optimized wxStringTokenizer: it's now slightly faster in wchar_t build and much faster in UTF-8 build

git-svn-id: https://svn.wxwidgets.org/svn/wx/wxWidgets/trunk@47707 c3d73ce0-8a6f-49c7-b76d-6d57e0e08775

include/wx/tokenzr.h

@@ -72,12 +72,12 @@ public:
     // get current tokenizer state
         // returns the part of the string which remains to tokenize (*not* the
         // initial string)
-    wxString GetString() const { return m_string.substr(m_pos); }
+    wxString GetString() const { return wxString(m_pos, m_string.end()); }
 
         // returns the current position (i.e. one index after the last
         // returned token or 0 if GetNextToken() has never been called) in the
         // original string
-    size_t GetPosition() const { return m_pos; }
+    size_t GetPosition() const { return m_pos - m_string.begin(); }
 
     // misc
         // get the current mode - can be different from the one passed to the
@@ -111,10 +111,24 @@ public:
 protected:
     bool IsOk() const { return m_mode != wxTOKEN_INVALID; }
 
-    wxString m_string,              // the string we tokenize
+    bool DoHasMoreTokens() const;
-             m_delims;              // all possible delimiters
 
-    size_t   m_pos;                 // the current position in m_string
+    enum MoreTokensState
+    {
+        MoreTokens_Unknown,
+        MoreTokens_Yes,
+        MoreTokens_No
+    };
+
+    MoreTokensState m_hasMoreTokens;
+
+    wxString m_string;              // the string we tokenize
+    wxString::const_iterator m_stringEnd;
+    // FIXME-UTF8: use wxWcharBuffer
+    wxWxCharBuffer m_delims;        // all possible delimiters
+    size_t m_delimsLen;
+
+    wxString::const_iterator m_pos; // the current position in m_string
 
     wxStringTokenizerMode m_mode;   // see wxTOKEN_XXX values
 

src/common/tokenzr.cpp

@@ -38,6 +38,42 @@
 // implementation
 // ============================================================================
 
+// ----------------------------------------------------------------------------
+// helpers
+// ----------------------------------------------------------------------------
+
+static wxString::const_iterator
+find_first_of(const wxChar *delims, size_t len,
+              const wxString::const_iterator& from,
+              const wxString::const_iterator& end)
+{
+    wxASSERT_MSG( from <= end,  _T("invalid index") );
+
+    for ( wxString::const_iterator i = from; i != end; ++i )
+    {
+        if ( wxTmemchr(delims, *i, len) )
+            return i;
+    }
+
+    return end;
+}
+
+static wxString::const_iterator
+find_first_not_of(const wxChar *delims, size_t len,
+                  const wxString::const_iterator& from,
+                  const wxString::const_iterator& end)
+{
+    wxASSERT_MSG( from <= end,  _T("invalid index") );
+
+    for ( wxString::const_iterator i = from; i != end; ++i )
+    {
+        if ( !wxTmemchr(delims, *i, len) )
+            return i;
+    }
+
+    return end;
+}
+
 // ----------------------------------------------------------------------------
 // wxStringTokenizer construction
 // ----------------------------------------------------------------------------
@@ -78,7 +114,13 @@ void wxStringTokenizer::SetString(const wxString& str,
         }
     }
 
-    m_delims = delims;
+#if wxUSE_UNICODE // FIXME-UTF8: only wc_str()
+    m_delims = delims.wc_str();
+#else
+    m_delims = delims.mb_str();
+#endif
+    m_delimsLen = delims.length();
+
     m_mode = mode;
 
     Reinit(str);
@@ -89,8 +131,10 @@ void wxStringTokenizer::Reinit(const wxString& str)
     wxASSERT_MSG( IsOk(), _T("you should call SetString() first") );
 
     m_string = str;
-    m_pos = 0;
+    m_stringEnd = m_string.end();
+    m_pos = m_string.begin();
     m_lastDelim = _T('\0');
+    m_hasMoreTokens = MoreTokens_Unknown;
 }
 
 // ----------------------------------------------------------------------------
@@ -99,10 +143,29 @@ void wxStringTokenizer::Reinit(const wxString& str)
 
 // do we have more of them?
 bool wxStringTokenizer::HasMoreTokens() const
+{
+    // GetNextToken() calls HasMoreTokens() and so HasMoreTokens() is called
+    // twice in every interation in the following common usage patten:
+    //     while ( HasMoreTokens() )
+    //        GetNextToken();
+    // We optimize this case by caching HasMoreTokens() return value here:
+    if ( m_hasMoreTokens == MoreTokens_Unknown )
+    {
+        bool r = DoHasMoreTokens();
+        wxConstCast(this, wxStringTokenizer)->m_hasMoreTokens =
+            r ? MoreTokens_Yes : MoreTokens_No;
+        return r;
+    }
+    else
+        return m_hasMoreTokens == MoreTokens_Yes;
+}
+
+bool wxStringTokenizer::DoHasMoreTokens() const
 {
     wxCHECK_MSG( IsOk(), false, _T("you should call SetString() first") );
 
-    if ( m_string.find_first_not_of(m_delims, m_pos) != wxString::npos )
+    if ( find_first_not_of(m_delims, m_delimsLen, m_pos, m_stringEnd)
+         != m_stringEnd )
     {
         // there are non delimiter characters left, so we do have more tokens
         return true;
@@ -114,7 +177,7 @@ bool wxStringTokenizer::HasMoreTokens() const
         case wxTOKEN_RET_DELIMS:
             // special hack for wxTOKEN_RET_EMPTY: we should return the initial
             // empty token even if there are only delimiters after it
-            return m_pos == 0 && !m_string.empty();
+            return !m_string.empty() && m_pos == m_string.begin();
 
         case wxTOKEN_RET_EMPTY_ALL:
             // special hack for wxTOKEN_RET_EMPTY_ALL: we can know if we had
@@ -123,7 +186,7 @@ bool wxStringTokenizer::HasMoreTokens() const
             // up to the end of the string in GetNextToken(), but if it is not
             // NUL yet we still have this last token to return even if m_pos is
             // already at m_string.length()
-            return m_pos < m_string.length() || m_lastDelim != _T('\0');
+            return m_pos < m_stringEnd || m_lastDelim != _T('\0');
 
         case wxTOKEN_INVALID:
         case wxTOKEN_DEFAULT:
@@ -147,7 +210,7 @@ size_t wxStringTokenizer::CountTokens() const
     //     important if its implementation here is not as efficient as it
     //     could be -- but OTOH like this we're sure to get the correct answer
     //     in all modes
-    wxStringTokenizer tkz(m_string.c_str() + m_pos, m_delims, m_mode);
+    wxStringTokenizer tkz(wxString(m_pos, m_stringEnd), m_delims, m_mode);
 
     size_t count = 0;
     while ( tkz.HasMoreTokens() )
@@ -174,18 +237,21 @@ wxString wxStringTokenizer::GetNextToken()
             break;
         }
 
+        m_hasMoreTokens = MoreTokens_Unknown;
+
         // find the end of this token
-        size_t pos = m_string.find_first_of(m_delims, m_pos);
+        wxString::const_iterator pos =
+            find_first_of(m_delims, m_delimsLen, m_pos, m_stringEnd);
 
         // and the start of the next one
-        if ( pos == wxString::npos )
+        if ( pos == m_stringEnd )
         {
             // no more delimiters, the token is everything till the end of
             // string
-            token.assign(m_string, m_pos, wxString::npos);
+            token.assign(m_pos, m_stringEnd);
 
             // skip the token
-            m_pos = m_string.length();
+            m_pos = m_stringEnd;
 
             // it wasn't terminated
             m_lastDelim = _T('\0');
@@ -194,16 +260,15 @@ wxString wxStringTokenizer::GetNextToken()
         {
             // in wxTOKEN_RET_DELIMS mode we return the delimiter character
             // with token, otherwise leave it out
-            size_t len = pos - m_pos;
             if ( m_mode == wxTOKEN_RET_DELIMS )
-                len++;
+                ++pos;
 
-            token.assign(m_string, m_pos, len);
+            token.assign(m_pos, pos);
 
             // skip the token and the trailing delimiter
             m_pos = pos + 1;
 
-            m_lastDelim = m_string[pos];
+            m_lastDelim = (pos == m_stringEnd) ? _T('\0') : (wxChar)*pos;
         }
     }
     while ( !AllowEmpty() && token.empty() );