What do people on here think of the string class that David Eberly uses in his '3D Game Engine Architecture' book..?

Hi Rez,
Thanks for the reply..:)

Here is the string class..:

#ifndef WM3STRING_H
#define WM3STRING_H

#include "Wm3System.h"

namespace Wm3
{

class WM3_ITEM String
{
public:
// construction and destruction
String (const char* acText = 0);
String (int iLength, const char* acText);
String (const String& rkString);
~String ();

// member access
int GetLength () const;
operator const char* () const;

// assignment, comparisons, implicit conversion (support for hash tables)
String& operator= (const String& rkString);
String& operator+= (const String& rkString);
String operator+ (const String& rkString);
bool operator== (const String& rkString) const;
bool operator!= (const String& rkString) const;
operator unsigned int () const;

// memory and disk footprint (support for streaming)
int GetMemoryUsed () const;
int GetDiskUsed () const;

// Case-related
String ToUpper () const;
String ToLower () const;

private:
// Text is stored as null-terminated character string in memory. The
// length counts all but the non-null character. When written to disk,
// the length is stored first and all but the non-null character are
// stored second.
int m_iLength;
char* m_acText;
};

#include "Wm3String.inl"

}

#endif

The interface seems nice and lightweight, but I don't really see an advantage over the stl std::string class. Why not just use that since it's standard C++?

That is what I was thinking....most other books I have read(all in fact...I think) all use standard std::string.

Oh...bassically he says that its more efficient for streaming strings than the std::string..

..efficiency for real-time rendering and all that..