X-Git-Url: https://git.saurik.com/wxWidgets.git/blobdiff_plain/f416695a1bf8ac887e2b6223f8ad72954dc2dc03..99d82720737ebf91d3a387d53dd0cf6c8cf9a643:/include/wx/string.h diff --git a/include/wx/string.h b/include/wx/string.h index c5283522d5..d1c3b6d783 100644 --- a/include/wx/string.h +++ b/include/wx/string.h @@ -57,6 +57,36 @@ #include "wx/stringops.h" #include "wx/unichar.h" +// by default we cache the mapping of the positions in UTF-8 string to the byte +// offset as this results in noticeable performance improvements for loops over +// strings using indices; comment out this line to disable this +// +// notice that this optimization is well worth using even in debug builds as it +// changes asymptotic complexity of algorithms using indices to iterate over +// wxString back to expected linear from quadratic +// +// also notice that wxTLS_TYPE() (__declspec(thread) in this case) is unsafe to +// use in DLL build under pre-Vista Windows so we disable this code for now, if +// anybody really needs to use UTF-8 build under Windows with this optimization +// it would have to be re-tested and probably corrected +#if wxUSE_UNICODE_UTF8 && !defined(__WXMSW__) + #define wxUSE_STRING_POS_CACHE 1 +#else + #define wxUSE_STRING_POS_CACHE 0 +#endif + +#if wxUSE_STRING_POS_CACHE + #include "wx/tls.h" + + // change this 0 to 1 to enable additional (very expensive) asserts + // verifying that string caching logic works as expected + #if 0 + #define wxSTRING_CACHE_ASSERT(cond) wxASSERT(cond) + #else + #define wxSTRING_CACHE_ASSERT(cond) + #endif +#endif // wxUSE_STRING_POS_CACHE + class WXDLLIMPEXP_FWD_BASE wxString; // unless this symbol is predefined to disable the compatibility functions, do @@ -65,6 +95,11 @@ class WXDLLIMPEXP_FWD_BASE wxString; #define WXWIN_COMPATIBILITY_STRING_PTR_AS_ITER 1 #endif +namespace wxPrivate +{ + template struct wxStringAsBufHelper; +} + // --------------------------------------------------------------------------- // macros // --------------------------------------------------------------------------- @@ -503,6 +538,14 @@ private: static size_t LenToImpl(size_t len) { return len; } static size_t PosFromImpl(size_t pos) { return pos; } + // we don't want to define these as empty inline functions as it could + // result in noticeable (and quite unnecessary in non-UTF-8 build) slowdown + // in debug build where the inline functions are not effectively inlined + #define wxSTRING_INVALIDATE_CACHE() + #define wxSTRING_INVALIDATE_CACHED_LENGTH() + #define wxSTRING_UPDATE_CACHED_LENGTH(n) + #define wxSTRING_SET_CACHED_LENGTH(n) + #else // wxUSE_UNICODE_UTF8 static wxCharBuffer ImplStr(const char* str, @@ -517,12 +560,273 @@ private: static SubstrBufFromWC ImplStr(const wchar_t* str, size_t n) { return ConvertStr(str, n, wxMBConvUTF8()); } +#if wxUSE_STRING_POS_CACHE + // this is an extremely simple cache used by PosToImpl(): each cache element + // contains the string it applies to and the index corresponding to the last + // used position in this wxString in its m_impl string + // + // NB: notice that this struct (and nested Element one) must be a POD or we + // wouldn't be able to use a thread-local variable of this type, in + // particular it should have no ctor -- we rely on statics being + // initialized to 0 instead + struct Cache + { + enum { SIZE = 8 }; + + struct Element + { + const wxString *str; // the string to which this element applies + size_t pos, // the cached index in this string + impl, // the corresponding position in its m_impl + len; // cached length or npos if unknown + + // reset cached index to 0 + void ResetPos() { pos = impl = 0; } + + // reset position and length + void Reset() { ResetPos(); len = npos; } + }; + + // cache the indices mapping for the last few string used + Element cached[SIZE]; + + // the last used index + unsigned lastUsed; + }; + +#ifndef wxHAS_COMPILER_TLS + // we must use an accessor function and not a static variable when the TLS + // variables support is implemented in the library (and not by the compiler) + // because the global s_cache variable could be not yet initialized when a + // ctor of another global object is executed and if that ctor uses any + // wxString methods, bad things happen + // + // however notice that this approach does not work when compiler TLS is used, + // at least not with g++ 4.1.2 under amd64 as it apparently compiles code + // using this accessor incorrectly when optimizations are enabled (-O2 is + // enough) -- luckily we don't need it then neither as static __thread + // variables are initialized by 0 anyhow then and so we can use the variable + // directly + WXEXPORT static Cache& GetCache() + { + static wxTLS_TYPE(Cache) s_cache; + + return wxTLS_VALUE(s_cache); + } + + // this helper struct is used to ensure that GetCache() is called during + // static initialization time, i.e. before any threads creation, as otherwise + // the static s_cache construction inside GetCache() wouldn't be MT-safe + friend struct wxStrCacheInitializer; +#else // wxHAS_COMPILER_TLS + static wxTLS_TYPE(Cache) ms_cache; + static Cache& GetCache() { return wxTLS_VALUE(ms_cache); } +#endif // !wxHAS_COMPILER_TLS/wxHAS_COMPILER_TLS + + static Cache::Element *GetCacheBegin() { return GetCache().cached; } + static Cache::Element *GetCacheEnd() { return GetCacheBegin() + Cache::SIZE; } + static unsigned& LastUsedCacheElement() { return GetCache().lastUsed; } + + // this is used in debug builds only to provide a convenient function, + // callable from a debugger, to show the cache contents + friend struct wxStrCacheDumper; + + // uncomment this to have access to some profiling statistics on program + // termination + //#define wxPROFILE_STRING_CACHE + +#ifdef wxPROFILE_STRING_CACHE + static struct PosToImplCacheStats + { + unsigned postot, // total non-trivial calls to PosToImpl + poshits, // cache hits from PosToImpl() + mishits, // cached position beyond the needed one + sumpos, // sum of all positions, used to compute the + // average position after dividing by postot + sumofs, // sum of all offsets after using the cache, used to + // compute the average after dividing by hits + lentot, // number of total calls to length() + lenhits; // number of cache hits in length() + } ms_cacheStats; + + friend struct wxStrCacheStatsDumper; + + #define wxCACHE_PROFILE_FIELD_INC(field) ms_cacheStats.field++ + #define wxCACHE_PROFILE_FIELD_ADD(field, val) ms_cacheStats.field += (val) +#else // !wxPROFILE_STRING_CACHE + #define wxCACHE_PROFILE_FIELD_INC(field) + #define wxCACHE_PROFILE_FIELD_ADD(field, val) +#endif // wxPROFILE_STRING_CACHE/!wxPROFILE_STRING_CACHE + + // note: it could seem that the functions below shouldn't be inline because + // they are big, contain loops and so the compiler shouldn't be able to + // inline them anyhow, however moving them into string.cpp does decrease the + // code performance by ~5%, at least when using g++ 4.1 so do keep them here + // unless tests show that it's not advantageous any more + + // return the pointer to the cache element for this string or NULL if not + // cached + Cache::Element *FindCacheElement() const + { + // profiling seems to show a small but consistent gain if we use this + // simple loop instead of starting from the last used element (there are + // a lot of misses in this function...) + Cache::Element * const cacheBegin = GetCacheBegin(); +#ifndef wxHAS_COMPILER_TLS + // during destruction tls calls may return NULL, in this case return NULL + // immediately without accessing anything else + if ( cacheBegin == NULL ) + return NULL; +#endif + Cache::Element * const cacheEnd = GetCacheEnd(); + for ( Cache::Element *c = cacheBegin; c != cacheEnd; c++ ) + { + if ( c->str == this ) + return c; + } + + return NULL; + } + + // unlike FindCacheElement(), this one always returns a valid pointer to the + // cache element for this string, it may have valid last cached position and + // its corresponding index in the byte string or not + Cache::Element *GetCacheElement() const + { + Cache::Element * const cacheBegin = GetCacheBegin(); + Cache::Element * const cacheEnd = GetCacheEnd(); + Cache::Element * const cacheStart = cacheBegin + LastUsedCacheElement(); + + // check the last used first, this does no (measurable) harm for a miss + // but does help for simple loops addressing the same string all the time + if ( cacheStart->str == this ) + return cacheStart; + + // notice that we're going to check cacheStart again inside this call but + // profiling shows that it's still faster to use a simple loop like + // inside FindCacheElement() than manually looping with wrapping starting + // from the cache entry after the start one + Cache::Element *c = FindCacheElement(); + if ( !c ) + { + // claim the next cache entry for this string + c = cacheStart; + if ( ++c == cacheEnd ) + c = cacheBegin; + + c->str = this; + c->Reset(); + + // and remember the last used element + LastUsedCacheElement() = c - cacheBegin; + } + + return c; + } + + size_t DoPosToImpl(size_t pos) const + { + wxCACHE_PROFILE_FIELD_INC(postot); + + // NB: although the case of pos == 1 (and offset from cached position + // equal to 1) are common, nothing is gained by writing special code + // for handling them, the compiler (at least g++ 4.1 used) seems to + // optimize the code well enough on its own + + wxCACHE_PROFILE_FIELD_ADD(sumpos, pos); + + Cache::Element * const cache = GetCacheElement(); + + // cached position can't be 0 so if it is, it means that this entry was + // used for length caching only so far, i.e. it doesn't count as a hit + // from our point of view + if ( cache->pos ) + { + wxCACHE_PROFILE_FIELD_INC(poshits); + } + + if ( pos == cache->pos ) + return cache->impl; + + // this seems to happen only rarely so just reset the cache in this case + // instead of complicating code even further by seeking backwards in this + // case + if ( cache->pos > pos ) + { + wxCACHE_PROFILE_FIELD_INC(mishits); + + cache->ResetPos(); + } + + wxCACHE_PROFILE_FIELD_ADD(sumofs, pos - cache->pos); + + + wxStringImpl::const_iterator i(m_impl.begin() + cache->impl); + for ( size_t n = cache->pos; n < pos; n++ ) + wxStringOperations::IncIter(i); + + cache->pos = pos; + cache->impl = i - m_impl.begin(); + + wxSTRING_CACHE_ASSERT( + (int)cache->impl == (begin() + pos).impl() - m_impl.begin() ); + + return cache->impl; + } + + void InvalidateCache() + { + Cache::Element * const cache = FindCacheElement(); + if ( cache ) + cache->Reset(); + } + + void InvalidateCachedLength() + { + Cache::Element * const cache = FindCacheElement(); + if ( cache ) + cache->len = npos; + } + + void SetCachedLength(size_t len) + { + // we optimistically cache the length here even if the string wasn't + // present in the cache before, this seems to do no harm and the + // potential for avoiding length recomputation for long strings looks + // interesting + GetCacheElement()->len = len; + } + + void UpdateCachedLength(ptrdiff_t delta) + { + Cache::Element * const cache = FindCacheElement(); + if ( cache && cache->len != npos ) + { + wxSTRING_CACHE_ASSERT( (ptrdiff_t)cache->len + delta >= 0 ); + + cache->len += delta; + } + } + + #define wxSTRING_INVALIDATE_CACHE() InvalidateCache() + #define wxSTRING_INVALIDATE_CACHED_LENGTH() InvalidateCachedLength() + #define wxSTRING_UPDATE_CACHED_LENGTH(n) UpdateCachedLength(n) + #define wxSTRING_SET_CACHED_LENGTH(n) SetCachedLength(n) +#else // !wxUSE_STRING_POS_CACHE + size_t DoPosToImpl(size_t pos) const + { + return (begin() + pos).impl() - m_impl.begin(); + } + + #define wxSTRING_INVALIDATE_CACHE() + #define wxSTRING_INVALIDATE_CACHED_LENGTH() + #define wxSTRING_UPDATE_CACHED_LENGTH(n) + #define wxSTRING_SET_CACHED_LENGTH(n) +#endif // wxUSE_STRING_POS_CACHE/!wxUSE_STRING_POS_CACHE + size_t PosToImpl(size_t pos) const { - if ( pos == 0 || pos == npos ) - return pos; - else - return (begin() + pos).impl() - m_impl.begin(); + return pos == 0 || pos == npos ? pos : DoPosToImpl(pos); } void PosLenToImpl(size_t pos, size_t len, size_t *implPos, size_t *implLen) const; @@ -554,22 +858,26 @@ public: typedef size_t size_type; typedef wxUniChar const_reference; -#if wxUSE_STL +#if wxUSE_STD_STRING #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 + #define WX_DEFINE_ITERATOR_CATEGORY(cat) typedef cat iterator_category; #else - #define WX_STR_ITERATOR_TAG void /* dummy type */ + // not defining iterator_category at all in this case is better than defining + // it as some dummy type -- at least it results in more intelligible error + // messages + #define WX_DEFINE_ITERATOR_CATEGORY(cat) #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; \ + WX_DEFINE_ITERATOR_CATEGORY(WX_STR_ITERATOR_TAG) \ typedef wxUniChar value_type; \ typedef int difference_type; \ typedef reference_type reference; \ @@ -668,7 +976,7 @@ public: } reference operator*() - { return wxUniCharRef::CreateForString(m_node, m_cur); } + { return wxUniCharRef::CreateForString(*str(), m_cur); } iterator operator+(ptrdiff_t n) const { return iterator(str(), wxStringOperations::AddToIter(m_cur, n)); } @@ -679,7 +987,7 @@ public: 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); } + wxString* str() const { return const_cast(m_node.m_str); } wxStringIteratorNode m_node; @@ -732,6 +1040,10 @@ public: size_t IterToImplPos(wxString::iterator i) const { return wxStringImpl::const_iterator(i.impl()) - m_impl.begin(); } + iterator GetIterForNthChar(size_t n) + { return iterator(this, m_impl.begin() + PosToImpl(n)); } + const_iterator GetIterForNthChar(size_t n) const + { return const_iterator(this, m_impl.begin() + PosToImpl(n)); } #else // !wxUSE_UNICODE_UTF8 class WXDLLIMPEXP_BASE iterator @@ -783,6 +1095,9 @@ public: const_iterator(const wxString *WXUNUSED(str), underlying_iterator ptr) : m_cur(ptr) {} }; + + iterator GetIterForNthChar(size_t n) { return begin() + n; } + const_iterator GetIterForNthChar(size_t n) const { return begin() + n; } #endif // wxUSE_UNICODE_UTF8/!wxUSE_UNICODE_UTF8 #undef WX_STR_ITERATOR_TAG @@ -797,7 +1112,7 @@ public: public: typedef T iterator_type; - typedef typename T::iterator_category iterator_category; + WX_DEFINE_ITERATOR_CATEGORY(typename T::iterator_category) typedef typename T::value_type value_type; typedef typename T::difference_type difference_type; typedef typename T::reference reference; @@ -953,8 +1268,11 @@ public: wxString(const wxWCharBuffer& buf) { assign(buf.data()); } // FIXME-UTF8: fix for embedded NUL and buffer length + // NB: this version uses m_impl.c_str() to force making a copy of the + // string, so that "wxString(str.c_str())" idiom for passing strings + // between threads works wxString(const wxCStrData& cstr) - : m_impl(cstr.AsString().m_impl) { } + : m_impl(cstr.AsString().m_impl.c_str()) { } // 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 @@ -966,6 +1284,17 @@ public: wxString(const wxString& str, size_t nLength) { assign(str, nLength); } + +#if wxUSE_STRING_POS_CACHE + ~wxString() + { + // we need to invalidate our cache entry as another string could be + // recreated at the same address (unlikely, but still possible, with the + // heap-allocated strings but perfectly common with stack-allocated ones) + InvalidateCache(); + } +#endif // wxUSE_STRING_POS_CACHE + // even if we're not built with wxUSE_STL == 1 it is very convenient to allow // 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 @@ -1014,6 +1343,13 @@ public: #endif #endif // wxUSE_STL + wxString Clone() const + { + // make a deep copy of the string, i.e. the returned string will have + // ref count = 1 with refcounted implementation + return wxString::FromImpl(wxStringImpl(m_impl.c_str(), m_impl.length())); + } + // first valid index position const_iterator begin() const { return const_iterator(this, m_impl.begin()); } iterator begin() { return iterator(this, m_impl.begin()); } @@ -1034,7 +1370,32 @@ public: // std::string methods: #if wxUSE_UNICODE_UTF8 - size_t length() const { return end() - begin(); } // FIXME-UTF8: optimize! + size_t length() const + { +#if wxUSE_STRING_POS_CACHE + wxCACHE_PROFILE_FIELD_INC(lentot); + + Cache::Element * const cache = GetCacheElement(); + + if ( cache->len == npos ) + { + // it's probably not worth trying to be clever and using cache->pos + // here as it's probably 0 anyhow -- you usually call length() before + // starting to index the string + cache->len = end() - begin(); + } + else + { + wxCACHE_PROFILE_FIELD_INC(lenhits); + + wxSTRING_CACHE_ASSERT( (int)cache->len == end() - begin() ); + } + + return cache->len; +#else // !wxUSE_STRING_POS_CACHE + return end() - begin(); +#endif // wxUSE_STRING_POS_CACHE/!wxUSE_STRING_POS_CACHE + } #else size_t length() const { return m_impl.length(); } #endif @@ -1044,8 +1405,9 @@ public: bool empty() const { return m_impl.empty(); } - size_type capacity() const { return m_impl.capacity(); } // FIXME-UTF8 - void reserve(size_t sz) { m_impl.reserve(sz); } // FIXME-UTF8 + // NB: these methods don't have a well-defined meaning in UTF-8 case + size_type capacity() const { return m_impl.capacity(); } + void reserve(size_t sz) { m_impl.reserve(sz); } void resize(size_t nSize, wxUniChar ch = wxT('\0')) { @@ -1056,6 +1418,8 @@ public: #if wxUSE_UNICODE_UTF8 if ( nSize < len ) { + wxSTRING_INVALIDATE_CACHE(); + // we can't use wxStringImpl::resize() for truncating the string as it // counts in bytes, not characters erase(nSize); @@ -1065,10 +1429,16 @@ public: // 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 + } + else // can use (presumably faster) resize() version #endif // wxUSE_UNICODE_UTF8 + { + wxSTRING_INVALIDATE_CACHED_LENGTH(); + m_impl.resize(nSize, (wxStringCharType)ch); + } } wxString substr(size_t nStart = 0, size_t nLen = npos) const @@ -1088,18 +1458,9 @@ public: // truncate the string to given length wxString& Truncate(size_t uiLen); // empty string contents - void Empty() - { - Truncate(0); - - wxASSERT_MSG( empty(), _T("string not empty after call to Empty()?") ); - } + void Empty() { clear(); } // empty the string and free memory - void Clear() - { - wxString tmp(wxEmptyString); - swap(tmp); - } + void Clear() { clear(); } // contents test // Is an ascii value @@ -1112,16 +1473,16 @@ public: // data access (all indexes are 0 based) // read access wxUniChar at(size_t n) const - { return *(begin() + n); } // FIXME-UTF8: optimize? + { return wxStringOperations::DecodeChar(m_impl.begin() + PosToImpl(n)); } wxUniChar GetChar(size_t n) const { return at(n); } // read/write access wxUniCharRef at(size_t n) - { return *(begin() + n); } // FIXME-UTF8: optimize? + { return *GetIterForNthChar(n); } wxUniCharRef GetWritableChar(size_t n) { return at(n); } // write access - void SetChar(size_t n, wxUniChar ch) + void SetChar(size_t n, wxUniChar ch) { at(n) = ch; } // get last character @@ -1165,11 +1526,53 @@ public: { return at(n); } #endif // size_t != unsigned int - // explicit conversion to C string (use this with printf()!) + + /* + Overview of wxString conversions, implicit and explicit: + + - wxString has a std::[w]string-like c_str() method, however it does + not return a C-style string directly but instead returns wxCStrData + helper object which is convertible to either "char *" narrow string + or "wchar_t *" wide string. Usually the correct conversion will be + applied by the compiler automatically but if this doesn't happen you + need to explicitly choose one using wxCStrData::AsChar() or AsWChar() + methods or another wxString conversion function. + + - One of the places where the conversion does *NOT* happen correctly is + when c_str() is passed to a vararg function such as printf() so you + must *NOT* use c_str() with them. Either use wxPrintf() (all wx + functions do handle c_str() correctly, even if they appear to be + vararg (but they're not, really)) or add an explicit AsChar() or, if + compatibility with previous wxWidgets versions is important, add a + cast to "const char *". + + - In non-STL mode only, wxString is also implicitly convertible to + wxCStrData. The same warning as above applies. + + - c_str() is polymorphic as it can be converted to either narrow or + wide string. If you explicitly need one or the other, choose to use + mb_str() (for narrow) or wc_str() (for wide) instead. Notice that + these functions can return either the pointer to string directly (if + this is what the string uses internally) or a temporary buffer + containing the string and convertible to it. Again, conversion will + usually be done automatically by the compiler but beware of the + vararg functions: you need an explicit cast when using them. + + - There are also non-const versions of mb_str() and wc_str() called + char_str() and wchar_str(). They are only meant to be used with + non-const-correct functions and they always return buffers. + + - Finally wx_str() returns whatever string representation is used by + wxString internally. It may be either a narrow or wide string + depending on wxWidgets build mode but it will always be a raw pointer + (and not a buffer). + */ + + // explicit conversion to wxCStrData wxCStrData c_str() const { return wxCStrData(this); } wxCStrData data() const { return c_str(); } - // implicit conversion to C string + // implicit conversion to wxCStrData operator wxCStrData() const { return c_str(); } // the first two operators conflict with operators for conversion to @@ -1201,6 +1604,30 @@ public: { return mb_str(conv); } wxWritableWCharBuffer wchar_str() const { return wc_str(); } + // conversion to the buffer of the given type T (= char or wchar_t) and + // also optionally return the buffer length + // + // this is mostly/only useful for the template functions + // + // FIXME-VC6: the second argument only exists for VC6 which doesn't support + // explicit template function selection, do not use it unless + // you must support VC6! + template + wxCharTypeBuffer tchar_str(size_t *len = NULL, + T * WXUNUSED(dummy) = NULL) const + { +#if wxUSE_UNICODE + // we need a helper dispatcher depending on type + return wxPrivate::wxStringAsBufHelper::Get(*this, len); +#else // ANSI + // T can only be char in ANSI build + if ( len ) + *len = length(); + + return wxCharTypeBuffer::CreateNonOwned(wx_str()); +#endif // Unicode build kind + } + // conversion to/from plain (i.e. 7 bit) ASCII: this is useful for // converting numbers or strings which are certain not to contain special // chars (typically system functions, X atoms, environment variables etc.) @@ -1229,7 +1656,7 @@ public: // conversion to/from UTF-8: #if wxUSE_UNICODE_UTF8 - static wxString FromUTF8(const char *utf8) + static wxString FromUTF8Unchecked(const char *utf8) { if ( !utf8 ) return wxEmptyString; @@ -1237,23 +1664,51 @@ public: wxASSERT( wxStringOperations::IsValidUtf8String(utf8) ); return FromImpl(wxStringImpl(utf8)); } - static wxString FromUTF8(const char *utf8, size_t len) + static wxString FromUTF8Unchecked(const char *utf8, size_t len) { if ( !utf8 ) return wxEmptyString; if ( len == npos ) - return FromUTF8(utf8); + return FromUTF8Unchecked(utf8); wxASSERT( wxStringOperations::IsValidUtf8String(utf8, len) ); return FromImpl(wxStringImpl(utf8, len)); } + + static wxString FromUTF8(const char *utf8) + { + if ( !utf8 || !wxStringOperations::IsValidUtf8String(utf8) ) + return ""; + + return FromImpl(wxStringImpl(utf8)); + } + static wxString FromUTF8(const char *utf8, size_t len) + { + if ( len == npos ) + return FromUTF8(utf8); + + if ( !utf8 || !wxStringOperations::IsValidUtf8String(utf8, len) ) + return ""; + + return FromImpl(wxStringImpl(utf8, len)); + } + const char* utf8_str() const { return wx_str(); } const char* ToUTF8() const { return wx_str(); } + + // this function exists in UTF-8 build only and returns the length of the + // internal UTF-8 representation + size_t utf8_length() const { return m_impl.length(); } #elif wxUSE_UNICODE_WCHAR - static wxString FromUTF8(const char *utf8) - { return wxString(utf8, wxMBConvUTF8()); } - static wxString FromUTF8(const char *utf8, size_t len) + static wxString FromUTF8(const char *utf8, size_t len = npos) { return wxString(utf8, wxMBConvUTF8(), len); } + static wxString FromUTF8Unchecked(const char *utf8, size_t len = npos) + { + const wxString s(utf8, wxMBConvUTF8(), len); + wxASSERT_MSG( !utf8 || !*utf8 || !s.empty(), + "string must be valid UTF-8" ); + return s; + } const wxCharBuffer utf8_str() const { return mb_str(wxMBConvUTF8()); } const wxCharBuffer ToUTF8() const { return utf8_str(); } #else // ANSI @@ -1261,9 +1716,20 @@ public: { 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); + size_t wlen; + wxWCharBuffer buf(wxMBConvUTF8().cMB2WC(utf8, len == npos ? wxNO_LEN : len, &wlen)); + return wxString(buf.data(), wlen); + } + static wxString FromUTF8Unchecked(const char *utf8, size_t len = npos) + { + size_t wlen; + wxWCharBuffer buf(wxMBConvUTF8().cMB2WC(utf8, + len == npos ? wxNO_LEN : len, + &wlen)); + wxASSERT_MSG( !utf8 || !*utf8 || wlen, + "string must be valid UTF-8" ); + + return wxString(buf.data(), wlen); } const wxCharBuffer utf8_str() const { return wxMBConvUTF8().cWC2MB(wc_str()); } @@ -1308,7 +1774,7 @@ public: const wxWX2MBbuf mbc_str() const { return mb_str(*wxConvCurrent); } #if wxUSE_UNICODE_WCHAR - const wxChar* wc_str() const { return wx_str(); } + const wchar_t* wc_str() const { return wx_str(); } #elif wxUSE_UNICODE_UTF8 const wxWCharBuffer wc_str() const; #endif @@ -1326,7 +1792,7 @@ public: const wxChar* mb_str() const { return wx_str(); } // for compatibility with wxUSE_UNICODE version - const wxChar* mb_str(const wxMBConv& WXUNUSED(conv)) const { return wx_str(); } + const char* mb_str(const wxMBConv& WXUNUSED(conv)) const { return wx_str(); } const wxWX2MBbuf mbc_str() const { return mb_str(); } @@ -1336,23 +1802,45 @@ public: const wxCharBuffer fn_str() const { return wxConvFile.cWC2WX( wc_str( wxConvLibc ) ); } #endif // Unicode/ANSI +#if wxUSE_UNICODE_UTF8 + const wxWCharBuffer t_str() const { return wc_str(); } +#elif wxUSE_UNICODE_WCHAR + const wchar_t* t_str() const { return wx_str(); } +#else + const char* t_str() const { return wx_str(); } +#endif + + // overloaded assignment // from another wxString wxString& operator=(const wxString& stringSrc) - { if (&stringSrc != this) m_impl = stringSrc.m_impl; return *this; } + { + if ( this != &stringSrc ) + { + wxSTRING_INVALIDATE_CACHE(); + + m_impl = stringSrc.m_impl; + } + + return *this; + } + wxString& operator=(const wxCStrData& cstr) { return *this = cstr.AsString(); } // from a character wxString& operator=(wxUniChar ch) { + wxSTRING_INVALIDATE_CACHE(); + #if wxUSE_UNICODE_UTF8 if ( !ch.IsAscii() ) m_impl = wxStringOperations::EncodeChar(ch); else -#endif +#endif // wxUSE_UNICODE_UTF8 m_impl = (wxStringCharType)ch; return *this; } + wxString& operator=(wxUniCharRef ch) { return operator=((wxUniChar)ch); } wxString& operator=(char ch) @@ -1365,15 +1853,48 @@ public: // so we need to compensate in that case #if wxUSE_STL_BASED_WXSTRING wxString& operator=(const char *psz) - { if (psz) m_impl = ImplStr(psz); else Clear(); return *this; } + { + wxSTRING_INVALIDATE_CACHE(); + + 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_INVALIDATE_CACHE(); + + if ( pwz ) + m_impl = ImplStr(pwz); + else + clear(); + + return *this; + } +#else // !wxUSE_STL_BASED_WXSTRING wxString& operator=(const char *psz) - { m_impl = ImplStr(psz); return *this; } + { + wxSTRING_INVALIDATE_CACHE(); + + m_impl = ImplStr(psz); + + return *this; + } + wxString& operator=(const wchar_t *pwz) - { m_impl = ImplStr(pwz); return *this; } -#endif + { + wxSTRING_INVALIDATE_CACHE(); + + m_impl = ImplStr(pwz); + + return *this; + } +#endif // wxUSE_STL_BASED_WXSTRING/!wxUSE_STL_BASED_WXSTRING + wxString& operator=(const unsigned char *psz) { return operator=((const char*)psz); } @@ -1513,7 +2034,7 @@ public: const wxChar *fmt = _T("%") wxLongLongFmtSpec _T("u"); return (*this) << Format(fmt , ull); } -#endif +#endif // wxLongLong_t && !wxLongLongIsLong // insert a float into string wxString& operator<<(float f) { return (*this) << Format(_T("%f"), f); } @@ -1596,16 +2117,16 @@ public: wxString Left(size_t nCount) const; // get last nCount characters wxString Right(size_t nCount) const; - // get all characters before the first occurance of ch + // get all characters before the first occurrence of ch // (returns the whole string if ch not found) wxString BeforeFirst(wxUniChar ch) const; - // get all characters before the last occurence of ch + // get all characters before the last occurrence of ch // (returns empty string if ch not found) wxString BeforeLast(wxUniChar ch) const; - // get all characters after the first occurence of ch + // get all characters after the first occurrence of ch // (returns empty string if ch not found) wxString AfterFirst(wxUniChar ch) const; - // get all characters after the last occurence of ch + // get all characters after the last occurrence of ch // (returns the whole string if ch not found) wxString AfterLast(wxUniChar ch) const; @@ -1617,12 +2138,17 @@ public: // convert to upper case in place, return the string itself wxString& MakeUpper(); // convert to upper case, return the copy of the string - // Here's something to remember: BC++ doesn't like returns in inlines. - wxString Upper() const ; + wxString Upper() const { return wxString(*this).MakeUpper(); } // convert to lower case in place, return the string itself wxString& MakeLower(); // convert to lower case, return the copy of the string - wxString Lower() const ; + wxString Lower() const { return wxString(*this).MakeLower(); } + // convert the first character to the upper case and the rest to the + // lower one, return the modified string itself + wxString& MakeCapitalized(); + // convert the first character to the upper case and the rest to the + // lower one, return the copy of the string + wxString Capitalize() const { return wxString(*this).MakeCapitalized(); } // trimming/padding whitespace (either side) and truncating // remove spaces from left or from right (default) side @@ -1665,7 +2191,7 @@ public: int Find(const wxWCharBuffer& sub) const { return Find(sub.data()); } - // replace first (or all of bReplaceAll) occurences of substring with + // replace first (or all of bReplaceAll) occurrences of substring with // another string, returns the number of replacements made size_t Replace(const wxString& strOld, const wxString& strNew, @@ -1783,7 +2309,7 @@ public: #endif // wxNEEDS_WXSTRING_PRINTF_MIXIN // use Cmp() - inline int CompareTo(const wxChar* psz, caseCompare cmp = exact) const + int CompareTo(const wxChar* psz, caseCompare cmp = exact) const { return cmp == exact ? Cmp(psz) : CmpNoCase(psz); } // use length() @@ -1855,30 +2381,54 @@ public: // append elements str[pos], ..., str[pos+n] wxString& append(const wxString& str, size_t pos, size_t n) { - size_t from, len; - str.PosLenToImpl(pos, n, &from, &len); - m_impl.append(str.m_impl, from, len); - return *this; + wxSTRING_UPDATE_CACHED_LENGTH(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) - { m_impl.append(str.m_impl); return *this; } + { + wxSTRING_UPDATE_CACHED_LENGTH(str.length()); + + m_impl.append(str.m_impl); + return *this; + } + // append first n (or all if n == npos) characters of sz wxString& append(const char *sz) - { m_impl.append(ImplStr(sz)); return *this; } + { + wxSTRING_INVALIDATE_CACHED_LENGTH(); + + m_impl.append(ImplStr(sz)); + return *this; + } + wxString& append(const wchar_t *sz) - { m_impl.append(ImplStr(sz)); return *this; } + { + wxSTRING_INVALIDATE_CACHED_LENGTH(); + + 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_INVALIDATE_CACHED_LENGTH(); + + 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_UPDATE_CACHED_LENGTH(n); + + SubstrBufFromWC str(ImplStr(sz, n)); + m_impl.append(str.data, str.len); + return *this; } wxString& append(const wxCStrData& str) @@ -1898,13 +2448,23 @@ public: wxString& append(size_t n, wxUniChar ch) { #if wxUSE_UNICODE_UTF8 - if ( !ch.IsAscii() ) - m_impl.append(wxStringOperations::EncodeNChars(n, ch)); - else + if ( !ch.IsAscii() ) + { + wxSTRING_INVALIDATE_CACHED_LENGTH(); + + m_impl.append(wxStringOperations::EncodeNChars(n, ch)); + } + else // ASCII #endif - m_impl.append(n, (wxStringCharType)ch); - return *this; + { + wxSTRING_UPDATE_CACHED_LENGTH(n); + + m_impl.append(n, (wxStringCharType)ch); + } + + return *this; } + wxString& append(size_t n, wxUniCharRef ch) { return append(n, wxUniChar(ch)); } wxString& append(size_t n, char ch) @@ -1916,7 +2476,12 @@ public: // append from first to last wxString& append(const_iterator first, const_iterator last) - { m_impl.append(first.impl(), last.impl()); return *this; } + { + wxSTRING_INVALIDATE_CACHED_LENGTH(); + + 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); } @@ -1928,36 +2493,74 @@ public: // same as `this_string = str' wxString& assign(const wxString& str) - { m_impl = str.m_impl; return *this; } + { + wxSTRING_SET_CACHED_LENGTH(str.length()); + + 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; + wxSTRING_SET_CACHED_LENGTH(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) { - size_t from, len; - str.PosLenToImpl(pos, n, &from, &len); - m_impl.assign(str.m_impl, from, len); - return *this; + size_t from, len; + str.PosLenToImpl(pos, n, &from, &len); + m_impl.assign(str.m_impl, from, len); + + // it's important to call this after PosLenToImpl() above in case str is + // the same string as this one + wxSTRING_SET_CACHED_LENGTH(n); + + return *this; } + // same as `= first n (or all if n == npos) characters of sz' wxString& assign(const char *sz) - { m_impl.assign(ImplStr(sz)); return *this; } + { + wxSTRING_INVALIDATE_CACHE(); + + m_impl.assign(ImplStr(sz)); + + return *this; + } + wxString& assign(const wchar_t *sz) - { m_impl.assign(ImplStr(sz)); return *this; } + { + wxSTRING_INVALIDATE_CACHE(); + + 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_SET_CACHED_LENGTH(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_SET_CACHED_LENGTH(n); + + SubstrBufFromWC str(ImplStr(sz, n)); + m_impl.assign(str.data, str.len); + + return *this; } wxString& assign(const wxCStrData& str) @@ -1976,13 +2579,16 @@ public: // same as `= n copies of ch' wxString& assign(size_t n, wxUniChar ch) { + wxSTRING_SET_CACHED_LENGTH(n); + #if wxUSE_UNICODE_UTF8 - if ( !ch.IsAscii() ) - m_impl.assign(wxStringOperations::EncodeNChars(n, ch)); - else + if ( !ch.IsAscii() ) + m_impl.assign(wxStringOperations::EncodeNChars(n, ch)); + else #endif - m_impl.assign(n, (wxStringCharType)ch); - return *this; + m_impl.assign(n, (wxStringCharType)ch); + + return *this; } wxString& assign(size_t n, wxUniCharRef ch) @@ -1996,7 +2602,13 @@ public: // assign from first to last wxString& assign(const_iterator first, const_iterator last) - { m_impl.assign(first.impl(), last.impl()); return *this; } + { + wxSTRING_INVALIDATE_CACHE(); + + 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); } @@ -2029,58 +2641,93 @@ public: // insert another string wxString& insert(size_t nPos, const wxString& str) - { insert(begin() + nPos, str.begin(), str.end()); return *this; } + { insert(GetIterForNthChar(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) { - size_t from, len; - str.PosLenToImpl(nStart, n, &from, &len); - m_impl.insert(PosToImpl(nPos), str.m_impl, from, len); - return *this; + wxSTRING_UPDATE_CACHED_LENGTH(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 char *sz) - { m_impl.insert(PosToImpl(nPos), ImplStr(sz)); return *this; } + { + wxSTRING_INVALIDATE_CACHE(); + + 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_INVALIDATE_CACHE(); + + 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_UPDATE_CACHED_LENGTH(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; + wxSTRING_UPDATE_CACHED_LENGTH(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) { + wxSTRING_UPDATE_CACHED_LENGTH(n); + #if wxUSE_UNICODE_UTF8 - if ( !ch.IsAscii() ) - m_impl.insert(PosToImpl(nPos), wxStringOperations::EncodeNChars(n, ch)); - else + if ( !ch.IsAscii() ) + m_impl.insert(PosToImpl(nPos), wxStringOperations::EncodeNChars(n, ch)); + else #endif - m_impl.insert(PosToImpl(nPos), n, (wxStringCharType)ch); - return *this; + m_impl.insert(PosToImpl(nPos), n, (wxStringCharType)ch); + return *this; } + iterator insert(iterator it, wxUniChar ch) { + wxSTRING_UPDATE_CACHED_LENGTH(1); + #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 + 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)); + return iterator(this, m_impl.insert(it.impl(), (wxStringCharType)ch)); } + void insert(iterator it, const_iterator first, const_iterator last) - { m_impl.insert(it.impl(), first.impl(), last.impl()); } + { + wxSTRING_INVALIDATE_CACHE(); + + 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); } @@ -2092,150 +2739,238 @@ public: void insert(iterator it, size_type n, wxUniChar ch) { + wxSTRING_UPDATE_CACHED_LENGTH(n); + #if wxUSE_UNICODE_UTF8 - if ( !ch.IsAscii() ) - m_impl.insert(IterToImplPos(it), wxStringOperations::EncodeNChars(n, ch)); - else + if ( !ch.IsAscii() ) + m_impl.insert(IterToImplPos(it), wxStringOperations::EncodeNChars(n, ch)); + else #endif - m_impl.insert(it.impl(), n, (wxStringCharType)ch); + 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; + wxSTRING_INVALIDATE_CACHE(); + + 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())); } + { + wxSTRING_INVALIDATE_CACHE(); + + return iterator(this, m_impl.erase(first.impl(), last.impl())); + } + iterator erase(iterator first) - { return iterator(this, m_impl.erase(first.impl())); } + { + wxSTRING_UPDATE_CACHED_LENGTH(-1); + + return iterator(this, m_impl.erase(first.impl())); + } #ifdef wxSTRING_BASE_HASNT_CLEAR void clear() { erase(); } #else - void clear() { m_impl.clear(); } + void clear() + { + wxSTRING_SET_CACHED_LENGTH(0); + + 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_INVALIDATE_CACHE(); + + 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; + wxSTRING_INVALIDATE_CACHE(); + + 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; + wxSTRING_INVALIDATE_CACHE(); + + 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); + wxSTRING_INVALIDATE_CACHE(); + + 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 + if ( !ch.IsAscii() ) + m_impl.replace(from, len, wxStringOperations::EncodeNChars(nCount, ch)); + else #endif - m_impl.replace(from, len, nCount, (wxStringCharType)ch); - return *this; + 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); + wxSTRING_INVALIDATE_CACHE(); - size_t from2, len2; - str.PosLenToImpl(nStart2, nLen2, &from2, &len2); + size_t from, len; + PosLenToImpl(nStart, nLen, &from, &len); - m_impl.replace(from, len, str.m_impl, from2, len2); - return *this; + 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); + wxSTRING_INVALIDATE_CACHE(); - SubstrBufFromMB str(ImplStr(sz, nCount)); + size_t from, len; + PosLenToImpl(nStart, nLen, &from, &len); - m_impl.replace(from, len, str.data, str.len); - return *this; + 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); + wxSTRING_INVALIDATE_CACHE(); - SubstrBufFromWC str(ImplStr(sz, nCount)); + size_t from, len; + PosLenToImpl(nStart, nLen, &from, &len); - m_impl.replace(from, len, str.data, str.len); - return *this; + 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_INVALIDATE_CACHE(); + + 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_INVALIDATE_CACHE(); + + 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_INVALIDATE_CACHE(); + + 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_INVALIDATE_CACHE(); + + 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_INVALIDATE_CACHE(); + + 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_INVALIDATE_CACHE(); + + m_impl.replace(first.impl(), last.impl(), s.m_impl); + + return *this; + } + wxString& replace(iterator first, iterator last, size_type n, wxUniChar ch) { + wxSTRING_INVALIDATE_CACHE(); + #if wxUSE_UNICODE_UTF8 - if ( !ch.IsAscii() ) - m_impl.replace(first.impl(), last.impl(), - wxStringOperations::EncodeNChars(n, ch)); - else + 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; + 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_INVALIDATE_CACHE(); + + 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; } @@ -2245,7 +2980,17 @@ public: // swap two strings void swap(wxString& str) - { m_impl.swap(str.m_impl); } + { +#if wxUSE_STRING_POS_CACHE + // we modify not only this string but also the other one directly so we + // need to invalidate cache for both of them (we could also try to + // exchange their cache entries but it seems unlikely to be worth it) + InvalidateCache(); + str.InvalidateCache(); +#endif // wxUSE_STRING_POS_CACHE + + m_impl.swap(str.m_impl); + } // find a substring size_t find(const wxString& str, size_t nStart = 0) const @@ -2269,7 +3014,7 @@ public: 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 + // find the first occurrence of character ch after nStart size_t find(wxUniChar ch, size_t nStart = 0) const { #if wxUSE_UNICODE_UTF8 @@ -2335,7 +3080,7 @@ public: 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: + // find first/last occurrence 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; @@ -2426,7 +3171,7 @@ public: 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 + // find first/last occurrence 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 @@ -2548,14 +3293,34 @@ public: // string += string wxString& operator+=(const wxString& s) - { m_impl += s.m_impl; return *this; } + { + wxSTRING_INVALIDATE_CACHED_LENGTH(); + + m_impl += s.m_impl; + return *this; + } // string += C string wxString& operator+=(const char *psz) - { m_impl += ImplStr(psz); return *this; } + { + wxSTRING_INVALIDATE_CACHED_LENGTH(); + + m_impl += ImplStr(psz); + return *this; + } wxString& operator+=(const wchar_t *pwz) - { m_impl += ImplStr(pwz); return *this; } + { + wxSTRING_INVALIDATE_CACHED_LENGTH(); + + m_impl += ImplStr(pwz); + return *this; + } wxString& operator+=(const wxCStrData& s) - { m_impl += s.AsString().m_impl; return *this; } + { + wxSTRING_INVALIDATE_CACHED_LENGTH(); + + m_impl += s.AsString().m_impl; + return *this; + } wxString& operator+=(const wxCharBuffer& s) { return operator+=(s.data()); } wxString& operator+=(const wxWCharBuffer& s) @@ -2563,13 +3328,15 @@ public: // string += char wxString& operator+=(wxUniChar ch) { + wxSTRING_UPDATE_CACHED_LENGTH(1); + #if wxUSE_UNICODE_UTF8 - if ( !ch.IsAscii() ) - m_impl += wxStringOperations::EncodeChar(ch); - else + if ( !ch.IsAscii() ) + m_impl += wxStringOperations::EncodeChar(ch); + else #endif - m_impl += (wxStringCharType)ch; - return *this; + m_impl += (wxStringCharType)ch; + return *this; } wxString& operator+=(wxUniCharRef ch) { return *this += wxUniChar(ch); } wxString& operator+=(int ch) { return *this += wxUniChar(ch); } @@ -2581,11 +3348,23 @@ private: #if !wxUSE_STL_BASED_WXSTRING // helpers for wxStringBuffer and wxStringBufferLength wxStringCharType *DoGetWriteBuf(size_t nLen) - { return m_impl.DoGetWriteBuf(nLen); } + { + return m_impl.DoGetWriteBuf(nLen); + } + void DoUngetWriteBuf() - { m_impl.DoUngetWriteBuf(); } + { + wxSTRING_INVALIDATE_CACHE(); + + m_impl.DoUngetWriteBuf(); + } + void DoUngetWriteBuf(size_t nLen) - { m_impl.DoUngetWriteBuf(nLen); } + { + wxSTRING_SET_CACHED_LENGTH(nLen); + + m_impl.DoUngetWriteBuf(nLen); + } #endif // !wxUSE_STL_BASED_WXSTRING #ifndef wxNEEDS_WXSTRING_PRINTF_MIXIN @@ -2705,6 +3484,67 @@ inline wxString operator+(wchar_t ch, const wxString& string) #define wxGetEmptyString() wxString() +// ---------------------------------------------------------------------------- +// helper functions which couldn't be defined inline +// ---------------------------------------------------------------------------- + +namespace wxPrivate +{ + +#if wxUSE_UNICODE_WCHAR + +template <> +struct wxStringAsBufHelper +{ + static wxCharBuffer Get(const wxString& s, size_t *len) + { + wxCharBuffer buf(s.mb_str()); + if ( len ) + *len = buf ? strlen(buf) : 0; + return buf; + } +}; + +template <> +struct wxStringAsBufHelper +{ + static wxWCharBuffer Get(const wxString& s, size_t *len) + { + if ( len ) + *len = s.length(); + return wxWCharBuffer::CreateNonOwned(s.wx_str()); + } +}; + +#elif wxUSE_UNICODE_UTF8 + +template <> +struct wxStringAsBufHelper +{ + static wxCharBuffer Get(const wxString& s, size_t *len) + { + if ( len ) + *len = s.utf8_length(); + return wxCharBuffer::CreateNonOwned(s.wx_str()); + } +}; + +template <> +struct wxStringAsBufHelper +{ + static wxWCharBuffer Get(const wxString& s, size_t *len) + { + wxWCharBuffer wbuf(s.wc_str()); + if ( len ) + *len = wxWcslen(wbuf); + return wbuf; + } +}; + +#endif // Unicode build kind + +} // namespace wxPrivate + // ---------------------------------------------------------------------------- // wxStringBuffer: a tiny class allowing to get a writable pointer into string // ---------------------------------------------------------------------------- @@ -2765,15 +3605,37 @@ private: #endif // !wxUSE_STL_BASED_WXSTRING template -class WXDLLIMPEXP_BASE wxStringTypeBufferBase +class wxStringTypeBufferBase { public: typedef T CharType; wxStringTypeBufferBase(wxString& str, size_t lenWanted = 1024) : m_str(str), m_buf(lenWanted) - { } - + { + // for compatibility with old wxStringBuffer which provided direct + // access to wxString internal buffer, initialize ourselves with the + // string initial contents + + // FIXME-VC6: remove the ugly (CharType *)NULL and use normal + // tchar_str + size_t len; + const wxCharTypeBuffer buf(str.tchar_str(&len, (CharType *)NULL)); + if ( buf ) + { + if ( len > lenWanted ) + { + // in this case there is not enough space for terminating NUL, + // ensure that we still put it there + m_buf.data()[lenWanted] = 0; + len = lenWanted - 1; + } + + memcpy(m_buf.data(), buf, (len + 1)*sizeof(CharType)); + } + //else: conversion failed, this can happen when trying to get Unicode + // string contents into a char string + } operator CharType*() { return m_buf.data(); } @@ -2783,23 +3645,25 @@ protected: }; template -class WXDLLIMPEXP_BASE wxStringTypeBufferLengthBase +class wxStringTypeBufferLengthBase : public wxStringTypeBufferBase { public: - typedef T CharType; - wxStringTypeBufferLengthBase(wxString& str, size_t lenWanted = 1024) - : m_str(str), m_buf(lenWanted), m_len(0), m_lenSet(false) + : wxStringTypeBufferBase(str, lenWanted), + m_len(0), + m_lenSet(false) { } - operator CharType*() { return m_buf.data(); } + ~wxStringTypeBufferLengthBase() + { + wxASSERT_MSG( this->m_lenSet, "forgot to call SetLength()" ); + } + void SetLength(size_t length) { m_len = length; m_lenSet = true; } protected: - wxString& m_str; - wxCharTypeBuffer m_buf; - size_t m_len; - bool m_lenSet; + size_t m_len; + bool m_lenSet; }; template @@ -2807,7 +3671,9 @@ class wxStringTypeBuffer : public wxStringTypeBufferBase { public: wxStringTypeBuffer(wxString& str, size_t lenWanted = 1024) - : wxStringTypeBufferBase(str, lenWanted) {} + : wxStringTypeBufferBase(str, lenWanted) + { } + ~wxStringTypeBuffer() { this->m_str.assign(this->m_buf.data()); @@ -2821,11 +3687,11 @@ class wxStringTypeBufferLength : public wxStringTypeBufferLengthBase { public: wxStringTypeBufferLength(wxString& str, size_t lenWanted = 1024) - : wxStringTypeBufferLengthBase(str, lenWanted) {} + : wxStringTypeBufferLengthBase(str, lenWanted) + { } ~wxStringTypeBufferLength() { - wxASSERT(this->m_lenSet); this->m_str.assign(this->m_buf.data(), this->m_len); } @@ -2859,12 +3725,12 @@ public: ~wxStringInternalBufferLength() { - wxASSERT(m_lenSet); m_str.m_impl.assign(m_buf.data(), m_len); } DECLARE_NO_COPY_CLASS(wxStringInternalBufferLength) }; + #endif // wxUSE_STL_BASED_WXSTRING @@ -2883,25 +3749,57 @@ typedef wxStringInternalBufferLength wxUTF8StringBufferLength; WXDLLIMPEXP_TEMPLATE_INSTANCE_BASE( wxStringTypeBufferBase ) -class WXDLLIMPEXP_BASE wxUTF8StringBuffer : public wxStringTypeBufferBase +// Note about inlined dtors in the classes below: this is done not for +// performance reasons but just to avoid linking errors in the MSVC DLL build +// under Windows: if a class has non-inline methods it must be declared as +// being DLL-exported but, due to an extremely interesting feature of MSVC 7 +// and later, any template class which is used as a base of a DLL-exported +// class is implicitly made DLL-exported too, as explained at the bottom of +// http://msdn.microsoft.com/en-us/library/twa2aw10.aspx (just to confirm: yes, +// _inheriting_ from a class can change whether it is being exported from DLL) +// +// But this results in link errors because the base template class is not DLL- +// exported, whether it is declared with WXDLLIMPEXP_BASE or not, because it +// does have only inline functions. So the simplest fix is to just make all the +// functions of these classes inline too. + +class wxUTF8StringBuffer : public wxStringTypeBufferBase { public: wxUTF8StringBuffer(wxString& str, size_t lenWanted = 1024) : wxStringTypeBufferBase(str, lenWanted) {} - ~wxUTF8StringBuffer(); + ~wxUTF8StringBuffer() + { + wxMBConvStrictUTF8 conv; + size_t wlen = conv.ToWChar(NULL, 0, m_buf); + wxCHECK_RET( wlen != wxCONV_FAILED, "invalid UTF-8 data in string buffer?" ); + + wxStringInternalBuffer wbuf(m_str, wlen); + conv.ToWChar(wbuf, wlen, m_buf); + } DECLARE_NO_COPY_CLASS(wxUTF8StringBuffer) }; WXDLLIMPEXP_TEMPLATE_INSTANCE_BASE( wxStringTypeBufferLengthBase ) -class WXDLLIMPEXP_BASE wxUTF8StringBufferLength - : public wxStringTypeBufferLengthBase +class wxUTF8StringBufferLength : public wxStringTypeBufferLengthBase { public: wxUTF8StringBufferLength(wxString& str, size_t lenWanted = 1024) : wxStringTypeBufferLengthBase(str, lenWanted) {} - ~wxUTF8StringBufferLength(); + ~wxUTF8StringBufferLength() + { + wxCHECK_RET(m_lenSet, "length not set"); + + wxMBConvStrictUTF8 conv; + size_t wlen = conv.ToWChar(NULL, 0, m_buf, m_len); + wxCHECK_RET( wlen != wxCONV_FAILED, "invalid UTF-8 data in string buffer?" ); + + wxStringInternalBufferLength wbuf(m_str, wlen); + conv.ToWChar(wbuf, wlen, m_buf, m_len); + wbuf.SetLength(wlen); + } DECLARE_NO_COPY_CLASS(wxUTF8StringBufferLength) }; @@ -3057,7 +3955,7 @@ inline wxCStrData::wxCStrData(const wxCStrData& data) inline wxCStrData::~wxCStrData() { if ( m_owned ) - delete wx_const_cast(wxString*, m_str); // cast to silence warnings + delete const_cast(m_str); // cast to silence warnings } // simple cases for AsChar() and AsWChar(), the complicated ones are @@ -3180,7 +4078,7 @@ void wxStringIteratorNode::DoSet(const wxString *str, if ( str ) { m_next = str->m_iterators.ptr; - wx_const_cast(wxString*, m_str)->m_iterators.ptr = this; + const_cast(m_str)->m_iterators.ptr = this; if ( m_next ) m_next->m_prev = this; } @@ -3197,7 +4095,7 @@ void wxStringIteratorNode::clear() 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; + const_cast(m_str)->m_iterators.ptr = m_next; m_next = m_prev = NULL; m_citer = NULL;