pegasus/src/Pegasus/Common/String.cpp - diff

Return to String.cpp CVS log

Up to [Pegasus] / pegasus / src / Pegasus / Common

Diff for /pegasus/src/Pegasus/Common/String.cpp between version 1.111 and 1.111.6.16

version 1.111, 2005/05/25 00:48:26

version 1.111.6.16, 2005/10/14 15:18:37

Line 27

//==============================================================================

// Author: Mike Brasher (mbrasher@bmc.com)

// Author: Mike Brasher (mbrasher@austin.rr.com)

// Modified By: Roger Kumpf, Hewlett-Packard Company (roger_kumpf@hp.com)

// Modified By:

// Roger Kumpf, Hewlett-Packard Company (roger_kumpf@hp.com)

// Josephine Eskaline Joyce, IBM (jojustin@in.ibm.com) for Bug#3297

// David Dillard, VERITAS Software Corp.

// David Dillard, VERITAS Software Corp. (david.dillard@veritas.com)

// (david.dillard@veritas.com)

// Mike Brasher (mike-brasher@austin.rr.com)

//%/////////////////////////////////////////////////////////////////////////////

#include <cassert>

#include <cstring>

#include "String.h"

#include "Array.h"

#include "AutoPtr.h"

#include "InternalException.h"

#include <Pegasus/Common/CommonUTF.h>

#include "CommonUTF.h"

#include "MessageLoader.h"

#include "StringRep.h"

#ifdef PEGASUS_HAS_ICU

#include <unicode/ustring.h>

#include <unicode/uchar.h>

#endif

PEGASUS_USING_STD;

PEGASUS_NAMESPACE_BEGIN

///////////////////////////////////////////////////////////////////////////////

//==============================================================================

// CString

// Compile-time macros (undefined by default).

///////////////////////////////////////////////////////////////////////////////

// PEGASUS_STRING_NO_THROW -- suppresses throwing of exceptions

// PEGASUS_STRING_NO_UTF8 -- don't generate slower UTF8 code.

//==============================================================================

CString::CString()

//==============================================================================

: _rep(0)

// File-scope definitions:

//==============================================================================

// Note: this table is much faster than the system toupper(). Please do not

// change.

const Uint8 _toUpperTable[256] =

{

}

0x00,0x01,0x02,0x03,0x04,0x05,0x06,0x07,

0x08,0x09,0x0A,0x0B,0x0C,0x0D,0x0E,0x0F,

0x10,0x11,0x12,0x13,0x14,0x15,0x16,0x17,

0x18,0x19,0x1A,0x1B,0x1C,0x1D,0x1E,0x1F,

0x20,0x21,0x22,0x23,0x24,0x25,0x26,0x27,

0x28,0x29,0x2A,0x2B,0x2C,0x2D,0x2E,0x2F,

0x30,0x31,0x32,0x33,0x34,0x35,0x36,0x37,

0x38,0x39,0x3A,0x3B,0x3C,0x3D,0x3E,0x3F,

0x40,0x41,0x42,0x43,0x44,0x45,0x46,0x47,

0x48,0x49,0x4A,0x4B,0x4C,0x4D,0x4E,0x4F,

0x50,0x51,0x52,0x53,0x54,0x55,0x56,0x57,

0x58,0x59,0x5A,0x5B,0x5C,0x5D,0x5E,0x5F,

0x60,0x41,0x42,0x43,0x44,0x45,0x46,0x47,

0x48,0x49,0x4A,0x4B,0x4C,0x4D,0x4E,0x4F,

0x50,0x51,0x52,0x53,0x54,0x55,0x56,0x57,

0x58,0x59,0x5A,0x7B,0x7C,0x7D,0x7E,0x7F,

0x80,0x81,0x82,0x83,0x84,0x85,0x86,0x87,

0x88,0x89,0x8A,0x8B,0x8C,0x8D,0x8E,0x8F,

0x90,0x91,0x92,0x93,0x94,0x95,0x96,0x97,

0x98,0x99,0x9A,0x9B,0x9C,0x9D,0x9E,0x9F,

0xA0,0xA1,0xA2,0xA3,0xA4,0xA5,0xA6,0xA7,

0xA8,0xA9,0xAA,0xAB,0xAC,0xAD,0xAE,0xAF,

0xB0,0xB1,0xB2,0xB3,0xB4,0xB5,0xB6,0xB7,

0xB8,0xB9,0xBA,0xBB,0xBC,0xBD,0xBE,0xBF,

0xC0,0xC1,0xC2,0xC3,0xC4,0xC5,0xC6,0xC7,

0xC8,0xC9,0xCA,0xCB,0xCC,0xCD,0xCE,0xCF,

0xD0,0xD1,0xD2,0xD3,0xD4,0xD5,0xD6,0xD7,

0xD8,0xD9,0xDA,0xDB,0xDC,0xDD,0xDE,0xDF,

0xE0,0xE1,0xE2,0xE3,0xE4,0xE5,0xE6,0xE7,

0xE8,0xE9,0xEA,0xEB,0xEC,0xED,0xEE,0xEF,

0xF0,0xF1,0xF2,0xF3,0xF4,0xF5,0xF6,0xF7,

0xF8,0xF9,0xFA,0xFB,0xFC,0xFD,0xFE,0xFF,

};

CString::CString(const CString& cstr)

// Note: this table is much faster than the system tulower(). Please do not

// change.

const Uint8 _toLowerTable[256] =

{

_rep = 0;

0x00,0x01,0x02,0x03,0x04,0x05,0x06,0x07,

0x08,0x09,0x0A,0x0B,0x0C,0x0D,0x0E,0x0F,

0x10,0x11,0x12,0x13,0x14,0x15,0x16,0x17,

0x18,0x19,0x1A,0x1B,0x1C,0x1D,0x1E,0x1F,

0x20,0x21,0x22,0x23,0x24,0x25,0x26,0x27,

0x28,0x29,0x2A,0x2B,0x2C,0x2D,0x2E,0x2F,

0x30,0x31,0x32,0x33,0x34,0x35,0x36,0x37,

0x38,0x39,0x3A,0x3B,0x3C,0x3D,0x3E,0x3F,

0x40,0x61,0x62,0x63,0x64,0x65,0x66,0x67,

0x68,0x69,0x6A,0x6B,0x6C,0x6D,0x6E,0x6F,

0x70,0x71,0x72,0x73,0x74,0x75,0x76,0x77,

0x78,0x79,0x7A,0x5B,0x5C,0x5D,0x5E,0x5F,

0x60,0x61,0x62,0x63,0x64,0x65,0x66,0x67,

0x68,0x69,0x6A,0x6B,0x6C,0x6D,0x6E,0x6F,

0x70,0x71,0x72,0x73,0x74,0x75,0x76,0x77,

0x78,0x79,0x7A,0x7B,0x7C,0x7D,0x7E,0x7F,

0x80,0x81,0x82,0x83,0x84,0x85,0x86,0x87,

0x88,0x89,0x8A,0x8B,0x8C,0x8D,0x8E,0x8F,

0x90,0x91,0x92,0x93,0x94,0x95,0x96,0x97,

0x98,0x99,0x9A,0x9B,0x9C,0x9D,0x9E,0x9F,

0xA0,0xA1,0xA2,0xA3,0xA4,0xA5,0xA6,0xA7,

0xA8,0xA9,0xAA,0xAB,0xAC,0xAD,0xAE,0xAF,

0xB0,0xB1,0xB2,0xB3,0xB4,0xB5,0xB6,0xB7,

0xB8,0xB9,0xBA,0xBB,0xBC,0xBD,0xBE,0xBF,

0xC0,0xC1,0xC2,0xC3,0xC4,0xC5,0xC6,0xC7,

0xC8,0xC9,0xCA,0xCB,0xCC,0xCD,0xCE,0xCF,

0xD0,0xD1,0xD2,0xD3,0xD4,0xD5,0xD6,0xD7,

0xD8,0xD9,0xDA,0xDB,0xDC,0xDD,0xDE,0xDF,

0xE0,0xE1,0xE2,0xE3,0xE4,0xE5,0xE6,0xE7,

0xE8,0xE9,0xEA,0xEB,0xEC,0xED,0xEE,0xEF,

0xF0,0xF1,0xF2,0xF3,0xF4,0xF5,0xF6,0xF7,

0xF8,0xF9,0xFA,0xFB,0xFC,0xFD,0xFE,0xFF,

};

if (cstr._rep)

// Converts 16-bit characters to upper case. This routine is faster than the

// system toupper(). Please do not change.

inline Uint16 _toUpper(Uint16 x)

{

_rep = (void*)new char[strlen((char*)cstr._rep)+1];

return (x & 0xFF00) ? x : _toUpperTable[x];

strcpy((char*)_rep, (char*)cstr._rep);

}

// Converts 16-bit characters to lower case. This routine is faster than the

// system toupper(). Please do not change.

inline Uint16 _toLower(Uint16 x)

{

return (x & 0xFF00) ? x : _toLowerTable[x];

}

CString::CString(char* cstr)

// Rounds x up to the nearest power of two (or just returns 8 if x < 8).

: _rep(cstr)

static Uint32 _roundUpToPow2(Uint32 x)

{

#ifndef PEGASUS_STRING_NO_THROW

if (x > 0x0FFFFFFF)

throw PEGASUS_STD(bad_alloc)();

#endif

if (x < 8)

return 8;

x--;

x |= (x >> 1);

x |= (x >> 2);

x |= (x >> 4);

x |= (x >> 8);

x |= (x >> 16);

x++;

return x;

}

template<class P, class Q>

static void _copy(P* p, const Q* q, size_t n)

{

// The following employs loop unrolling for efficiency. Please do not

// eliminate.

while (n >= 8)

{

p[0] = q[0];

p[1] = q[1];

p[2] = q[2];

p[3] = q[3];

p[4] = q[4];

p[5] = q[5];

p[6] = q[6];

p[7] = q[7];

p += 8;

q += 8;

n -= 8;

}

while (n >= 4)

{

p[0] = q[0];

p[1] = q[1];

p[2] = q[2];

p[3] = q[3];

p += 4;

q += 4;

n -= 4;

}

CString::~CString()

while (n--)

*p++ = *q++;

}

static Uint16* _find(const Uint16* s, size_t n, Uint16 c)

{

if (_rep)

// The following employs loop unrolling for efficiency. Please do not

// eliminate.

while (n >= 4)

{

delete [] (char*)_rep;

if (s[0] == c)

return (Uint16*)s;

if (s[1] == c)

return (Uint16*)&s[1];

if (s[2] == c)

return (Uint16*)&s[2];

if (s[3] == c)

return (Uint16*)&s[3];

n -= 4;

s += 4;

}

if (n)

{

if (*s == c)

return (Uint16*)s;

s++;

n--;

}

CString& CString::operator=(const CString& cstr)

if (n)

{

if (&cstr != this)

if (*s == c)

return (Uint16*)s;

s++;

n--;

}

if (n && *s == c)

return (Uint16*)s;

// Not found!

return 0;

}

static int _compare(const Uint16* s1, const Uint16* s2)

{

if (_rep)

while (*s1 && *s2)

{

delete [] (char*)_rep;

int r = *s1++ - *s2++;

_rep = 0;

if (r)

return r;

}

if (cstr._rep)

{

if (*s2)

_rep = (char*)new char[strlen((char*)cstr._rep)+1];

return -1;

strcpy((char*)_rep, (char*)cstr._rep);

else if (*s1)

return 1;

return 0;

}

static int _compareNoUTF8(const Uint16* s1, const char* s2)

{

Uint16 c1;

Uint16 c2;

{

c1 = *s1++;

c2 = *s2++;

if (c1 == 0)

return c1 - c2;

}

return *this;

while (c1 == c2);

return c1 - c2;

}

CString::operator const char*() const

static int _compare(const Uint16* s1, const Uint16* s2, size_t n)

{

return (char*)_rep;

// This should only be called when s1 and s2 have the same length.

while (n-- && (*s1++ - *s2++) == 0)

;

return s1[-1] - s2[-1];

}

///////////////////////////////////////////////////////////////////////////////

static inline void _copy(Uint16* s1, const Uint16* s2, size_t n)

{

// String

memcpy(s1, s2, n * sizeof(Uint16));

}

///////////////////////////////////////////////////////////////////////////////

const String String::EMPTY = String();

void StringThrowOutOfBounds()

{

throw IndexOutOfBoundsException();

}

inline Uint32 _StrLen(const Char16* str)

inline void _checkNullPointer(const void* ptr)

{

if (!str)

#ifdef PEGASUS_STRING_NO_THROW

if (!ptr)

throw NullPointer();

Uint32 n = 0;

#endif

}

static void _StringThrowBadUTF8(Uint32 index)

{

MessageLoaderParms parms(

"Common.String.BAD_UTF8",

"The byte sequence starting at index $0 "

"is not valid UTF-8 encoding.",

index);

throw Exception(parms);

}

static size_t _copyFromUTF8(

Uint16* dest,

const char* src,

size_t n,

size_t& utf8_error_index)

{

Uint16* p = dest;

const Uint8* q = (const Uint8*)src;

while (*str++)

// Process leading 7-bit ASCII characters (to avoid UTF8 overhead later).

n++;

// Use loop-unrolling.

return n;

while (n >=8 && ((q[0]|q[1]|q[2]|q[3]|q[4]|q[5]|q[6]|q[7]) & 0x80) == 0)

{

p[0] = q[0];

p[1] = q[1];

p[2] = q[2];

p[3] = q[3];

p[4] = q[4];

p[5] = q[5];

p[6] = q[6];

p[7] = q[7];

p += 8;

q += 8;

n -= 8;

}

while (n >=4 && ((q[0]|q[1]|q[2]|q[3]) & 0x80) == 0)

// Converts a utf-8 char buffer to utf-16 and appends the utf-16 to the Array.

// n is the length of the input char *, if stopAtTerm is 0

// A terminator character is appended to the end.

// Note that each input char is converted individually, which gives

// the fastest performance.

void _convertAndAppend(const char* str, Array<Char16>& c16a, Uint32 n, Uint8 stopAtTerm)

{

Uint32 i = 0;

p[0] = q[0];

while ((stopAtTerm && *str) || (!stopAtTerm && i < n))

p[1] = q[1];

p[2] = q[2];

p[3] = q[3];

p += 4;

q += 4;

n -= 4;

}

switch (n)

{

if (*(Uint8*)str <= 0x7f)

case 0:

return p - dest;

case 1:

if (q[0] < 128)

{

// Current byte sequence is in the us-ascii range.

p[0] = q[0];

c16a.append(Uint8(*str++));

return p + 1 - dest;

}

else

break;

case 2:

if (((q[0]|q[1]) & 0x80) == 0)

{

p[0] = q[0];

// Current byte sequence is not in the us-ascii range.

p[1] = q[1];

return p + 2 - dest;

}

break;

case 3:

if (((q[0]|q[1]|q[2]) & 0x80) == 0)

{

p[0] = q[0];

p[1] = q[1];

p[2] = q[2];

return p + 3 - dest;

}

break;

}

// Check if the byte sequence is valid utf-8, and if so,

// Process remaining characters.

// call the converter to utf-16

Uint16 tgt[3];

while (n)

tgt[1] = 0;

{

Uint8 c = UTF_8_COUNT_TRAIL_BYTES(*str);

// Optimize for 7-bit ASCII case.

if ( (!stopAtTerm && i + c >= n) ||

(!isValid_U8((const Uint8 *)str, c+1)) )

if (*q < 128)

{

// Note about error conditions.

*p++ = *q++;

// It is possible that the last utf-8 char before the

n--;

// end of input string extends past the end of the input string.

// This is caught in both cases -

// If counting up to n, then the test above catches it.

// If converting until terminator found, then a terminator

// in the middle of a multi-byte utf-8 char is invalid.

MessageLoaderParms parms("Common.String.BAD_UTF8",

"The byte sequence starting at index $0 is not valid UTF-8 encoding.",

i);

throw Exception(parms);

}

else

{

// str is incremented by this call to the start of the next char

Uint8 c = UTF_8_COUNT_TRAIL_BYTES(*q) + 1;

Uint16 * tgtBuf = tgt;

UTF8toUTF16((const Uint8 **)&str, (Uint8 *)&str[c+1], &tgtBuf, &tgtBuf[2]);

if (c > n || !isValid_U8(q, c) ||

c16a.append(tgt[0]);

UTF8toUTF16(&q, q + c, &p, p + n) != 0)

if (tgt[1])

{

// Its a utf-16 surrogate pair (uses 2 Char16's)

utf8_error_index = q - (const Uint8*)src;

c16a.append(tgt[1]);

return size_t(-1);

}

// bump by the trailing byte count

n -= c;

i += c;

}

i++;

return p - dest;

} // end while

c16a.append('\0');

}

class StringRep

// Note: dest must be at least three times src (plus an extra byte for

// terminator).

static inline size_t _copyToUTF8(char* dest, const Uint16* src, size_t n)

{

public:

// The following employs loop unrolling for efficiency. Please do not

StringRep()

// eliminate.

{}

StringRep(const StringRep& r)

: c16a(r.c16a)

{}

StringRep(const Char16* str)

: c16a(str, _StrLen(str) + 1)

{}

Array<Char16> c16a;

const Uint16* q = src;

};

Uint8* p = (Uint8*)dest;

String::String()

while (n >= 4 && q[0] < 128 && q[1] < 128 && q[2] < 128 && q[3] < 128)

{

_rep = new StringRep;

p[0] = q[0];

_rep->c16a.append('\0');

p[1] = q[1];

p[2] = q[2];

p[3] = q[3];

p += 4;

q += 4;

n -= 4;

}

String::String(const String& str)

switch (n)

{

if (str._rep != NULL)

case 0:

return p - (Uint8*)dest;

case 1:

if (q[0] < 128)

{

_rep = new StringRep(*str._rep);

p[0] = q[0];

return p + 1 - (Uint8*)dest;

}

else

break;

case 2:

if (q[0] < 128 && q[1] < 128)

{

_rep = new StringRep();

p[0] = q[0];

p[1] = q[1];

return p + 2 - (Uint8*)dest;

}

break;

case 3:

if (q[0] < 128 && q[1] < 128 && q[2] < 128)

{

p[0] = q[0];

p[1] = q[1];

p[2] = q[2];

return p + 3 - (Uint8*)dest;

}

break;

}

// If this line was reached, there must be characters greater than 128.

String::String(const String& str, Uint32 n)

UTF16toUTF8(&q, q + n, &p, p + 3 * n);

return p - (Uint8*)dest;

}

static inline size_t _convert(

Uint16* p, const char* q, size_t n, size_t& utf8_error_index)

{

_rep = new StringRep;

#ifdef PEGASUS_STRING_NO_UTF8

assign(str.getChar16Data(), n);

_copy(p, q, n);

return n;

#else

return _copyFromUTF8(p, q, n, utf8_error_index);

#endif

}

String::String(const Char16* str)

//==============================================================================

// class CString

//==============================================================================

CString::CString(const CString& cstr) : _rep(0)

{

if ( str == 0 )

if (cstr._rep)

{

throw NullPointer();

size_t n = strlen(cstr._rep) + 1;

_rep = (char*)operator new(n);

memcpy(_rep, cstr._rep, n);

}

_rep = new StringRep(str);

}

String::String(const Char16* str, Uint32 n)

CString& CString::operator=(const CString& cstr)

{

if (&cstr != this)

{

if ( str == 0 )

if (_rep)

{

throw NullPointer();

operator delete(_rep);

_rep = 0;

}

_rep = new StringRep;

if (cstr._rep)

assign(str, n);

{

size_t n = strlen(cstr._rep) + 1;

_rep = (char*)operator new(n);

memcpy(_rep, cstr._rep, n);

}

String::String(const char* str)

return *this;

}

//==============================================================================

// class StringRep

//==============================================================================

StringRep StringRep::_emptyRep;

inline StringRep* StringRep::alloc(size_t cap)

{

if ( str == 0 )

#ifndef PEGASUS_STRING_NO_THROW

// Any string bigger than this is seriously suspect.

if (cap > 0x0FFFFFFF)

throw PEGASUS_STD(bad_alloc)();

#endif

StringRep* rep = (StringRep*)::operator new(

sizeof(StringRep) + cap * sizeof(Uint16));

rep->cap = cap;

new(&rep->refs) AtomicInt(1);

return rep;

}

static inline void _reserve(StringRep*& rep, Uint32 cap)

{

throw NullPointer();

if (cap > rep->cap || rep->refs.value() != 1)

{

size_t n = _roundUpToPow2(cap);

StringRep* newRep = StringRep::alloc(n);

newRep->size = rep->size;

_copy(newRep->data, rep->data, rep->size + 1);

StringRep::unref(rep);

rep = newRep;

}

_rep = new StringRep;

StringRep* StringRep::create(const Uint16* data, size_t size)

AutoPtr<StringRep> tempRep(_rep);

{

// An exception can be thrown, so use a temp AutoPtr.

StringRep* rep = StringRep::alloc(size);

_convertAndAppend(str, _rep->c16a, 0, 1);

rep->size = size;

tempRep.release();

_copy(rep->data, data, size);

rep->data[size] = '\0';

return rep;

}

String::String(const char* str, Uint32 n)

StringRep* StringRep::copyOnWrite(StringRep* rep)

{

if ( str == 0 )

// Return a new copy of rep. Release rep.

StringRep* newRep = StringRep::alloc(rep->size);

newRep->size = rep->size;

_copy(newRep->data, rep->data, rep->size);

newRep->data[newRep->size] = '\0';

StringRep::unref(rep);

return newRep;

}

StringRep* StringRep::create(const char* data, size_t size)

{

throw NullPointer();

StringRep* rep = StringRep::alloc(size);

size_t utf8_error_index;

rep->size = _convert((Uint16*)rep->data, data, size, utf8_error_index);

#ifndef PEGASUS_STRING_NO_THROW

if (rep->size == size_t(-1))

{

StringRep::free(rep);

_StringThrowBadUTF8(utf8_error_index);

}

#endif

_rep = new StringRep;

rep->data[rep->size] = '\0';

AutoPtr<StringRep> tempRep(_rep);

// An exception can be thrown, so use a temp AutoPtr.

return rep;

_convertAndAppend(str, _rep->c16a, n, 0);

tempRep.release();

}

String::~String()

Uint32 StringRep::length(const Uint16* str)

{

delete _rep;

// Note: We could unroll this but it is rarely called.

const Uint16* end = (Uint16*)str;

while (*end++)

;

return end - str - 1;

}

String& String::operator=(const String& str)

//==============================================================================

// class String

//==============================================================================

const String String::EMPTY;

String::String(const String& str, Uint32 n)

{

if (&str != this)

_checkBounds(n, str._rep->size);

_rep = StringRep::create(str._rep->data, n);

}

String::String(const Char16* str)

{

assign(str);

_checkNullPointer(str);

_rep = StringRep::create((Uint16*)str, StringRep::length((Uint16*)str));

}

return *this;

String::String(const Char16* str, Uint32 n)

{

_checkNullPointer(str);

_rep = StringRep::create((Uint16*)str, n);

}

String& String::assign(const String& str)

String::String(const char* str)

{

_rep->c16a = str._rep->c16a;

_checkNullPointer(str);

return *this;

// Set this just in case create() throws an exception.

_rep = &StringRep::_emptyRep;

_rep = StringRep::create(str, strlen(str));

}

String& String::assign(const Char16* str)

String::String(const char* str, Uint32 n)

{

if ( str == 0 )

_checkNullPointer(str);

// Set this just in case create() throws an exception.

_rep = &StringRep::_emptyRep;

_rep = StringRep::create(str, n);

}

String::String(const String& s1, const String& s2)

{

throw NullPointer();

size_t n1 = s1._rep->size;

size_t n2 = s2._rep->size;

size_t n = n1 + n2;

_rep = StringRep::alloc(n);

_copy(_rep->data, s1._rep->data, n1);

_copy(_rep->data + n1, s2._rep->data, n2);

_rep->size = n;

_rep->data[n] = '\0';

}

_rep->c16a.clear();

String::String(const String& s1, const char* s2)

_rep->c16a.append(str, _StrLen(str) + 1);

{

return *this;

_checkNullPointer(s2);

size_t n1 = s1._rep->size;

size_t n2 = strlen(s2);

_rep = StringRep::alloc(n1 + n2);

_copy(_rep->data, s1._rep->data, n1);

size_t utf8_error_index;

size_t tmp = _convert((Uint16*)_rep->data + n1, s2, n2, utf8_error_index);

#ifndef PEGASUS_STRING_NO_THROW

if (tmp == size_t(-1))

{

StringRep::free(_rep);

_rep = &StringRep::_emptyRep;

_StringThrowBadUTF8(utf8_error_index);

}

#endif

String& String::assign(const Char16* str, Uint32 n)

_rep->size = n1 + tmp;

_rep->data[_rep->size] = '\0';

}

String::String(const char* s1, const String& s2)

{

if ( str == 0 )

_checkNullPointer(s1);

size_t n1 = strlen(s1);

size_t n2 = s2._rep->size;

_rep = StringRep::alloc(n1 + n2);

size_t utf8_error_index;

size_t tmp = _convert((Uint16*)_rep->data, s1, n1, utf8_error_index);

#ifndef PEGASUS_STRING_NO_THROW

if (tmp == size_t(-1))

{

throw NullPointer();

StringRep::free(_rep);

_rep = &StringRep::_emptyRep;

_StringThrowBadUTF8(utf8_error_index);

}

#endif

_rep->c16a.clear();

_rep->size = n2 + tmp;

_rep->c16a.append(str, n);

_copy(_rep->data + n1, s2._rep->data, n2);

_rep->c16a.append('\0');

_rep->data[_rep->size] = '\0';

return *this;

}

String& String::assign(const char* str)

String& String::assign(const String& str)

{

if ( str == 0 )

if (_rep != str._rep)

{

throw NullPointer();

StringRep::unref(_rep);

StringRep::ref(_rep = str._rep);

}

_rep->c16a.clear();

_convertAndAppend(str, _rep->c16a, 0, 1);

return *this;

}

String& String::assign(const char* str, Uint32 n)

String& String::assign(const Char16* str, Uint32 n)

{

if ( str == 0 )

_checkNullPointer(str);

if (n > _rep->cap || _rep->refs.value() != 1)

{

throw NullPointer();

StringRep::unref(_rep);

_rep = StringRep::alloc(n);

}

_rep->c16a.clear();

_rep->size = n;

_convertAndAppend(str, _rep->c16a, n, 0);

_copy(_rep->data, (Uint16*)str, n);

_rep->data[n] = '\0';

return *this;

}

void String::clear()

String& String::assign(const char* str, Uint32 n)

{

_checkNullPointer(str);

if (n > _rep->cap || _rep->refs.value() != 1)

{

_rep->c16a.clear();

StringRep::unref(_rep);

_rep->c16a.append('\0');

_rep = StringRep::alloc(n);

}

void String::reserveCapacity(Uint32 capacity)

size_t utf8_error_index;

_rep->size = _convert(_rep->data, str, n, utf8_error_index);

#ifndef PEGASUS_STRING_NO_THROW

if (_rep->size == size_t(-1))

{

_rep->c16a.reserveCapacity(capacity + 1);

StringRep::free(_rep);

_rep = &StringRep::_emptyRep;

_StringThrowBadUTF8(utf8_error_index);

}

#endif

_rep->data[_rep->size] = 0;

Uint32 String::size() const

return *this;

}

void String::clear()

{

#if defined (PEGASUS_OS_VMS)

if (_rep->size)

// This prevents returning a minus number.

// Seems as though the first time through

// the XML parser something doesn't get

// initialized and there is no check for

// a negative number in the parser!

Uint32 foo;

foo = _rep->c16a.size();

if (foo == 0)

{

return 0;

if (_rep->refs.value() == 1)

{

_rep->size = 0;

_rep->data[0] = '\0';

}

else

{

return (foo -1);

StringRep::unref(_rep);

_rep = &StringRep::_emptyRep;

}

#else

return _rep->c16a.size() - 1;

#endif

}

const Char16* String::getChar16Data() const

void String::reserveCapacity(Uint32 cap)

{

return _rep->c16a.getData();

_reserve(_rep, cap);

}

Char16& String::operator[](Uint32 index)

CString String::getCString() const

{

if (index > size())

// A UTF8 string can have three times as many characters as its UTF16

throw IndexOutOfBoundsException();

// counterpart, so we allocate extra memory for the worst case. In the

// best case, we may need only one third of the memory allocated. But

return _rep->c16a[index];

// downsizing the string afterwards is expensive and unecessary since

// CString objects are usually short-lived (disappearing after only a few

// instructions). CString objects are typically created on the stack as

// means to obtain a char* pointer.

#ifdef PEGASUS_STRING_NO_UTF8

char* str = (char*)operator new(_rep->size + 1);

_copy(str, _rep->data, _rep->size);

str[_rep->size] = '\0';

return CString(str);

#else

Uint32 n = 3 * _rep->size;

char* str = (char*)operator new(n + 1);

size_t size = _copyToUTF8(str, _rep->data, _rep->size);

str[size] = '\0';

return CString(str);

#endif

}

const Char16 String::operator[](Uint32 index) const

String& String::append(const Char16* str, Uint32 n)

{

if (index > size())

_checkNullPointer(str);

throw IndexOutOfBoundsException();

size_t oldSize = _rep->size;

size_t newSize = oldSize + n;

_reserve(_rep, newSize);

_copy(_rep->data + oldSize, (Uint16*)str, n);

_rep->size = newSize;

_rep->data[newSize] = '\0';

return _rep->c16a[index];

return *this;

}

String& String::append(const Char16& c)

String& String::append(const String& str)

{

_rep->c16a.insert(_rep->c16a.size() - 1, c);

return append((Char16*)str._rep->data, str._rep->size);

return *this;

}

String& String::append(const Char16* str, Uint32 n)

String& String::append(const char* str, Uint32 size)

{

if (str == 0)

_checkNullPointer(str);

size_t oldSize = _rep->size;

size_t cap = oldSize + size;

_reserve(_rep, cap);

size_t utf8_error_index;

size_t tmp = _convert(

(Uint16*)_rep->data + oldSize, str, size, utf8_error_index);

#ifndef PEGASUS_STRING_NO_THROW

if (tmp == size_t(-1))

{

throw NullPointer();

StringRep::free(_rep);

_rep = &StringRep::_emptyRep;

_StringThrowBadUTF8(utf8_error_index);

}

#endif

_rep->size += tmp;

_rep->data[_rep->size] = '\0';

_rep->c16a.reserveCapacity(_rep->c16a.size() + n);

_rep->c16a.remove(_rep->c16a.size() - 1);

_rep->c16a.append(str, n);

_rep->c16a.append('\0');

return *this;

}

String& String::append(const String& str)

void String::remove(Uint32 index, Uint32 n)

{

return append(str.getChar16Data(), str.size());

if (n == PEG_NOT_FOUND)

}

n = _rep->size - index;

void String::remove(Uint32 index, Uint32 size)

_checkBounds(index + n, _rep->size);

{

if (size == PEG_NOT_FOUND)

size = this->size() - index;

if (index + size > this->size())

if (_rep->refs.value() != 1)

throw IndexOutOfBoundsException();

_rep = StringRep::copyOnWrite(_rep);

assert(index + n <= _rep->size);

if (size)

size_t rem = _rep->size - (index + n);

_rep->c16a.remove(index, size);

Uint16* data = _rep->data;

if (rem)

memmove(data + index, data + index + n, rem * sizeof(Uint16));

_rep->size -= n;

data[_rep->size] = '\0';

}

String String::subString(Uint32 index, Uint32 length) const

String String::subString(Uint32 index, Uint32 n) const

{

if (index < size())

// Note: this implementation is very permissive but used for

// backwards compatibility.

if (index < _rep->size)

{

if ((length == PEG_NOT_FOUND) || (length > size() - index))

if (n == PEG_NOT_FOUND || n > _rep->size - index)

length = size() - index;

n = _rep->size - index;

return String(getChar16Data() + index, length);

return String((Char16*)_rep->data + index, n);

}

return String();

Line 468

Line 920

Uint32 String::find(Char16 c) const

{

const Char16* first = getChar16Data();

Uint16* p = (Uint16*)_find(_rep->data, _rep->size, c);

for (const Char16* p = first; *p; p++)

if (p)

{

return p - _rep->data;

if (*p == c)

return p - first;

}

return PEG_NOT_FOUND;

}

Uint32 String::find(Uint32 index, Char16 c) const

{

const Char16* data = getChar16Data();

_checkBounds(index, _rep->size);

for (Uint32 i = index, n = size(); i < n; i++)

if (index >= _rep->size)

{

return PEG_NOT_FOUND;

if (data[i] == c)

return i;

Uint16* p = (Uint16*)_find(_rep->data + index, _rep->size - index, c);

}

if (p)

return p - _rep->data;

return PEG_NOT_FOUND;

}

Uint32 String::find(const String& s) const

Uint32 StringFindAux(

const StringRep* _rep, const Char16* s, Uint32 n)

{

const Char16* pSubStr = s.getChar16Data();

_checkNullPointer(s);

const Char16* pStr = getChar16Data();

Uint32 subStrLen = s.size();

const Uint16* data = _rep->data;

Uint32 strLen = size();

size_t rem = _rep->size;

if (subStrLen > strLen)

while (n <= rem)

{

return PEG_NOT_FOUND;

Uint16* p = (Uint16*)_find(data, rem, s[0]);

}

// loop to find first char match

if (!p)

Uint32 loc = 0;

break;

for( ; loc <= (strLen-subStrLen); loc++)

{

if (memcmp(p, s, n * sizeof(Uint16)) == 0)

if (*pStr++ == *pSubStr) // match first char

return p - _rep->data;

{

// point to substr 2nd char

p++;

const Char16* p = pSubStr + 1;

rem -= p - data;

data = p;

// Test remaining chars for equal

Uint32 i = 1;

for (; i < subStrLen; i++)

if (*pStr++ != *p++ )

{pStr-=i; break;} // break from loop

if (i == subStrLen)

return loc;

}

return PEG_NOT_FOUND;

}

Uint32 String::find(const char* s) const

{

_checkNullPointer(s);

// Note: could optimize away creation of temporary, but this is rarely

// called.

return find(String(s));

}

Uint32 String::reverseFind(Char16 c) const

{

const Char16* first = getChar16Data();

Uint16 x = c;

const Char16* last = getChar16Data() + size();

Uint16* p = _rep->data;

Uint16* q = _rep->data + _rep->size;

while (last != first)

while (q != p)

{

if (*--last == c)

if (*--q == x)

return last - first;

return q - p;

}

return PEG_NOT_FOUND;

Line 542

Line 996

void String::toLower()

{

#ifdef PEGASUS_HAS_ICU

if (InitializeICU::initICUSuccessful())

{

if (_rep->refs.value() != 1)

_rep = StringRep::copyOnWrite(_rep);

// This will do a locale-insensitive, but context-sensitive convert.

// Context-sensitive prevents any optimizations that try to

// Since context-sensitive casing looks at adjacent chars, this

// convert just the ascii before calling ICU.

// prevents optimizations where the us-ascii is converted before

// calling ICU.

// The string may shrink or expand after the convert.

int32_t sz = size();

//// First calculate size of resulting string. u_strToLower() returns

UChar* destbuf = new UChar[sz + 1];

//// only the size when zero is passed as the destination size argument.

const UChar* srcbuf = (const UChar *)getChar16Data();

UErrorCode err = U_ZERO_ERROR;

int32_t needed = u_strToLower(destbuf, sz + 1 , srcbuf, sz, NULL, &err);

int32_t newSize = u_strToLower(

if (err == U_BUFFER_OVERFLOW_ERROR)

NULL, 0, (UChar*)_rep->data, _rep->size, NULL, &err);

{

delete [] destbuf;

destbuf = new UChar[needed + 1];

err = U_ZERO_ERROR;

u_strToLower(destbuf, needed + 1 , srcbuf, sz, NULL, &err);

}

if (U_FAILURE(err))

{

delete [] destbuf;

throw Exception(u_errorName(err));

}

if (needed == sz)

//// Reserve enough space for the result.

{

Char16* from = (Char16*)destbuf;

if ((Uint32)newSize > _rep->cap)

for (Char16* to = &_rep->c16a[0]; *to; to++, from++)

_reserve(_rep, newSize);

{

*to = *from;

}

else

{

assign((Char16 *)destbuf, needed);

}

delete [] destbuf;

//// Perform the conversion (overlapping buffers are allowed).

u_strToLower((UChar*)_rep->data, newSize,

(UChar*)_rep->data, _rep->size, NULL, &err);

_rep->size = newSize;

return;

}

else

#endif

#endif /* PEGASUS_HAS_ICU */

{

for (Char16* p = &_rep->c16a[0]; *p; p++)

if (_rep->refs.value() != 1)

_rep = StringRep::copyOnWrite(_rep);

Uint16* p = _rep->data;

size_t n = _rep->size;

for (; n--; p++)

{

if (*p <= PEGASUS_MAX_PRINTABLE_CHAR)

if (!(*p & 0xFF00))

*p = tolower(*p);

*p = _toLower(*p);

}

void String::toUpper()

{

#ifdef PEGASUS_HAS_ICU

if (InitializeICU::initICUSuccessful())

{

if (_rep->refs.value() != 1)

_rep = StringRep::copyOnWrite(_rep);

// This will do a locale-insensitive, but context-sensitive convert.

// Context-sensitive prevents any optimizations that try to

// Since context-sensitive casing looks at adjacent chars, this

// convert just the ascii before calling ICU.

// prevents optimizations where the us-ascii is converted before

// calling ICU.

// The string may shrink or expand after the convert.

int32_t sz = size();

//// First calculate size of resulting string. u_strToUpper() returns

UChar* destbuf = new UChar[sz + 1];

//// only the size when zero is passed as the destination size argument.

const UChar* srcbuf = (const UChar *)getChar16Data();

UErrorCode err = U_ZERO_ERROR;

int32_t needed = u_strToUpper(destbuf, sz + 1 , srcbuf, sz, NULL, &err);

int32_t newSize = u_strToUpper(

if (err == U_BUFFER_OVERFLOW_ERROR)

NULL, 0, (UChar*)_rep->data, _rep->size, NULL, &err);

{

delete [] destbuf;

destbuf = new UChar[needed + 1];

err = U_ZERO_ERROR;

u_strToUpper(destbuf, needed + 1 , srcbuf, sz, NULL, &err);

}

if (U_FAILURE(err))

{

delete [] destbuf;

throw Exception(u_errorName(err));

}

if (needed == sz)

//// Reserve enough space for the result.

{

Char16* from = (Char16*)destbuf;

for (Char16* to = &_rep->c16a[0]; *to; to++, from++)

{

*to = *from;

}

else

{

assign((Char16 *)destbuf, needed);

}

delete [] destbuf;

if ((Uint32)newSize > _rep->cap)

}

_reserve(_rep, newSize);

else

#endif

{

for (Char16* p = &_rep->c16a[0]; *p; p++)

{

if (*p <= PEGASUS_MAX_PRINTABLE_CHAR)

*p = toupper(*p);

}

int String::compare(const String& s1, const String& s2, Uint32 n)

//// Perform the conversion (overlapping buffers are allowed).

{

const Char16* s1c16 = s1.getChar16Data();

const Char16* s2c16 = s2.getChar16Data();

while (n--)

u_strToUpper((UChar*)_rep->data, newSize,

{

(UChar*)_rep->data, _rep->size, NULL, &err);

int r = *s1c16++ - *s2c16++;

if (r)

_rep->size = newSize;

return r;

return;

}

return 0;

#endif /* PEGASUS_HAS_ICU */

if (_rep->refs.value() != 1)

_rep = StringRep::copyOnWrite(_rep);

Uint16* p = _rep->data;

size_t n = _rep->size;

for (; n--; p++)

*p = _toUpper(*p);

}

int String::compare(const String& s1, const String& s2)

int String::compare(const String& s1, const String& s2, Uint32 n)

{

const Char16* s1c16 = s1.getChar16Data();

assert(n <= s1._rep->size);

const Char16* s2c16 = s2.getChar16Data();

assert(n <= s2._rep->size);

while (*s1c16 && *s2c16)

// Ignoring error in which n is greater than s1.size() or s2.size()

{

return _compare(s1._rep->data, s2._rep->data, n);

int r = *s1c16++ - *s2c16++;

}

if (r)

int String::compare(const String& s1, const String& s2)

return r;

{

return _compare(s1._rep->data, s2._rep->data);

}

if (*s2c16)

int String::compare(const String& s1, const char* s2)

return -1;

{

else if (*s1c16)

_checkNullPointer(s2);

return 1;

return 0;

#ifdef PEGASUS_STRING_NO_UTF8

return _compareNoUTF8(s1._rep->data, s2);

#else

// ATTN: optimize this!

return String::compare(s1, String(s2));

#endif

}

int String::compareNoCase(const String& s1, const String& s2)

int String::compareNoCase(const String& str1, const String& str2)

{

#ifdef PEGASUS_HAS_ICU

if (InitializeICU::initICUSuccessful())

{

return u_strcasecmp((const UChar*)s1.getChar16Data(),

return u_strcasecmp(

(const UChar*)s2.getChar16Data(),

str1._rep->data, str2._rep->data, U_FOLD_CASE_DEFAULT);

U_FOLD_CASE_DEFAULT);

}

#endif

const Char16* _s1 = s1.getChar16Data();

const Char16* _s2 = s2.getChar16Data();

while (*_s1 && *_s2)

#endif /* PEGASUS_HAS_ICU */

{

int r;

if (*_s1 <= PEGASUS_MAX_PRINTABLE_CHAR &&

const Uint16* s1 = str1._rep->data;

*_s2 <= PEGASUS_MAX_PRINTABLE_CHAR)

const Uint16* s2 = str2._rep->data;

{

r = tolower(*_s1++) - tolower(*_s2++);

while (*s1 && *s2)

}

else

{

r = *_s1++ - *_s2++;

int r = _toLower(*s1++) - _toLower(*s2++);

}

if (r)

return r;

}

if (*_s2)

if (*s2)

return -1;

else if (*_s1)

else if (*s1)

return 1;

return 0;

}

Boolean String::equal(const String& str1, const String& str2)

Boolean StringEqualNoCase(const String& s1, const String& s2)

{

return String::compare(str1, str2) == 0;

#ifdef PEGASUS_HAS_ICU

}

return String::compareNoCase(s1, s2) == 0;

Boolean String::equalNoCase(const String& str1, const String& str2)

#else /* PEGASUS_HAS_ICU */

// The following employs loop unrolling for efficiency. Please do not

// eliminate.

Uint16* p = (Uint16*)s1.getChar16Data();

Uint16* q = (Uint16*)s2.getChar16Data();

Uint32 n = s2.size();

while (n >= 8)

{

if (((p[0] - q[0]) && (_toUpper(p[0]) - _toUpper(q[0]))) ||

((p[1] - q[1]) && (_toUpper(p[1]) - _toUpper(q[1]))) ||

((p[2] - q[2]) && (_toUpper(p[2]) - _toUpper(q[2]))) ||

((p[3] - q[3]) && (_toUpper(p[3]) - _toUpper(q[3]))) ||

((p[4] - q[4]) && (_toUpper(p[4]) - _toUpper(q[4]))) ||

((p[5] - q[5]) && (_toUpper(p[5]) - _toUpper(q[5]))) ||

((p[6] - q[6]) && (_toUpper(p[6]) - _toUpper(q[6]))) ||

((p[7] - q[7]) && (_toUpper(p[7]) - _toUpper(q[7]))))

{

#ifdef PEGASUS_HAS_ICU

return compareNoCase(str1, str2) == 0;

#else

if (str1.size() != str2.size())

return false;

}

const Char16* p = str1.getChar16Data();

n -= 8;

const Char16* q = str2.getChar16Data();

p += 8;

q += 8;

Uint32 n = str1.size();

}

while (n--)

while (n >= 4)

{

if (*p <= PEGASUS_MAX_PRINTABLE_CHAR &&

if (((p[0] - q[0]) && (_toUpper(p[0]) - _toUpper(q[0]))) ||

*q <= PEGASUS_MAX_PRINTABLE_CHAR)

((p[1] - q[1]) && (_toUpper(p[1]) - _toUpper(q[1]))) ||

((p[2] - q[2]) && (_toUpper(p[2]) - _toUpper(q[2]))) ||

((p[3] - q[3]) && (_toUpper(p[3]) - _toUpper(q[3]))))

{

if (tolower(*p++) != tolower(*q++))

return false;

}

else if (*p++ != *q++)

n -= 4;

p += 4;

q += 4;

}

while (n--)

{

if (((p[0] - q[0]) && (_toUpper(p[0]) - _toUpper(q[0]))))

return false;

p++;

q++;

}

return true;

#endif

}

#endif /* PEGASUS_HAS_ICU */

}

CString String::getCString() const

Boolean String::equalNoCase(const String& s1, const char* s2)

{

Uint32 n = 3*size() + 1;

_checkNullPointer(s2);

char* str = new char[n];

const Char16* msg16 = getChar16Data();

#if defined(PEGASUS_HAS_ICU)

const Uint16 *strsrc = (Uint16 *)msg16;

return String::equalNoCase(s1, String(s2));

Uint16 *endsrc = (Uint16 *)&msg16[size()+1];

Uint8 *strtgt = (Uint8 *)str;

#elif defined(PEGASUS_STRING_NO_UTF8)

Uint8 *endtgt = (Uint8 *)&str[n];

UTF16toUTF8 (&strsrc,

const Uint16* p1 = (Uint16*)s1._rep->data;

endsrc,

const char* p2 = s2;

&strtgt,

size_t n = s1._rep->size;

endtgt);

char* str1 = new char[strlen(str)+1];

while (n--)

strcpy(str1,str);

{

delete [] str;

if (!*p2)

return false;

return CString(str1);

if (_toUpper(*p1++) != _toUpperTable[int(*p2++)])

return false;

}

#if 0

if (*p2)

// ATTN-RK-P3-20020603: This code is not completely correct

return false;

// Wildcard String matching function that may be useful in the future

// The following code was provided by Bob Blair.

/* _StringMatch Match input MatchString against a GLOB style pattern

Note that MatchChar is the char type so that this source

in portable to different string types. This is an internal function

Results: The return value is 1 if string matches pattern, and

return true;

0 otherwise. The matching operation permits the following

special characters in the pattern: *?\[] (see the manual

entry for details on what these mean).

#else /* PEGASUS_HAS_ICU */

Side effects: None.

// ATTN: optimize this!

return String::equalNoCase(s1, String(s2));

/* MatchChar defined as a separate entity because this function source used

#endif /* PEGASUS_HAS_ICU */

elsewhere was an unsigned char *. Here we use Uint16 to maintain 16 bit

size.

typedef Uint16 MatchChar;

inline Uint16 _ToLower(Uint16 ch)

{

// ICU_TODO: If ICU is available we should do this the correct way.

return ch <= PEGASUS_MAX_PRINTABLE_CHAR ? tolower(char(ch)) : ch;

}

inline Boolean _Equal(MatchChar ch1, MatchChar ch2, int nocase)

Boolean String::equal(const String& s1, const String& s2)

{

// ICU_TODO: If ICU is available we should do this the correct way.

if (nocase)

return _ToLower(ch1) == _ToLower(ch2);

return ch1 == ch2;

}

static const MatchChar *

_matchrange(const MatchChar *range, MatchChar c, int nocase)

{

const MatchChar *p = range;

const MatchChar *rstart = range + 1;

const MatchChar *rend = 0;

MatchChar compchar;

for (rend = rstart; *rend && *rend != ']'; rend++);

if (*rend == ']') { // if there is an end to this pattern

for (compchar = *rstart; rstart != rend; rstart++) {

if (_Equal(*rstart, c, nocase))

return ++rend;

if (*rstart == '-') {

rstart++;

if (c >= compchar && c <= *rstart)

return ++rend;

}

return (const MatchChar *)0;

}

static int

_StringMatch(

const MatchChar *testString,

const MatchChar *pattern,

int nocase ) /* Ignore case if this is true */

{

const MatchChar *pat = pattern;

const MatchChar *str = testString;

unsigned int done = 0;

unsigned int res = 0; // the result: 1 == match

while (!done) { // main loop walks through pattern and test string

//cerr << "Comparing <" << *pat << "> and <" << *str << ">" << endl;

if (!*pat) { //end of pattern

done = 1; // we're done

if (!*str) //end of test, too?

res = 1; // then we matched

} else { //Not end of pattern

if (!*str) { // but end of test

done = 1; // We're done

if (*pat == '*') // If pattern openends

res = 1; // then we matched

} else { //Not end of test

if (*pat == '*') { //Ambiguuity found

if (!*++pat) { //and it ends pattern

done = 1; // then we're done

res = 1; // and match

} else { //if it doesn't end

while (!done) { // until we're done

if (_StringMatch(str, pat, nocase)) { // we recurse

done = 1; //if it recurses true

res = 1; // we done and match

} else { //it recurses false

if (!*str) // see if test is done

done = 1; // yes: we done

else // not done:

str++; // keep testing

} // end test on recursive call

} // end looping on recursive calls

} // end logic when pattern is ambiguous

} else { //pattern not ambiguus

if (*pat == '?') { //pattern is 'any'

pat++, str++; // so move along

} else if (*pat == '[') { //see if it's a range

pat = _matchrange(pat, *str, nocase); // and is a match

if (!pat) { //It is not a match

done = 1; // we're done

res = 0; // no match

} else { //Range matches

str++, pat++; // keep going

}

} else { // only case left is individual characters

if (!_Equal(*pat++, *str++, nocase)) // if they don't match

done = 1; // bail.

}

} // end ("pattern is not ambiguous (*)" logic

} // end logic when pattern and string still have data

} // end logic when pattern still has data

} // end main loop

return res;

}

/** match matches a string against a GLOB style pattern.

Return trues if the String parameter matches the pattern. C-Shell style

glob matching is used.

@param str String to be matched against the pattern

@param pattern Pattern to use in the match

@return Boolean true if str matches pattern

The pattern definition is as follows:

<pre>

* Matches any number of any characters

? Match exactly one character

[chars] Match any character in chars

[chara-charb] Match any character in the range between chara and charb

</pre>

The literal characters *, ?, [, ] can be included in a string by

escaping them with backslash "\". Ranges of characters can be concatenated.

<pre>

examples:

Boolean result = String::match("This is a test", "*is*");

Boolean works = String::match("abcdef123", "*[0-9]");

</pre>

Boolean String::match(const String& str, const String& pattern)

{

return _StringMatch(

return s1._rep->size == s2._rep->size && memcmp(s1._rep->data,

(Uint16*)str.getChar16Data(), (Uint16*)pattern.getChar16Data(), 0) != 0;

s2._rep->data, s1._rep->size * sizeof(Uint16)) == 0;

}

/** matchNoCase Matches a String against a GLOB style pattern independent

Boolean String::equal(const String& s1, const char* s2)

of case.

Returns true if the str parameter matches the pattern. C-Shell style

glob matching is used. Ignore case in all comparisons. Case is

ignored in the match.

@parm str String containing the string to be matched\

@parm pattern GLOB style patterh to use in the match.

@return Boolean true if str matches patterh

@see match

Boolean String::matchNoCase(const String& str, const String& pattern)

{

return _StringMatch(

#ifdef PEGASUS_STRING_NO_UTF8

(Uint16*)str.getChar16Data(), (Uint16*)pattern.getChar16Data(), 1) != 0;

}

#endif

_checkNullPointer(s2);

///////////////////////////////////////////////////////////////////////////////

const Uint16* p = (Uint16*)s1._rep->data;

const char* q = s2;

// String-related functions

///////////////////////////////////////////////////////////////////////////////

Boolean operator==(const String& str1, const String& str2)

while (*p && *q)

{

return String::equal(str1, str2);

if (*p++ != Uint16(*q++))

return false;

}

Boolean operator==(const String& str1, const char* str2)

return !(*p || *q);

{

return String::equal(str1, str2);

}

Boolean operator==(const char* str1, const String& str2)

#else /* PEGASUS_STRING_NO_UTF8 */

{

return String::equal(str1, str2);

}

Boolean operator!=(const String& str1, const String& str2)

return String::equal(s1, String(s2));

{

return !String::equal(str1, str2);

#endif /* PEGASUS_STRING_NO_UTF8 */

}

PEGASUS_STD(ostream)& operator<<(PEGASUS_STD(ostream)& os, const String& str)

{

#if defined(PEGASUS_OS_OS400)

CString cstr = str.getCString();

const char* utf8str = cstr;

os << utf8str;

return os;

#else

#if defined(PEGASUS_HAS_ICU)

#elif defined(PEGASUS_HAS_ICU)

if (InitializeICU::initICUSuccessful())

{

char *buf = NULL;

const int size = str.size() * 6;

UnicodeString UniStr((const UChar *)str.getChar16Data(), (int32_t)str.size());

UnicodeString UniStr(

(const UChar *)str.getChar16Data(), (int32_t)str.size());

Uint32 bufsize = UniStr.extract(0,size,buf);

buf = new char[bufsize+1];

UniStr.extract(0,bufsize,buf);

os << buf;

os.flush();

delete [] buf;

return os;

}

else

#endif // End of PEGASUS_HAS_ICU #else leg.

#endif // PEGASUS_HAS_ICU

{

for (Uint32 i = 0, n = str.size(); i < n; i++)

{

Uint16 code = str[i];

if (code > 0 && code <= PEGASUS_MAX_PRINTABLE_CHAR)

if (code > 0 && !(code & 0xFF00))

{

os << char(code);

}

else

{

// Print in hex format:

Line 1024

Line 1328

os << buffer;

}

return os;

#endif // PEGASUS_OS_OS400

}

String operator+(const String& str1, const String& str2)

void StringAppendCharAux(StringRep*& _rep)

{

return String(str1).append(str2);

StringRep* tmp;

}

Boolean operator<(const String& str1, const String& str2)

if (_rep->cap)

{

return String::compare(str1, str2) < 0;

tmp = StringRep::alloc(2 * _rep->cap);

tmp->size = _rep->size;

_copy(tmp->data, _rep->data, _rep->size);

}

else

{

tmp = StringRep::alloc(8);

tmp->size = 0;

}

StringRep::unref(_rep);

_rep = tmp;

}

PEGASUS_NAMESPACE_END

================================================================================

String optimizations:

1. Added mechanism allowing certain functions to be inlined only when

used by internal Pegasus modules. External modules (i.e., providers)

link to a non-inline version, which allows for binary compatibility.

2. Implemented copy-on-write with atomic increment/decrement. This

yieled a 10% improvement for the 'gc' benchmark and a 11% improvment

for the 'ni1000' benchmark.

3. Employed loop unrolling in several places. For example, see:

static Uint16* _find(const Uint16* s, size_t n, Uint16 c);

4. Used the "empty-rep" optimization (described in whitepaper from the

GCC Developers Summit). This reduced default construction to a simple

pointer assignment.

inline String::String() : _rep(&_emptyRep) { }

5. Implemented Uint16 versions of toupper() and tolower() using tables.

For example:

Boolean operator<=(const String& str1, const String& str2)

static const char _upper[] =

{

return String::compare(str1, str2) <= 0;

0,1,2,...255

};

inline Uint16 _toUpper(Uint16 x)

{

return (x & 0xFF00) ? x : _upper[x];

}

Boolean operator>(const String& str1, const String& str2)

This outperforms the system implementation by avoiding an anding

operation.

6. Implemented char* version of the following member functions to

eliminate unecessary creation of anonymous string objects

(temporaries).

String(const String& s1, const char* s2);

String(const char* s1, const String& s2);

String& String::operator=(const char* str);

Uint32 String::find(const char* s) const;

bool String::equal(const String& s1, const char* s2);

static int String::compare(const String& s1, const char* s2);

String& String::append(const char* str);

String& String::append(const char* str, Uint32 size);

static bool String::equalNoCase(const String& s1, const char* s2);

String& operator=(const char* str)

String& String::assign(const char* str)

String& String::append(const char* str)

Boolean operator==(const String& s1, const char* s2)

Boolean operator==(const char* s1, const String& s2)

Boolean operator!=(const String& s1, const char* s2)

Boolean operator!=(const char* s1, const String& s2)

Boolean operator<(const String& s1, const char* s2)

Boolean operator<(const char* s1, const String& s2)

Boolean operator>(const String& s1, const char* s2)

Boolean operator>(const char* s1, const String& s2)

Boolean operator<=(const String& s1, const char* s2)

Boolean operator<=(const char* s1, const String& s2)

Boolean operator>=(const String& s1, const char* s2)

Boolean operator>=(const char* s1, const String& s2)

String operator+(const String& s1, const char* s2)

String operator+(const char* s1, const String& s2)

7. Optimized _roundUpToPow2(), used in rounding the capacity to the next

power of two (algorithm from the book "Hacker's Delight").

static Uint32 _roundUpToPow2(Uint32 x)

{

return String::compare(str1, str2) > 0;

if (x < 8)

return 8;

x--;

x |= (x >> 1);

x |= (x >> 2);

x |= (x >> 4);

x |= (x >> 8);

x |= (x >> 16);

x++;

return x;

}

Boolean operator>=(const String& str1, const String& str2)

8. Implemented "concatenating constructors" to eliminate temporaries

created by operator+(). This scheme employs the "return-value

optimization" described by Stan Lippman.

inline String operator+(const String& s1, const String& s2)

{

return String::compare(str1, str2) >= 0;

return String(s1, s2, 0);

}

PEGASUS_NAMESPACE_END

9. Experimented to find the optimial initial size for a short string.

Eight seems to offer the best tradeoff between space and time.

10. Inlined all members of the Char16 class.

11. Used Uint16 internally in the String class. This showed no improvememnt

since Char16 was already fully inlined and was essentially reduced to

Uint16 in any case.

12. Implemented conditional logic (#if) allowing error checking logic to

be excluded to better performance. Examples include bounds checking

and null-pointer checking.

13. Used memcpy() and memcmp() where possible. These are implemented using

the rep family of intructions under Intel and are much faster.

14. Used loop unrolling, jump-tables, and short-circuiting to reduce UTF8

copy routine overhead.

15. Added ASCII7 form of the constructor and assign().

String s("hello world", String::ASCII7);

s.assignASCII7("hello world");

This avoids slower UTF8 processing when not needed.

================================================================================

TO-DO:

(+) [DONE] Use PEGASUS_USE_EXPERIMENTAL_INTERFACES

(+) [DONE] Submit BUG-2754 (Windows buffer limit).

(+) [DONE] Eliminate char versions of find() and append().

(+) [DONE] Remove PEGASUS_MAX_PRINTABLE_CHARACTER from Config.h

(+) [DONE] Change _next_pow_2() to _roundUpToPow2().

(+) [DONE] Change '99' to '2' in StringRep constructor (comment as well).

(+) [DONE] Comment StringRep allocation layout.

(+) [DONE] Conceal private inline functions.

(+) [DONE] Shorten inclusion of StringInline.h in String.h.

(+) [DONE] Change USE_INTERNAL_INLINE TO DISABLE_INTERNAL_INLINE or get

rid of altogether.

(+) [DONE] useCamelNotationOnAllFunctionNames.

(+) [DONE] Check for overlow condition in StringRep::alloc().

(+) [DONE] Remove tabs (used vim ":set expandtab" and ":retab").

(+) [DONE] Fix throw-related memory leak.

(+) [DONE] Look at PEP223 for coding security guidelines.

(+) [DONE] Use old AtomicInt for now (new AtomicInt part of bug #4250).

(+) [DONE] Removed appendASCII() and the ASCII form of the constructor.

(+) DOC++ String.h - will open new bug?

(+) Added PEGASUS_DISABLE_INTERNAL_INLINES macro (to permit suppression

on certain platforms).

================================================================================

Legend:

Removed from v.1.111
changed lines
	Added in v.1.111.6.16

No CVS admin address has been configured