterator erase(iterator first)
- { return wxStringBase::erase(first); }
+ // as find, but from the end
+ size_t rfind(const wxString& str, size_t nStart = npos) const
+ { return PosFromImpl(m_impl.rfind(str.m_impl, PosToImpl(nStart))); }
-#ifdef wxSTRING_BASE_HASNT_CLEAR
- void clear() { erase(); }
+ // as find, but from the end
+ size_t rfind(const char* sz, size_t nStart = npos, size_t n = npos) const
+ {
+ SubstrBufFromMB str(ImplStr(sz, n));
+ return PosFromImpl(m_impl.rfind(str.data, PosToImpl(nStart), str.len));
+ }
+ size_t rfind(const wchar_t* sz, size_t nStart = npos, size_t n = npos) const
+ {
+ SubstrBufFromWC str(ImplStr(sz, n));
+ return PosFromImpl(m_impl.rfind(str.data, PosToImpl(nStart), str.len));
+ }
+ size_t rfind(const wxCharBuffer& s, size_t nStart = npos, size_t n = npos) const
+ { return rfind(s.data(), nStart, n); }
+ size_t rfind(const wxWCharBuffer& s, size_t nStart = npos, size_t n = npos) const
+ { return rfind(s.data(), nStart, n); }
+ size_t rfind(const wxCStrData& s, size_t nStart = npos, size_t n = npos) const
+ { return rfind(s.AsWChar(), nStart, n); }
+ // as find, but from the end
+ size_t rfind(wxUniChar ch, size_t nStart = npos) const
+ {
+ return PosFromImpl(m_impl.rfind(wxStringOperations::EncodeChar(ch),
+ PosToImpl(nStart)));
+ }
+ size_t rfind(wxUniCharRef ch, size_t nStart = npos) const
+ { return rfind(wxUniChar(ch), nStart); }
+ size_t rfind(char ch, size_t nStart = npos) const
+ { return rfind(wxUniChar(ch), nStart); }
+ size_t rfind(unsigned char ch, size_t nStart = npos) const
+ { return rfind(wxUniChar(ch), nStart); }
+ size_t rfind(wchar_t ch, size_t nStart = npos) const
+ { return rfind(wxUniChar(ch), nStart); }
+
+ // find first/last occurence of any character (not) in the set:
+#if wxUSE_STL_BASED_WXSTRING && !wxUSE_UNICODE_UTF8
+ // FIXME-UTF8: this is not entirely correct, because it doesn't work if
+ // sizeof(wchar_t)==2 and surrogates are present in the string;
+ // should we care? Probably not.
+ size_t find_first_of(const wxString& str, size_t nStart = 0) const
+ { return m_impl.find_first_of(str.m_impl, nStart); }
+ size_t find_first_of(const char* sz, size_t nStart = 0) const
+ { return m_impl.find_first_of(ImplStr(sz), nStart); }
+ size_t find_first_of(const wchar_t* sz, size_t nStart = 0) const
+ { return m_impl.find_first_of(ImplStr(sz), nStart); }
+ size_t find_first_of(const char* sz, size_t nStart, size_t n) const
+ { return m_impl.find_first_of(ImplStr(sz), nStart, n); }
+ size_t find_first_of(const wchar_t* sz, size_t nStart, size_t n) const
+ { return m_impl.find_first_of(ImplStr(sz), nStart, n); }
+ size_t find_first_of(wxUniChar c, size_t nStart = 0) const
+ { return m_impl.find_first_of((wxChar)c, nStart); }
+
+ size_t find_last_of(const wxString& str, size_t nStart = npos) const
+ { return m_impl.find_last_of(str.m_impl, nStart); }
+ size_t find_last_of(const char* sz, size_t nStart = npos) const
+ { return m_impl.find_last_of(ImplStr(sz), nStart); }
+ size_t find_last_of(const wchar_t* sz, size_t nStart = npos) const
+ { return m_impl.find_last_of(ImplStr(sz), nStart); }
+ size_t find_last_of(const char* sz, size_t nStart, size_t n) const
+ { return m_impl.find_last_of(ImplStr(sz), nStart, n); }
+ size_t find_last_of(const wchar_t* sz, size_t nStart, size_t n) const
+ { return m_impl.find_last_of(ImplStr(sz), nStart, n); }
+ size_t find_last_of(wxUniChar c, size_t nStart = npos) const
+ { return m_impl.find_last_of((wxChar)c, nStart); }
+
+ size_t find_first_not_of(const wxString& str, size_t nStart = 0) const
+ { return m_impl.find_first_not_of(str.m_impl, nStart); }
+ size_t find_first_not_of(const char* sz, size_t nStart = 0) const
+ { return m_impl.find_first_not_of(ImplStr(sz), nStart); }
+ size_t find_first_not_of(const wchar_t* sz, size_t nStart = 0) const
+ { return m_impl.find_first_not_of(ImplStr(sz), nStart); }
+ size_t find_first_not_of(const char* sz, size_t nStart, size_t n) const
+ { return m_impl.find_first_not_of(ImplStr(sz), nStart, n); }
+ size_t find_first_not_of(const wchar_t* sz, size_t nStart, size_t n) const
+ { return m_impl.find_first_not_of(ImplStr(sz), nStart, n); }
+ size_t find_first_not_of(wxUniChar c, size_t nStart = 0) const
+ { return m_impl.find_first_not_of((wxChar)c, nStart); }
+
+ size_t find_last_not_of(const wxString& str, size_t nStart = npos) const
+ { return m_impl.find_last_not_of(str.m_impl, nStart); }
+ size_t find_last_not_of(const char* sz, size_t nStart = npos) const
+ { return m_impl.find_last_not_of(ImplStr(sz), nStart); }
+ size_t find_last_not_of(const wchar_t* sz, size_t nStart = npos) const
+ { return m_impl.find_last_not_of(ImplStr(sz), nStart); }
+ size_t find_last_not_of(const char* sz, size_t nStart, size_t n) const
+ { return m_impl.find_last_not_of(ImplStr(sz), nStart, n); }
+ size_t find_last_not_of(const wchar_t* sz, size_t nStart, size_t n) const
+ { return m_impl.find_last_not_of(ImplStr(sz), nStart, n); }
+ size_t find_last_not_of(wxUniChar c, size_t nStart = npos) const
+ { return m_impl.find_last_not_of((wxChar)c, nStart); }
+#else
+ // we can't use std::string implementation in UTF-8 build, because the
+ // character sets would be interpreted wrongly:
+
+ // as strpbrk() but starts at nStart, returns npos if not found
+ size_t find_first_of(const wxString& str, size_t nStart = 0) const
+#if wxUSE_UNICODE // FIXME-UTF8: temporary
+ { return find_first_of(str.wc_str(), nStart); }
+#else
+ { return find_first_of(str.mb_str(), nStart); }
+#endif
+ // same as above
+ size_t find_first_of(const char* sz, size_t nStart = 0) const;
+ size_t find_first_of(const wchar_t* sz, size_t nStart = 0) const;
+ size_t find_first_of(const char* sz, size_t nStart, size_t n) const;
+ size_t find_first_of(const wchar_t* sz, size_t nStart, size_t n) const;
+ // same as find(char, size_t)
+ size_t find_first_of(wxUniChar c, size_t nStart = 0) const
+ { return find(c, nStart); }
+ // find the last (starting from nStart) char from str in this string
+ size_t find_last_of (const wxString& str, size_t nStart = npos) const
+#if wxUSE_UNICODE // FIXME-UTF8: temporary
+ { return find_last_of(str.wc_str(), nStart); }
+#else
+ { return find_last_of(str.mb_str(), nStart); }
#endif
+ // same as above
+ size_t find_last_of (const char* sz, size_t nStart = npos) const;
+ size_t find_last_of (const wchar_t* sz, size_t nStart = npos) const;
+ size_t find_last_of(const char* sz, size_t nStart, size_t n) const;
+ size_t find_last_of(const wchar_t* sz, size_t nStart, size_t n) const;
+ // same as above
+ size_t find_last_of(wxUniChar c, size_t nStart = npos) const
+ { return rfind(c, nStart); }
- // replaces the substring of length nLen starting at nStart
- wxString& replace(size_t nStart, size_t nLen, const wxChar* sz)
- { return (wxString&)wxStringBase::replace(nStart, nLen, sz); }
- // replaces the substring of length nLen starting at nStart
- wxString& replace(size_t nStart, size_t nLen, const wxString& str)
- { return (wxString&)wxStringBase::replace(nStart, nLen, str); }
- // replaces the substring with nCount copies of ch
- wxString& replace(size_t nStart, size_t nLen, size_t nCount, wxUniChar ch)
- { return (wxString&)wxStringBase::replace(nStart, nLen, nCount, ch); }
- // replaces a substring with another substring
- wxString& replace(size_t nStart, size_t nLen,
- const wxString& str, size_t nStart2, size_t nLen2)
- { return (wxString&)wxStringBase::replace(nStart, nLen, str,
- nStart2, nLen2); }
- // replaces the substring with first nCount chars of sz
- wxString& replace(size_t nStart, size_t nLen,
- const wxChar* sz, size_t nCount)
- { return (wxString&)wxStringBase::replace(nStart, nLen, sz, nCount); }
- wxString& replace(iterator first, iterator last, const_pointer s)
- { return (wxString&)wxStringBase::replace(first, last, s); }
- wxString& replace(iterator first, iterator last, const_pointer s,
- size_type n)
- { return (wxString&)wxStringBase::replace(first, last, s, n); }
- wxString& replace(iterator first, iterator last, const wxString& s)
- { return (wxString&)wxStringBase::replace(first, last, s); }
- wxString& replace(iterator first, iterator last, size_type n, wxUniChar c)
- { return (wxString&)wxStringBase::replace(first, last, n, c); }
- wxString& replace(iterator first, iterator last,
- const_iterator first1, const_iterator last1)
- { return (wxString&)wxStringBase::replace(first, last, first1, last1); }
+ // find first/last occurence of any character not in the set
+
+ // as strspn() (starting from nStart), returns npos on failure
+ size_t find_first_not_of(const wxString& str, size_t nStart = 0) const
+#if wxUSE_UNICODE // FIXME-UTF8: temporary
+ { return find_first_not_of(str.wc_str(), nStart); }
+#else
+ { return find_first_not_of(str.mb_str(), nStart); }
+#endif
+ // same as above
+ size_t find_first_not_of(const char* sz, size_t nStart = 0) const;
+ size_t find_first_not_of(const wchar_t* sz, size_t nStart = 0) const;
+ size_t find_first_not_of(const char* sz, size_t nStart, size_t n) const;
+ size_t find_first_not_of(const wchar_t* sz, size_t nStart, size_t n) const;
+ // same as above
+ size_t find_first_not_of(wxUniChar ch, size_t nStart = 0) const;
+ // as strcspn()
+ size_t find_last_not_of(const wxString& str, size_t nStart = npos) const
+#if wxUSE_UNICODE // FIXME-UTF8: temporary
+ { return find_last_not_of(str.wc_str(), nStart); }
+#else
+ { return find_last_not_of(str.mb_str(), nStart); }
+#endif
+ // same as above
+ size_t find_last_not_of(const char* sz, size_t nStart = npos) const;
+ size_t find_last_not_of(const wchar_t* sz, size_t nStart = npos) const;
+ size_t find_last_not_of(const char* sz, size_t nStart, size_t n) const;
+ size_t find_last_not_of(const wchar_t* sz, size_t nStart, size_t n) const;
+ // same as above
+ size_t find_last_not_of(wxUniChar ch, size_t nStart = npos) const;
+#endif // wxUSE_STL_BASED_WXSTRING && !wxUSE_UNICODE_UTF8 or not
+
+ // provide char/wchar_t/wxUniCharRef overloads for char-finding functions
+ // above to resolve ambiguities:
+ size_t find_first_of(wxUniCharRef ch, size_t nStart = 0) const
+ { return find_first_of(wxUniChar(ch), nStart); }
+ size_t find_first_of(char ch, size_t nStart = 0) const
+ { return find_first_of(wxUniChar(ch), nStart); }
+ size_t find_first_of(unsigned char ch, size_t nStart = 0) const
+ { return find_first_of(wxUniChar(ch), nStart); }
+ size_t find_first_of(wchar_t ch, size_t nStart = 0) const
+ { return find_first_of(wxUniChar(ch), nStart); }
+ size_t find_last_of(wxUniCharRef ch, size_t nStart = npos) const
+ { return find_last_of(wxUniChar(ch), nStart); }
+ size_t find_last_of(char ch, size_t nStart = npos) const
+ { return find_last_of(wxUniChar(ch), nStart); }
+ size_t find_last_of(unsigned char ch, size_t nStart = npos) const
+ { return find_last_of(wxUniChar(ch), nStart); }
+ size_t find_last_of(wchar_t ch, size_t nStart = npos) const
+ { return find_last_of(wxUniChar(ch), nStart); }
+ size_t find_first_not_of(wxUniCharRef ch, size_t nStart = 0) const
+ { return find_first_not_of(wxUniChar(ch), nStart); }
+ size_t find_first_not_of(char ch, size_t nStart = 0) const
+ { return find_first_not_of(wxUniChar(ch), nStart); }
+ size_t find_first_not_of(unsigned char ch, size_t nStart = 0) const
+ { return find_first_not_of(wxUniChar(ch), nStart); }
+ size_t find_first_not_of(wchar_t ch, size_t nStart = 0) const
+ { return find_first_not_of(wxUniChar(ch), nStart); }
+ size_t find_last_not_of(wxUniCharRef ch, size_t nStart = npos) const
+ { return find_last_not_of(wxUniChar(ch), nStart); }
+ size_t find_last_not_of(char ch, size_t nStart = npos) const
+ { return find_last_not_of(wxUniChar(ch), nStart); }
+ size_t find_last_not_of(unsigned char ch, size_t nStart = npos) const
+ { return find_last_not_of(wxUniChar(ch), nStart); }
+ size_t find_last_not_of(wchar_t ch, size_t nStart = npos) const
+ { return find_last_not_of(wxUniChar(ch), nStart); }
+
+ // and additional overloads for the versions taking strings:
+ size_t find_first_of(const wxCStrData& sz, size_t nStart = 0) const
+ { return find_first_of(sz.AsString(), nStart); }
+ size_t find_first_of(const wxCharBuffer& sz, size_t nStart = 0) const
+ { return find_first_of(sz.data(), nStart); }
+ size_t find_first_of(const wxWCharBuffer& sz, size_t nStart = 0) const
+ { return find_first_of(sz.data(), nStart); }
+ size_t find_first_of(const wxCStrData& sz, size_t nStart, size_t n) const
+ { return find_first_of(sz.AsWChar(), nStart, n); }
+ size_t find_first_of(const wxCharBuffer& sz, size_t nStart, size_t n) const
+ { return find_first_of(sz.data(), nStart, n); }
+ size_t find_first_of(const wxWCharBuffer& sz, size_t nStart, size_t n) const
+ { return find_first_of(sz.data(), nStart, n); }
+
+ size_t find_last_of(const wxCStrData& sz, size_t nStart = 0) const
+ { return find_last_of(sz.AsString(), nStart); }
+ size_t find_last_of(const wxCharBuffer& sz, size_t nStart = 0) const
+ { return find_last_of(sz.data(), nStart); }
+ size_t find_last_of(const wxWCharBuffer& sz, size_t nStart = 0) const
+ { return find_last_of(sz.data(), nStart); }
+ size_t find_last_of(const wxCStrData& sz, size_t nStart, size_t n) const
+ { return find_last_of(sz.AsWChar(), nStart, n); }
+ size_t find_last_of(const wxCharBuffer& sz, size_t nStart, size_t n) const
+ { return find_last_of(sz.data(), nStart, n); }
+ size_t find_last_of(const wxWCharBuffer& sz, size_t nStart, size_t n) const
+ { return find_last_of(sz.data(), nStart, n); }
+
+ size_t find_first_not_of(const wxCStrData& sz, size_t nStart = 0) const
+ { return find_first_not_of(sz.AsString(), nStart); }
+ size_t find_first_not_of(const wxCharBuffer& sz, size_t nStart = 0) const
+ { return find_first_not_of(sz.data(), nStart); }
+ size_t find_first_not_of(const wxWCharBuffer& sz, size_t nStart = 0) const
+ { return find_first_not_of(sz.data(), nStart); }
+ size_t find_first_not_of(const wxCStrData& sz, size_t nStart, size_t n) const
+ { return find_first_not_of(sz.AsWChar(), nStart, n); }
+ size_t find_first_not_of(const wxCharBuffer& sz, size_t nStart, size_t n) const
+ { return find_first_not_of(sz.data(), nStart, n); }
+ size_t find_first_not_of(const wxWCharBuffer& sz, size_t nStart, size_t n) const
+ { return find_first_not_of(sz.data(), nStart, n); }
+
+ size_t find_last_not_of(const wxCStrData& sz, size_t nStart = 0) const
+ { return find_last_not_of(sz.AsString(), nStart); }
+ size_t find_last_not_of(const wxCharBuffer& sz, size_t nStart = 0) const
+ { return find_last_not_of(sz.data(), nStart); }
+ size_t find_last_not_of(const wxWCharBuffer& sz, size_t nStart = 0) const
+ { return find_last_not_of(sz.data(), nStart); }
+ size_t find_last_not_of(const wxCStrData& sz, size_t nStart, size_t n) const
+ { return find_last_not_of(sz.AsWChar(), nStart, n); }
+ size_t find_last_not_of(const wxCharBuffer& sz, size_t nStart, size_t n) const
+ { return find_last_not_of(sz.data(), nStart, n); }
+ size_t find_last_not_of(const wxWCharBuffer& sz, size_t nStart, size_t n) const
+ { return find_last_not_of(sz.data(), nStart, n); }
// string += string
wxString& operator+=(const wxString& s)
- { return (wxString&)wxStringBase::operator+=(s); }
+ { m_impl += s.m_impl; return *this; }
// string += C string
- wxString& operator+=(const wxChar *psz)
- { return (wxString&)wxStringBase::operator+=(psz); }
+ wxString& operator+=(const char *psz)
+ { m_impl += ImplStr(psz); return *this; }
+ wxString& operator+=(const wchar_t *pwz)
+ { m_impl += ImplStr(pwz); return *this; }
wxString& operator+=(const wxCStrData& s)
- { return (wxString&)wxStringBase::operator+=(s.AsString()); }
+ { m_impl += s.AsString().m_impl; return *this; }
+ wxString& operator+=(const wxCharBuffer& s)
+ { return operator+=(s.data()); }
+ wxString& operator+=(const wxWCharBuffer& s)
+ { return operator+=(s.data()); }
// string += char
wxString& operator+=(wxUniChar ch)
- { return (wxString&)wxStringBase::operator+=(ch); }
+ { m_impl += wxStringOperations::EncodeChar(ch); return *this; }
wxString& operator+=(wxUniCharRef ch) { return *this += wxUniChar(ch); }
+ wxString& operator+=(int ch) { return *this += wxUniChar(ch); }
wxString& operator+=(char ch) { return *this += wxUniChar(ch); }
+ wxString& operator+=(unsigned char ch) { return *this += wxUniChar(ch); }
wxString& operator+=(wchar_t ch) { return *this += wxUniChar(ch); }
private:
#if !wxUSE_STL_BASED_WXSTRING
// helpers for wxStringBuffer and wxStringBufferLength
- wxChar *DoGetWriteBuf(size_t nLen);
- void DoUngetWriteBuf();
- void DoUngetWriteBuf(size_t nLen);
+ wxStringCharType *DoGetWriteBuf(size_t nLen)
+ { return m_impl.DoGetWriteBuf(nLen); }
+ void DoUngetWriteBuf()
+ { m_impl.DoUngetWriteBuf(); }
+ void DoUngetWriteBuf(size_t nLen)
+ { m_impl.DoUngetWriteBuf(nLen); }
+#endif // !wxUSE_STL_BASED_WXSTRING
- friend class WXDLLIMPEXP_BASE wxStringBuffer;
- friend class WXDLLIMPEXP_BASE wxStringBufferLength;
+#ifndef wxNEEDS_WXSTRING_PRINTF_MIXIN
+ #if !wxUSE_UTF8_LOCALE_ONLY
+ int DoPrintfWchar(const wxChar *format, ...);
+ static wxString DoFormatWchar(const wxChar *format, ...);
+ #endif
+ #if wxUSE_UNICODE_UTF8
+ int DoPrintfUtf8(const char *format, ...);
+ static wxString DoFormatUtf8(const char *format, ...);
+ #endif
#endif
-#ifndef wxNEEDS_WXSTRING_PRINTF_MIXIN
- int DoPrintf(const wxChar *format, ...) ATTRIBUTE_PRINTF_2;
- static wxString DoFormat(const wxChar *format, ...) ATTRIBUTE_PRINTF_1;
+#if !wxUSE_STL_BASED_WXSTRING
+ // check string's data validity
+ bool IsValid() const { return m_impl.GetStringData()->IsValid(); }
+#endif
+
+private:
+ wxStringImpl m_impl;
+
+#ifdef __VISUALC__
+ // "struct 'ConvertedBuffer<T>' needs to have dll-interface to be used by
+ // clients of class 'wxString'" - this is private, we don't care
+ #pragma warning (disable:4251)
+#endif
+
+ // buffers for compatibility conversion from (char*)c_str() and
+ // (wchar_t*)c_str():
+ // FIXME-UTF8: bechmark various approaches to keeping compatibility buffers
+ template<typename T>
+ struct ConvertedBuffer
+ {
+ ConvertedBuffer() : m_buf(NULL) {}
+ ~ConvertedBuffer()
+ { free(m_buf); }
+
+ operator T*() const { return m_buf; }
+
+ ConvertedBuffer& operator=(T *str)
+ {
+ free(m_buf);
+ m_buf = str;
+ return *this;
+ }
+
+ T *m_buf;
+ };
+#if wxUSE_UNICODE && !wxUSE_UTF8_LOCALE_ONLY
+ ConvertedBuffer<char> m_convertedToChar;
+#endif
+#if !wxUSE_UNICODE_WCHAR
+ ConvertedBuffer<wchar_t> m_convertedToWChar;
+#endif
+
+#ifdef __VISUALC__
+ #pragma warning (default:4251)
#endif
+
+#if wxUSE_UNICODE_UTF8
+ // FIXME-UTF8: (try to) move this elsewhere (TLS) or solve differently
+ // assigning to character pointer to by wxString::interator may
+ // change the underlying wxStringImpl iterator, so we have to
+ // keep track of all iterators and update them as necessary:
+ struct wxStringIteratorNodeHead
+ {
+ wxStringIteratorNodeHead() : ptr(NULL) {}
+ wxStringIteratorNode *ptr;
+
+ // copying is disallowed as it would result in more than one pointer into
+ // the same linked list
+ DECLARE_NO_COPY_CLASS(wxStringIteratorNodeHead)
+ };
+
+ wxStringIteratorNodeHead m_iterators;
+
+ friend class WXDLLIMPEXP_FWD_BASE wxStringIteratorNode;
+ friend class WXDLLIMPEXP_FWD_BASE wxUniCharRef;
+#endif // wxUSE_UNICODE_UTF8
+
+ friend class WXDLLIMPEXP_FWD_BASE wxCStrData;
+ friend class wxStringInternalBuffer;
+ friend class wxStringInternalBufferLength;
};
#ifdef wxNEEDS_WXSTRING_PRINTF_MIXIN
#pragma warning (default:4275)
#endif
+// string iterator operators that satisfy STL Random Access Iterator
+// requirements:
+inline wxString::iterator operator+(int n, wxString::iterator i)
+ { return i + n; }
+inline wxString::iterator operator+(size_t n, wxString::iterator i)
+ { return i + n; }
+inline wxString::const_iterator operator+(int n, wxString::const_iterator i)
+ { return i + n; }
+inline wxString::const_iterator operator+(size_t n, wxString::const_iterator i)
+ { return i + n; }
+inline wxString::reverse_iterator operator+(int n, wxString::reverse_iterator i)
+ { return i + n; }
+inline wxString::reverse_iterator operator+(size_t n, wxString::reverse_iterator i)
+ { return i + n; }
+inline wxString::const_reverse_iterator operator+(int n, wxString::const_reverse_iterator i)
+ { return i + n; }
+inline wxString::const_reverse_iterator operator+(size_t n, wxString::const_reverse_iterator i)
+ { return i + n; }
+
// notice that even though for many compilers the friend declarations above are
// enough, from the point of view of C++ standard we must have the declarations
// here as friend ones are not injected in the enclosing namespace and without
// them the code fails to compile with conforming compilers such as xlC or g++4
wxString WXDLLIMPEXP_BASE operator+(const wxString& string1, const wxString& string2);
-wxString WXDLLIMPEXP_BASE operator+(const wxString& string, const wxChar *psz);
-wxString WXDLLIMPEXP_BASE operator+(const wxChar *psz, const wxString& string);
+wxString WXDLLIMPEXP_BASE operator+(const wxString& string, const char *psz);
+wxString WXDLLIMPEXP_BASE operator+(const wxString& string, const wchar_t *pwz);
+wxString WXDLLIMPEXP_BASE operator+(const char *psz, const wxString& string);
+wxString WXDLLIMPEXP_BASE operator+(const wchar_t *pwz, const wxString& string);
wxString WXDLLIMPEXP_BASE operator+(const wxString& string, wxUniChar ch);
wxString WXDLLIMPEXP_BASE operator+(wxUniChar ch, const wxString& string);
{ return wxUniChar(ch) + string; }
-#if wxUSE_STL_BASED_WXSTRING
- // return an empty wxString (not very useful with wxUSE_STL == 1)
- inline const wxString wxGetEmptyString() { return wxString(); }
-#else // !wxUSE_STL_BASED_WXSTRING
- // return an empty wxString (more efficient than wxString() here)
- inline const wxString& wxGetEmptyString()
- {
- return *(wxString *)&wxEmptyString;
- }
-#endif // wxUSE_STL_BASED_WXSTRING/!wxUSE_STL_BASED_WXSTRING
+#define wxGetEmptyString() wxString()
// ----------------------------------------------------------------------------
// wxStringBuffer: a tiny class allowing to get a writable pointer into string
// ----------------------------------------------------------------------------
-#if wxUSE_STL_BASED_WXSTRING
-
-class WXDLLIMPEXP_BASE wxStringBuffer
+#if !wxUSE_STL_BASED_WXSTRING
+// string buffer for direct access to string data in their native
+// representation:
+class wxStringInternalBuffer
{
public:
- wxStringBuffer(wxString& str, size_t lenWanted = 1024)
- : m_str(str), m_buf(lenWanted)
- { }
+ typedef wxStringCharType CharType;
+
+ wxStringInternalBuffer(wxString& str, size_t lenWanted = 1024)
+ : m_str(str), m_buf(NULL)
+ { m_buf = m_str.DoGetWriteBuf(lenWanted); }
- ~wxStringBuffer() { m_str.assign(m_buf.data(), wxStrlen(m_buf.data())); }
+ ~wxStringInternalBuffer() { m_str.DoUngetWriteBuf(); }
- operator wxChar*() { return m_buf.data(); }
+ operator wxStringCharType*() const { return m_buf; }
private:
- wxString& m_str;
-#if wxUSE_UNICODE
- wxWCharBuffer m_buf;
-#else
- wxCharBuffer m_buf;
-#endif
+ wxString& m_str;
+ wxStringCharType *m_buf;
- DECLARE_NO_COPY_CLASS(wxStringBuffer)
+ DECLARE_NO_COPY_CLASS(wxStringInternalBuffer)
};
-class WXDLLIMPEXP_BASE wxStringBufferLength
+class wxStringInternalBufferLength
{
public:
- wxStringBufferLength(wxString& str, size_t lenWanted = 1024)
- : m_str(str), m_buf(lenWanted), m_len(0), m_lenSet(false)
- { }
+ typedef wxStringCharType CharType;
+
+ wxStringInternalBufferLength(wxString& str, size_t lenWanted = 1024)
+ : m_str(str), m_buf(NULL), m_len(0), m_lenSet(false)
+ {
+ m_buf = m_str.DoGetWriteBuf(lenWanted);
+ wxASSERT(m_buf != NULL);
+ }
- ~wxStringBufferLength()
+ ~wxStringInternalBufferLength()
{
wxASSERT(m_lenSet);
- m_str.assign(m_buf.data(), m_len);
+ m_str.DoUngetWriteBuf(m_len);
}
- operator wxChar*() { return m_buf.data(); }
+ operator wxStringCharType*() const { return m_buf; }
void SetLength(size_t length) { m_len = length; m_lenSet = true; }
private:
- wxString& m_str;
-#if wxUSE_UNICODE
- wxWCharBuffer m_buf;
-#else
- wxCharBuffer m_buf;
-#endif
- size_t m_len;
- bool m_lenSet;
+ wxString& m_str;
+ wxStringCharType *m_buf;
+ size_t m_len;
+ bool m_lenSet;
- DECLARE_NO_COPY_CLASS(wxStringBufferLength)
+ DECLARE_NO_COPY_CLASS(wxStringInternalBufferLength)
};
-#else // if !wxUSE_STL_BASED_WXSTRING
+#endif // !wxUSE_STL_BASED_WXSTRING
-class WXDLLIMPEXP_BASE wxStringBuffer
+template<typename T>
+class WXDLLIMPEXP_BASE wxStringTypeBufferBase
{
public:
- wxStringBuffer(wxString& str, size_t lenWanted = 1024)
- : m_str(str), m_buf(NULL)
- { m_buf = m_str.DoGetWriteBuf(lenWanted); }
+ typedef T CharType;
+
+ wxStringTypeBufferBase(wxString& str, size_t lenWanted = 1024)
+ : m_str(str), m_buf(lenWanted)
+ { }
- ~wxStringBuffer() { m_str.DoUngetWriteBuf(); }
- operator wxChar*() const { return m_buf; }
+ operator CharType*() { return m_buf.data(); }
-private:
+protected:
+ wxString& m_str;
+ wxCharTypeBuffer<CharType> m_buf;
+};
+
+template<typename T>
+class WXDLLIMPEXP_BASE wxStringTypeBufferLengthBase
+{
+public:
+ typedef T CharType;
+
+ wxStringTypeBufferLengthBase(wxString& str, size_t lenWanted = 1024)
+ : m_str(str), m_buf(lenWanted), m_len(0), m_lenSet(false)
+ { }
+
+ operator CharType*() { return m_buf.data(); }
+ void SetLength(size_t length) { m_len = length; m_lenSet = true; }
+
+protected:
wxString& m_str;
- wxChar *m_buf;
+ wxCharTypeBuffer<CharType> m_buf;
+ size_t m_len;
+ bool m_lenSet;
+};
+
+template<typename T>
+class wxStringTypeBuffer : public wxStringTypeBufferBase<T>
+{
+public:
+ wxStringTypeBuffer(wxString& str, size_t lenWanted = 1024)
+ : wxStringTypeBufferBase<T>(str, lenWanted) {}
+ ~wxStringTypeBuffer()
+ {
+ this->m_str.assign(this->m_buf.data());
+ }
- DECLARE_NO_COPY_CLASS(wxStringBuffer)
+ DECLARE_NO_COPY_CLASS(wxStringTypeBuffer)
};
-class WXDLLIMPEXP_BASE wxStringBufferLength
+template<typename T>
+class wxStringTypeBufferLength : public wxStringTypeBufferLengthBase<T>
{
public:
- wxStringBufferLength(wxString& str, size_t lenWanted = 1024)
- : m_str(str), m_buf(NULL), m_len(0), m_lenSet(false)
+ wxStringTypeBufferLength(wxString& str, size_t lenWanted = 1024)
+ : wxStringTypeBufferLengthBase<T>(str, lenWanted) {}
+
+ ~wxStringTypeBufferLength()
{
- m_buf = m_str.DoGetWriteBuf(lenWanted);
- wxASSERT(m_buf != NULL);
+ wxASSERT(this->m_lenSet);
+ this->m_str.assign(this->m_buf.data(), this->m_len);
}
- ~wxStringBufferLength()
+ DECLARE_NO_COPY_CLASS(wxStringTypeBufferLength)
+};
+
+#if wxUSE_STL_BASED_WXSTRING
+
+WXDLLIMPEXP_TEMPLATE_INSTANCE_BASE( wxStringTypeBufferBase<wxStringCharType> )
+
+class wxStringInternalBuffer : public wxStringTypeBufferBase<wxStringCharType>
+{
+public:
+ wxStringInternalBuffer(wxString& str, size_t lenWanted = 1024)
+ : wxStringTypeBufferBase<wxStringCharType>(str, lenWanted) {}
+ ~wxStringInternalBuffer()
+ { m_str.m_impl.assign(m_buf.data()); }
+
+ DECLARE_NO_COPY_CLASS(wxStringInternalBuffer)
+};
+
+WXDLLIMPEXP_TEMPLATE_INSTANCE_BASE(
+ wxStringTypeBufferLengthBase<wxStringCharType> )
+
+class wxStringInternalBufferLength
+ : public wxStringTypeBufferLengthBase<wxStringCharType>
+{
+public:
+ wxStringInternalBufferLength(wxString& str, size_t lenWanted = 1024)
+ : wxStringTypeBufferLengthBase<wxStringCharType>(str, lenWanted) {}
+
+ ~wxStringInternalBufferLength()
{
wxASSERT(m_lenSet);
- m_str.DoUngetWriteBuf(m_len);
+ m_str.m_impl.assign(m_buf.data(), m_len);
}
- operator wxChar*() const { return m_buf; }
- void SetLength(size_t length) { m_len = length; m_lenSet = true; }
+ DECLARE_NO_COPY_CLASS(wxStringInternalBufferLength)
+};
+#endif // wxUSE_STL_BASED_WXSTRING
-private:
- wxString& m_str;
- wxChar *m_buf;
- size_t m_len;
- bool m_lenSet;
- DECLARE_NO_COPY_CLASS(wxStringBufferLength)
+#if wxUSE_STL_BASED_WXSTRING || wxUSE_UNICODE_UTF8
+typedef wxStringTypeBuffer<wxChar> wxStringBuffer;
+typedef wxStringTypeBufferLength<wxChar> wxStringBufferLength;
+#else // if !wxUSE_STL_BASED_WXSTRING && !wxUSE_UNICODE_UTF8
+typedef wxStringInternalBuffer wxStringBuffer;
+typedef wxStringInternalBufferLength wxStringBufferLength;
+#endif // !wxUSE_STL_BASED_WXSTRING && !wxUSE_UNICODE_UTF8
+
+#if wxUSE_UNICODE_UTF8
+typedef wxStringInternalBuffer wxUTF8StringBuffer;
+typedef wxStringInternalBufferLength wxUTF8StringBufferLength;
+#elif wxUSE_UNICODE_WCHAR
+
+WXDLLIMPEXP_TEMPLATE_INSTANCE_BASE( wxStringTypeBufferBase<char> )
+
+class WXDLLIMPEXP_BASE wxUTF8StringBuffer : public wxStringTypeBufferBase<char>
+{
+public:
+ wxUTF8StringBuffer(wxString& str, size_t lenWanted = 1024)
+ : wxStringTypeBufferBase<char>(str, lenWanted) {}
+ ~wxUTF8StringBuffer();
+
+ DECLARE_NO_COPY_CLASS(wxUTF8StringBuffer)
};
-#endif // !wxUSE_STL_BASED_WXSTRING
+WXDLLIMPEXP_TEMPLATE_INSTANCE_BASE( wxStringTypeBufferLengthBase<char> )
+
+class WXDLLIMPEXP_BASE wxUTF8StringBufferLength
+ : public wxStringTypeBufferLengthBase<char>
+{
+public:
+ wxUTF8StringBufferLength(wxString& str, size_t lenWanted = 1024)
+ : wxStringTypeBufferLengthBase<char>(str, lenWanted) {}
+ ~wxUTF8StringBufferLength();
+
+ DECLARE_NO_COPY_CLASS(wxUTF8StringBufferLength)
+};
+#endif // wxUSE_UNICODE_UTF8/wxUSE_UNICODE_WCHAR
+
// ---------------------------------------------------------------------------
// wxString comparison functions: operator versions are always case sensitive
// ---------------------------------------------------------------------------
-// note that when wxUSE_STL_BASED_WXSTRING == 1 the comparison operators taking
-// std::string are used and defining them also for wxString would only result
-// in compilation ambiguities when comparing std::string and wxString
-#if !wxUSE_STL_BASED_WXSTRING
+#define wxCMP_WXCHAR_STRING(p, s, op) 0 op s.Cmp(p)
+wxDEFINE_ALL_COMPARISONS(const wxChar *, const wxString&, wxCMP_WXCHAR_STRING)
+
+#undef wxCMP_WXCHAR_STRING
+
+// note that there is an optimization in operator==() and !=(): we (quickly)
+// checks the strings length first, before comparing their data
inline bool operator==(const wxString& s1, const wxString& s2)
{ return (s1.Len() == s2.Len()) && (s1.Cmp(s2) == 0); }
-inline bool operator==(const wxString& s1, const wxChar * s2)
- { return s1.Cmp(s2) == 0; }
-inline bool operator==(const wxChar * s1, const wxString& s2)
- { return s2.Cmp(s1) == 0; }
inline bool operator!=(const wxString& s1, const wxString& s2)
{ return (s1.Len() != s2.Len()) || (s1.Cmp(s2) != 0); }
-inline bool operator!=(const wxString& s1, const wxChar * s2)
- { return s1.Cmp(s2) != 0; }
-inline bool operator!=(const wxChar * s1, const wxString& s2)
- { return s2.Cmp(s1) != 0; }
inline bool operator< (const wxString& s1, const wxString& s2)
{ return s1.Cmp(s2) < 0; }
-inline bool operator< (const wxString& s1, const wxChar * s2)
- { return s1.Cmp(s2) < 0; }
-inline bool operator< (const wxChar * s1, const wxString& s2)
- { return s2.Cmp(s1) > 0; }
inline bool operator> (const wxString& s1, const wxString& s2)
{ return s1.Cmp(s2) > 0; }
-inline bool operator> (const wxString& s1, const wxChar * s2)
- { return s1.Cmp(s2) > 0; }
-inline bool operator> (const wxChar * s1, const wxString& s2)
- { return s2.Cmp(s1) < 0; }
inline bool operator<=(const wxString& s1, const wxString& s2)
{ return s1.Cmp(s2) <= 0; }
-inline bool operator<=(const wxString& s1, const wxChar * s2)
- { return s1.Cmp(s2) <= 0; }
-inline bool operator<=(const wxChar * s1, const wxString& s2)
- { return s2.Cmp(s1) >= 0; }
inline bool operator>=(const wxString& s1, const wxString& s2)
{ return s1.Cmp(s2) >= 0; }
-inline bool operator>=(const wxString& s1, const wxChar * s2)
- { return s1.Cmp(s2) >= 0; }
-inline bool operator>=(const wxChar * s1, const wxString& s2)
- { return s2.Cmp(s1) <= 0; }
-#if wxUSE_UNICODE
+inline bool operator==(const wxString& s1, const wxCStrData& s2)
+ { return s1 == s2.AsString(); }
+inline bool operator==(const wxCStrData& s1, const wxString& s2)
+ { return s1.AsString() == s2; }
+inline bool operator!=(const wxString& s1, const wxCStrData& s2)
+ { return s1 != s2.AsString(); }
+inline bool operator!=(const wxCStrData& s1, const wxString& s2)
+ { return s1.AsString() != s2; }
+
inline bool operator==(const wxString& s1, const wxWCharBuffer& s2)
{ return (s1.Cmp((const wchar_t *)s2) == 0); }
inline bool operator==(const wxWCharBuffer& s1, const wxString& s2)
{ return (s1.Cmp((const wchar_t *)s2) != 0); }
inline bool operator!=(const wxWCharBuffer& s1, const wxString& s2)
{ return (s2.Cmp((const wchar_t *)s1) != 0); }
-#else // !wxUSE_UNICODE
+
inline bool operator==(const wxString& s1, const wxCharBuffer& s2)
{ return (s1.Cmp((const char *)s2) == 0); }
inline bool operator==(const wxCharBuffer& s1, const wxString& s2)
{ return (s1.Cmp((const char *)s2) != 0); }
inline bool operator!=(const wxCharBuffer& s1, const wxString& s2)
{ return (s2.Cmp((const char *)s1) != 0); }
-#endif // wxUSE_UNICODE/!wxUSE_UNICODE
-#if wxUSE_UNICODE
inline wxString operator+(const wxString& string, const wxWCharBuffer& buf)
{ return string + (const wchar_t *)buf; }
inline wxString operator+(const wxWCharBuffer& buf, const wxString& string)
{ return (const wchar_t *)buf + string; }
-#else // !wxUSE_UNICODE
+
inline wxString operator+(const wxString& string, const wxCharBuffer& buf)
{ return string + (const char *)buf; }
inline wxString operator+(const wxCharBuffer& buf, const wxString& string)
{ return (const char *)buf + string; }
-#endif // wxUSE_UNICODE/!wxUSE_UNICODE
-
-#endif // !wxUSE_STL_BASED_WXSTRING
-// comparison with char (those are not defined by std::[w]string and so should
-// be always available)
+// comparison with char
inline bool operator==(const wxUniChar& c, const wxString& s) { return s.IsSameAs(c); }
inline bool operator==(const wxUniCharRef& c, const wxString& s) { return s.IsSameAs(c); }
inline bool operator==(char c, const wxString& s) { return s.IsSameAs(c); }
// comparison with C string in Unicode build
#if wxUSE_UNICODE
-inline bool operator==(const wxString& s1, const char* s2)
- { return s1 == wxString(s2); }
-inline bool operator==(const char* s1, const wxString& s2)
- { return wxString(s1) == s2; }
-inline bool operator!=(const wxString& s1, const char* s2)
- { return s1 != wxString(s2); }
-inline bool operator!=(const char* s1, const wxString& s2)
- { return wxString(s1) != s2; }
-inline bool operator< (const wxString& s1, const char* s2)
- { return s1 < wxString(s2); }
-inline bool operator< (const char* s1, const wxString& s2)
- { return wxString(s1) < s2; }
-inline bool operator> (const wxString& s1, const char* s2)
- { return s1 > wxString(s2); }
-inline bool operator> (const char* s1, const wxString& s2)
- { return wxString(s1) > s2; }
-inline bool operator<=(const wxString& s1, const char* s2)
- { return s1 <= wxString(s2); }
-inline bool operator<=(const char* s1, const wxString& s2)
- { return wxString(s1) <= s2; }
-inline bool operator>=(const wxString& s1, const char* s2)
- { return s1 >= wxString(s2); }
-inline bool operator>=(const char* s1, const wxString& s2)
- { return wxString(s1) >= s2; }
+
+#define wxCMP_CHAR_STRING(p, s, op) wxString(p) op s
+
+wxDEFINE_ALL_COMPARISONS(const char *, const wxString&, wxCMP_CHAR_STRING)
+
+#undef wxCMP_CHAR_STRING
+
#endif // wxUSE_UNICODE
+// we also need to provide the operators for comparison with wxCStrData to
+// resolve ambiguity between operator(const wxChar *,const wxString &) and
+// operator(const wxChar *, const wxChar *) for "p == s.c_str()"
+//
+// notice that these are (shallow) pointer comparisons, not (deep) string ones
+#define wxCMP_CHAR_CSTRDATA(p, s, op) p op s.AsChar()
+#define wxCMP_WCHAR_CSTRDATA(p, s, op) p op s.AsWChar()
+
+wxDEFINE_ALL_COMPARISONS(const wchar_t *, const wxCStrData&, wxCMP_WCHAR_CSTRDATA)
+wxDEFINE_ALL_COMPARISONS(const char *, const wxCStrData&, wxCMP_CHAR_CSTRDATA)
+
+#undef wxCMP_CHAR_CSTRDATA
+#undef wxCMP_WCHAR_CSTRDATA
+
// ---------------------------------------------------------------------------
// Implementation only from here until the end of file
// ---------------------------------------------------------------------------
-// don't pollute the library user's name space
-#undef wxASSERT_VALID_INDEX
-
#if wxUSE_STD_IOSTREAM
#include "wx/iosfwrap.h"
WXDLLIMPEXP_BASE wxSTD ostream& operator<<(wxSTD ostream&, const wxString&);
WXDLLIMPEXP_BASE wxSTD ostream& operator<<(wxSTD ostream&, const wxCStrData&);
+WXDLLIMPEXP_BASE wxSTD ostream& operator<<(wxSTD ostream&, const wxCharBuffer&);
+#ifndef __BORLANDC__
+WXDLLIMPEXP_BASE wxSTD ostream& operator<<(wxSTD ostream&, const wxWCharBuffer&);
+#endif
+
+#if wxUSE_UNICODE && defined(HAVE_WOSTREAM)
+
+WXDLLIMPEXP_BASE wxSTD wostream& operator<<(wxSTD wostream&, const wxString&);
+WXDLLIMPEXP_BASE wxSTD wostream& operator<<(wxSTD wostream&, const wxCStrData&);
+WXDLLIMPEXP_BASE wxSTD wostream& operator<<(wxSTD wostream&, const wxWCharBuffer&);
-#endif // wxSTD_STRING_COMPATIBILITY
+#endif // wxUSE_UNICODE && defined(HAVE_WOSTREAM)
+
+#endif // wxUSE_STD_IOSTREAM
// ---------------------------------------------------------------------------
// wxCStrData implementation
inline wxCStrData::wxCStrData(wchar_t *buf)
: m_str(new wxString(buf)), m_offset(0), m_owned(true) {}
+inline wxCStrData::wxCStrData(const wxCStrData& data)
+ : m_str(data.m_owned ? new wxString(*data.m_str) : data.m_str),
+ m_offset(data.m_offset),
+ m_owned(data.m_owned)
+{
+}
+
inline wxCStrData::~wxCStrData()
{
if ( m_owned )
- delete m_str;
+ delete wx_const_cast(wxString*, m_str); // cast to silence warnings
}
-#if wxUSE_UNICODE
+// simple cases for AsChar() and AsWChar(), the complicated ones are
+// in string.cpp
+#if wxUSE_UNICODE_WCHAR
inline const wchar_t* wxCStrData::AsWChar() const
-#else
+{
+ return m_str->wx_str() + m_offset;
+}
+#endif // wxUSE_UNICODE_WCHAR
+
+#if !wxUSE_UNICODE
+inline const char* wxCStrData::AsChar() const
+{
+ return m_str->wx_str() + m_offset;
+}
+#endif // !wxUSE_UNICODE
+
+#if wxUSE_UTF8_LOCALE_ONLY
inline const char* wxCStrData::AsChar() const
+{
+ return wxStringOperations::AddToIter(m_str->wx_str(), m_offset);
+}
+#endif // wxUSE_UTF8_LOCALE_ONLY
+
+inline const wxCharBuffer wxCStrData::AsCharBuf() const
+{
+#if !wxUSE_UNICODE
+ return wxCharBuffer::CreateNonOwned(AsChar());
+#else
+ return AsString().mb_str();
#endif
+}
+
+inline const wxWCharBuffer wxCStrData::AsWCharBuf() const
{
- if ( m_offset == 0 )
- return m_str->wx_str(); // FIXME
- else
- return (const wxChar*)(m_str->begin() + m_offset);
+#if wxUSE_UNICODE_WCHAR
+ return wxWCharBuffer::CreateNonOwned(AsWChar());
+#else
+ return AsString().wc_str();
+#endif
}
inline wxString wxCStrData::AsString() const
return m_str->Mid(m_offset);
}
-inline wxCStrData::operator wxString() const { return AsString(); }
+inline const wxStringCharType *wxCStrData::AsInternal() const
+{
+#if wxUSE_UNICODE_UTF8
+ return wxStringOperations::AddToIter(m_str->wx_str(), m_offset);
+#else
+ return m_str->wx_str() + m_offset;
+#endif
+}
inline wxUniChar wxCStrData::operator*() const
{
inline wxUniChar wxCStrData::operator[](size_t n) const
{
- return m_str->at(m_offset + n);
+ // NB: we intentionally use operator[] and not at() here because the former
+ // works for the terminating NUL while the latter does not
+ return (*m_str)[m_offset + n];
}
// ----------------------------------------------------------------------------
-// implementation of wxString inline methods using wxCStrData
+// more wxCStrData operators
// ----------------------------------------------------------------------------
-inline wxString& wxString::operator=(const wxCStrData& cstr)
+// we need to define those to allow "size_t pos = p - s.c_str()" where p is
+// some pointer into the string
+inline size_t operator-(const char *p, const wxCStrData& cs)
{
- return *this = cstr.AsString();
+ return p - cs.AsChar();
+}
+
+inline size_t operator-(const wchar_t *p, const wxCStrData& cs)
+{
+ return p - cs.AsWChar();
}
// ----------------------------------------------------------------------------
// implementation of wx[W]CharBuffer inline methods using wxCStrData
// ----------------------------------------------------------------------------
-#if wxUSE_UNICODE
+// FIXME-UTF8: move this to buffer.h
+inline wxCharBuffer::wxCharBuffer(const wxCStrData& cstr)
+ : wxCharTypeBufferBase(cstr.AsCharBuf())
+{
+}
inline wxWCharBuffer::wxWCharBuffer(const wxCStrData& cstr)
- : m_str(wxStrdupW(cstr))
+ : wxCharTypeBufferBase(cstr.AsWCharBuf())
{
}
-#else // !wxUSE_UNICODE
+#if wxUSE_UNICODE_UTF8
+// ----------------------------------------------------------------------------
+// implementation of wxStringIteratorNode inline methods
+// ----------------------------------------------------------------------------
-inline wxCharBuffer::wxCharBuffer(const wxCStrData& cstr)
- : m_str(wxStrdupA(cstr))
+void wxStringIteratorNode::DoSet(const wxString *str,
+ wxStringImpl::const_iterator *citer,
+ wxStringImpl::iterator *iter)
+{
+ m_prev = NULL;
+ m_iter = iter;
+ m_citer = citer;
+ m_str = str;
+ if ( str )
+ {
+ m_next = str->m_iterators.ptr;
+ wx_const_cast(wxString*, m_str)->m_iterators.ptr = this;
+ if ( m_next )
+ m_next->m_prev = this;
+ }
+ else
+ {
+ m_next = NULL;
+ }
+}
+
+void wxStringIteratorNode::clear()
{
+ if ( m_next )
+ m_next->m_prev = m_prev;
+ if ( m_prev )
+ m_prev->m_next = m_next;
+ else if ( m_str ) // first in the list
+ wx_const_cast(wxString*, m_str)->m_iterators.ptr = m_next;
+
+ m_next = m_prev = NULL;
+ m_citer = NULL;
+ m_iter = NULL;
+ m_str = NULL;
}
+#endif // wxUSE_UNICODE_UTF8
-#endif // wxUSE_UNICODE/!wxUSE_UNICODE
+#if WXWIN_COMPATIBILITY_2_8
+ // lot of code out there doesn't explicitly include wx/crt.h, but uses
+ // CRT wrappers that are now declared in wx/wxcrt.h and wx/wxcrtvararg.h,
+ // so let's include this header now that wxString is defined and it's safe
+ // to do it:
+ #include "wx/crt.h"
+#endif
-#endif // _WX_WXSTRINGH__
+#endif // _WX_WXSTRING_H_