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 1.20 and 1.111.2.15

version 1.20, 2001/05/24 00:48:37

version 1.111.2.15, 2005/10/04 22:25:15

Line 1

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

//%2005////////////////////////////////////////////////////////////////////////

// Company, L.P.; IBM Corp.; The Open Group; Tivoli Systems.

// IBM Corp.; EMC Corporation, The Open Group.

// IBM Corp.; EMC Corporation; VERITAS Software Corporation; The Open Group.

// EMC Corporation; VERITAS Software Corporation; The Open Group.

// Permission is hereby granted, free of charge, to any person obtaining a copy

// of this software and associated documentation files (the "Software"), to

// deal in the Software without restriction, including without limitation the

// rights to use, copy, modify, merge, publish, distribute, sublicense, and/or

// sell copies of the Software, and to permit persons to whom the Software is

// furnished to do so, subject to the following conditions:

// Permission is hereby granted, free of charge, to any person obtaining a

// THE ABOVE COPYRIGHT NOTICE AND THIS PERMISSION NOTICE SHALL BE INCLUDED IN

// copy of this software and associated documentation files (the "Software"),

// ALL COPIES OR SUBSTANTIAL PORTIONS OF THE SOFTWARE. THE SOFTWARE IS PROVIDED

// to deal in the Software without restriction, including without limitation

// "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT

// the rights to use, copy, modify, merge, publish, distribute, sublicense,

// LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR

// and/or sell copies of the Software, and to permit persons to whom the

// PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT

// Software is furnished to do so, subject to the following conditions:

// HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN

// ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION

// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR

// WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,

// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL

// THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER

// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING

// FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER

// DEALINGS IN THE SOFTWARE.

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

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

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

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

#ifndef PEGASUS_USE_INTERNAL_INLINES

# define PEGASUS_USE_INTERNAL_INLINES

#endif

#include <cctype>

#include "String.h"

#include "Exception.h"

#include <cassert>

#include "String.h"

#include "InternalException.h"

#include <iostream>

#include "CommonUTF.h"

#include "MessageLoader.h"

#include "CharSet.h"

#ifdef PEGASUS_HAS_ICU

#include <unicode/ustring.h>

#include <unicode/uchar.h>

#endif

PEGASUS_NAMESPACE_BEGIN

#define PEGASUS_ARRAY_T String

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

#include <Pegasus/Common/ArrayImpl.h>

#undef PEGASUS_ARRAY_T

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

// PEGASUS_STRING_NO_THROW -- suppresses throwing of exceptions

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

// PEGASUS_USE_INTERNAL_INLINES -- enables internal inlining feature.

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

const String String::EMPTY;

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

// File-scope definitions:

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

inline Uint32 StrLen(const char* str)

// Converts 16-bit characters to upper case.

inline Uint16 _to_upper(Uint16 x)

{

if (!str)

return (x & 0xFF00) ? x : CharSet::to_upper(x);

throw NullPointer();

}

return strlen(str);

// Converts 16-bit characters to lower case.

inline Uint16 _to_lower(Uint16 x)

{

return (x & 0xFF00) ? x : CharSet::to_lower(x);

}

inline Uint32 StrLen(const Char16* str)

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

static Uint32 _next_pow_2(Uint32 x)

{

if (!str)

if (x < 8)

throw NullPointer();

return 8;

Uint32 n = 0;

x--;

x |= (x >> 1);

x |= (x >> 2);

x |= (x >> 4);

x |= (x >> 8);

x |= (x >> 16);

x++;

while (*str++)

return x;

n++;

}

return n;

template<class P, class Q>

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

{

// Use loop unrolling.

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;

}

String::String()

while (n >= 4)

{

_rep.append('\0');

p[0] = q[0];

p[1] = q[1];

p[2] = q[2];

p[3] = q[3];

p += 4;

q += 4;

n -= 4;

}

String::String(const String& x) : _rep(x._rep)

while (n--)

*p++ = *q++;

}

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

{

while (n >= 4)

{

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;

}

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

if (n)

{

_rep.append('\0');

if (*s == c)

append(x.getData(), n);

return (Uint16*)s;

s++;

n--;

}

String::String(const Char16* x) : _rep(x, StrLen(x) + 1)

if (n)

{

if (*s == c)

return (Uint16*)s;

s++;

n--;

}

if (n && *s == c)

return (Uint16*)s;

// Not found!

return 0;

}

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

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

{

while (*s1 && *s2)

{

assign(x, n);

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

if (r)

return r;

}

String::String(const char* str)

if (*s2)

return -1;

else if (*s1)

return 1;

return 0;

}

static int _compare_no_utf8(const Uint16* s1, const char* s2)

{

Uint32 n = ::strlen(str) + 1;

Uint16 c1;

reserve(n);

Uint16 c2;

while (n--)

_rep.append(*str++);

{

c1 = *s1++;

c2 = *s2++;

if (c1 == 0)

return c1 - c2;

}

while (c1 == c2);

return c1 - c2;

}

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

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

{

Uint32 n = _min(strlen(str), n_);

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

reserve(n + 1);

while (n--)

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

_rep.append(*str++);

;

_rep.append('\0');

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

}

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

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

{

_rep.clear();

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

_rep.append(x, StrLen(x) + 1);

return *this;

}

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

void String_throw_out_of_bounds()

{

_rep.clear();

throw IndexOutOfBoundsException();

Uint32 m = _min(StrLen(str), n);

_rep.append(str, m);

_rep.append('\0');

return *this;

}

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

#ifdef PEGASUS_STRING_NO_THROW

# define _check_null_pointer(ARG) /* empty */

#else

template<class T>

inline void _check_null_pointer(const T* ptr)

{

_rep.clear();

if (!ptr)

Uint32 n = strlen(x);

throw NullPointer();

_rep.reserve(n + 1);

}

#endif

while (n--)

static size_t _copy_from_utf8(Uint16* dest, const char* src, size_t n)

_rep.append(*x++);

{

Uint16* p = dest;

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

_rep.append('\0');

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

// Use loop-unrolling.

return *this;

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;

}

String& String::assign(const char* x, Uint32 n_)

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

{

_rep.clear();

p[0] = q[0];

p[1] = q[1];

p[2] = q[2];

p[3] = q[3];

p += 4;

q += 4;

n -= 4;

}

Uint32 n = _min(strlen(x), n_);

switch (n)

_rep.reserve(n + 1);

{

case 0:

return p - dest;

case 1:

if (q[0] < 128)

{

p[0] = q[0];

return p + 1 - dest;

}

break;

case 2:

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

{

p[0] = q[0];

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;

}

while (n--)

// Process remaining characters.

_rep.append(*x++);

_rep.append('\0');

while (n)

{

// Optimize for 7-bit ASCII case.

return *this;

if (*q < 128)

{

*p++ = *q++;

n--;

}

else

{

Uint8 c = UTF_8_COUNT_TRAIL_BYTES(*q) + 1;

char* String::allocateCString(Uint32 extraBytes, Boolean noThrow) const

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

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

{

Uint32 n = size() + 1;

MessageLoaderParms parms("Common.String.BAD_UTF8",

char* str = new char[n + extraBytes];

"The byte sequence starting at index $0 "

char* p = str;

"is not valid UTF-8 encoding.",

const Char16* q = getData();

q - (const Uint8*)src);

throw Exception(parms);

}

for (Uint32 i = 0; i < n; i++)

n -= c;

}

return p - dest;

}

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

// terminator).

static inline size_t _copy_to_utf8(char* dest, const Uint16* src, size_t n)

{

Uint16 c = *q++;

const Uint16* q = src;

*p++ = char(c);

Uint8* p = (Uint8*)dest;

if ((c & 0xff00) && !noThrow)

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

throw TruncatedCharacter();

{

p[0] = q[0];

p[1] = q[1];

p[2] = q[2];

p[3] = q[3];

p += 4;

q += 4;

n -= 4;

}

return str;

switch (n)

{

case 0:

return p - (Uint8*)dest;

case 1:

if (q[0] < 128)

{

p[0] = q[0];

return p + 1 - (Uint8*)dest;

}

break;

case 2:

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

{

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;

}

void String::appendToCString(

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

char* str,

Uint32 length,

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

Boolean noThrow) const

return p - (Uint8*)dest;

}

static inline size_t _convert(Uint16* p, const char* q, size_t n)

{

if (!str)

#ifdef PEGASUS_STRING_NO_UTF8

throw NullPointer();

_copy(p, q, n);

return n;

#else

return _copy_from_utf8(p, q, n);

#endif

}

Uint32 n = _min(size(), length);

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

// class CString

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

char* p = str + strlen(str);

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

const Char16* q = getData();

{

if (cstr._rep)

{

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

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

memcpy(_rep, cstr._rep, n);

}

for (Uint32 i = 0; i < n; i++)

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

{

if (&cstr != this)

{

Uint16 c = *q++;

if (_rep)

*p++ = char(c);

{

operator delete(_rep);

_rep = 0;

}

if ((c & 0xff00) && !noThrow)

if (cstr._rep)

throw TruncatedCharacter();

{

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

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

memcpy(_rep, cstr._rep, n);

}

*p = '\0';

return *this;

}

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

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

// class StringRep

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

StringRep StringRep::_empty_rep;

inline StringRep* StringRep::alloc(size_t cap)

{

if (i > size())

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

ThrowOutOfBounds();

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

rep->cap = cap;

Atomic_create(&rep->refs, 1);

return _rep[i];

return rep;

}

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

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

{

if (cap > rep->cap || Atomic_get(&rep->refs) != 1)

{

if (i > size())

size_t n = _next_pow_2(cap);

ThrowOutOfBounds();

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

new_rep->size = rep->size;

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

StringRep::unref(rep);

rep = new_rep;

}

return _rep[i];

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

{

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

rep->size = size;

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

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

return rep;

}

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

StringRep* StringRep::copy_on_write(StringRep* rep)

{

Uint32 m = _min(StrLen(str), n);

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

_rep.reserve(_rep.size() + m);

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

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

_rep.append(str, m);

new_rep->size = rep->size;

_rep.append('\0');

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

return *this;

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

StringRep::unref(rep);

return new_rep;

}

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

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

{

if (size == PEG_NOT_FOUND)

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

size = this->size() - pos;

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

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

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

return rep;

ThrowOutOfBounds();

}

if (size)

StringRep* StringRep::createASCII7(const char* data, size_t size)

_rep.remove(pos, size);

{

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

_copy((Uint16*)rep->data, data, size);

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

return rep;

}

int String::compare(const Char16* s1, const Char16* s2, Uint32 n)

Uint32 StringRep::length(const Uint16* str)

{

while (n--)

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

const Uint16* end = (Uint16*)str;

while (*end++)

;

return end - str - 1;

}

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

// class String

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

const String String::EMPTY;

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

{

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

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

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

}

if (r)

String::String(const Char16* str)

return r;

{

_check_null_pointer(str);

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

}

return 0;

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

{

_check_null_pointer(str);

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

}

int String::compareNoCase(const char* s1, const char* s2, Uint32 n)

String::String(const char* str)

{

while (n--)

_check_null_pointer(str);

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

}

#ifdef PEGASUS_USE_STRING_EXTENSIONS

String::String(const char* str, String::ASCII7Tag tag)

{

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

_check_null_pointer(str);

_rep = StringRep::createASCII7(str, strlen(str));

}

#endif /* PEGASUS_USE_STRING_EXTENSIONS */

if (r)

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

return r;

{

_check_null_pointer(str);

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

}

return 0;

#ifdef PEGASUS_USE_STRING_EXTENSIONS

String::String(const char* str, size_t n, String::ASCII7Tag tag)

{

_check_null_pointer(str);

_rep = StringRep::createASCII7(str, n);

}

#endif /* PEGASUS_USE_STRING_EXTENSIONS */

Boolean String::equal(const String& x, const String& y)

#ifdef PEGASUS_USE_STRING_EXTENSIONS

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

{

if (x.size() != y.size())

size_t n1 = s1._rep->size;

return false;

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';

}

#endif /* PEGASUS_USE_STRING_EXTENSIONS */

return String::compare(x.getData(), y.getData(), x.size()) == 0;

#ifdef PEGASUS_USE_STRING_EXTENSIONS

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

{

_check_null_pointer(s2);

size_t n1 = s1._rep->size;

size_t n2 = strlen(s2);

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

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

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

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

}

#endif /* PEGASUS_USE_STRING_EXTENSIONS */

Boolean String::equal(const String& x, const Char16* y)

#ifdef PEGASUS_USE_STRING_EXTENSIONS

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

{

if (x.size() != StrLen(y))

_check_null_pointer(s1);

return false;

size_t n1 = strlen(s1);

size_t n2 = s2._rep->size;

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

_rep->size = n2 + _convert((Uint16*)_rep->data, s1, n1);

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

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

}

#endif /* PEGASUS_USE_STRING_EXTENSIONS */

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

{

if (_rep != str._rep)

{

StringRep::unref(_rep);

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

}

return String::compare(x.getData(), y, x.size()) == 0;

return *this;

}

Boolean String::equal(const Char16* x, const String& y)

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

{

_check_null_pointer(str);

if (n > _rep->cap || Atomic_get(&_rep->refs) != 1)

{

return equal(y, x);

StringRep::unref(_rep);

_rep = StringRep::alloc(n);

}

Boolean String::equal(const String& x, const char* y)

_rep->size = n;

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

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

return *this;

}

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

{

_check_null_pointer(str);

if (n > _rep->cap || Atomic_get(&_rep->refs) != 1)

{

return equal(x, String(y));

StringRep::unref(_rep);

_rep = StringRep::alloc(n);

}

Boolean String::equal(const char* x, const String& y)

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

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

return *this;

}

#ifdef PEGASUS_USE_STRING_EXTENSIONS

String& String::assignASCII7(const char* str, Uint32 n)

{

_check_null_pointer(str);

if (n > _rep->cap || Atomic_get(&_rep->refs) != 1)

{

return equal(String(x), y);

StringRep::unref(_rep);

_rep = StringRep::alloc(n);

}

Boolean String::equalNoCase(const String& x, const String& y)

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

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

return *this;

}

#endif /* PEGASUS_USE_STRING_EXTENSIONS */

void String::clear()

{

if (x.size() != y.size())

if (_rep->size)

return false;

{

if (Atomic_get(&_rep->refs) == 1)

_rep->size = 0;

else

{

StringRep::unref(_rep);

_rep = &StringRep::_empty_rep;

}

const Char16* p = x.getData();

void String::reserveCapacity(Uint32 cap)

const Char16* q = y.getData();

{

_reserve(_rep, cap);

}

Uint32 n = x.size();

CString String::getCString() const

{

#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 = _copy_to_utf8(str, _rep->data, _rep->size);

str[size] = '\0';

return CString(str);

#endif

}

while (n--)

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

{

if (*p <= 127 && *q <= 127)

_check_null_pointer(str);

size_t old_size = _rep->size;

size_t new_size = old_size + n;

_reserve(_rep, new_size);

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

_rep->size = new_size;

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

return *this;

}

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

{

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

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

return false;

}

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

return false;

#ifdef PEGASUS_USE_STRING_EXTENSIONS

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

{

_check_null_pointer(str);

size_t old_size = _rep->size;

size_t cap = old_size + size;

_reserve(_rep, cap);

_rep->size += _convert((Uint16*)_rep->data + old_size, str, size);

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

return *this;

}

#endif /* PEGASUS_USE_STRING_EXTENSIONS */

return true;

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

{

if (n == PEG_NOT_FOUND)

n = _rep->size - index;

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

if (Atomic_get(&_rep->refs) != 1)

_rep = StringRep::copy_on_write(_rep);

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

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

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 pos, Uint32 length) const

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

{

if (pos < size())

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

// backwards compatibility.

if (index < _rep->size)

{

if (length == PEG_NOT_FOUND)

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

length = size() - pos;

n = _rep->size - index;

return String(getData() + pos, length);

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

}

else

return String();

}

Uint32 String::find(Char16 c) const

{

const Char16* first = getData();

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(const String& s) const

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

{

const Char16* pSubStr = s.getData();

_check_bounds(index, _rep->size);

const Char16* pStr = getData();

Uint32 subStrLen = s.size();

if (index >= _rep->size)

Uint32 strLen = size();

return PEG_NOT_FOUND;

// loop to find first char match

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

Uint32 loc = 0;

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

if (p)

{

return p - _rep->data;

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

{

return PEG_NOT_FOUND;

// point to substr 2nd char

const Char16* p = pSubStr + 1;

// Test remaining chars for equal

Uint32 i = 1;

for (; i < subStrLen; i++)

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

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

if (i == subStrLen)

return loc;

}

Uint32 String::_find_aux(const Char16* s, Uint32 n) const

{

_check_null_pointer(s);

const Uint16* data = _rep->data;

size_t rem = _rep->size;

while (n <= rem)

{

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

if (!p)

break;

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

return p - _rep->data;

p++;

rem -= p - data;

data = p;

}

return PEG_NOT_FOUND;

}

// ATTN:KS 5 apr 2000 Need to add the Char16* version.

#ifdef PEGASUS_USE_STRING_EXTENSIONS

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

{

_check_null_pointer(s);

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

// called.

return find(String(s));

}

#endif /* PEGASUS_USE_STRING_EXTENSIONS */

Uint32 String::reverseFind(Char16 c) const

{

const Char16* first = getData();

Uint16 x = c;

const Char16* last = getData() + 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 391

Line 867

void String::toLower()

{

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

#ifdef PEGASUS_HAS_ICU

if (InitializeICU::initICUSuccessful())

{

if (Atomic_get(&_rep->refs) != 1)

_rep = StringRep::copy_on_write(_rep);

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

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

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

// calling ICU.

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

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

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

UErrorCode err = U_ZERO_ERROR;

int32_t new_size = u_strToLower(

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

err = U_ZERO_ERROR;

//// Reserve enough space for the result.

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

_reserve(_rep, new_size);

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

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

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

_rep->size = new_size;

return;

}

#endif /* PEGASUS_HAS_ICU */

if (Atomic_get(&_rep->refs) != 1)

_rep = StringRep::copy_on_write(_rep);

Uint16* p = _rep->data;

size_t n = _rep->size;

for (; n--; p++)

{

if (!(*p & 0xFF00))

*p = _to_lower(*p);

}

void String::toUpper()

{

#ifdef PEGASUS_HAS_ICU

if (InitializeICU::initICUSuccessful())

{

if (*p <= 127)

if (Atomic_get(&_rep->refs) != 1)

*p = tolower(*p);

_rep = StringRep::copy_on_write(_rep);

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

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

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

// calling ICU.

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

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

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

UErrorCode err = U_ZERO_ERROR;

int32_t new_size = u_strToUpper(

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

err = U_ZERO_ERROR;

//// Reserve enough space for the result.

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

_reserve(_rep, new_size);

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

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

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

_rep->size = new_size;

return;

}

#endif /* PEGASUS_HAS_ICU */

if (Atomic_get(&_rep->refs) != 1)

_rep = StringRep::copy_on_write(_rep);

Uint16* p = _rep->data;

size_t n = _rep->size;

for (; n--; p++)

*p = _to_upper(*p);

}

void String::translate(Char16 fromChar, Char16 toChar)

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

{

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

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

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

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

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

}

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

{

if (*p == fromChar)

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

*p = toChar;

}

#ifdef PEGASUS_USE_STRING_EXTENSIONS

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

{

_check_null_pointer(s2);

#ifdef PEGASUS_STRING_NO_UTF8

return _compare_no_utf8(s1._rep->data, s2);

#else

// ATTN: optimize this!

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

#endif

}

#endif /* PEGASUS_USE_STRING_EXTENSIONS */

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

{

#ifdef PEGASUS_HAS_ICU

int String::compare(const Char16* s1, const Char16* s2)

if (InitializeICU::initICUSuccessful())

{

return u_strcasecmp(

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

}

#endif /* PEGASUS_HAS_ICU */

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

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

while (*s1 && *s2)

{

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

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

if (r)

return r;

Line 425

Line 1030

return 0;

}

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

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

{

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

#ifdef PEGASUS_HAS_ICU

os << x[i];

return os;

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

}

#else /* PEGASUS_HAS_ICU */

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

Uint16* q = (Uint16*)s2._rep->data;

Uint32 n = s2._rep->size;

void String::toLower(char* str)

while (n >= 8)

{

while (*str)

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

tolower(*str++);

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

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

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

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

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

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

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

{

return false;

}

String ToLower(const String& str)

n -= 8;

p += 8;

q += 8;

}

while (n >= 4)

{

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

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

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

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

{

String tmp(str);

return false;

}

n -= 4;

p += 4;

q += 4;

}

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

while (n--)

{

Char16 c = tmp[i];

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

return false;

if (c <= 127)

p++;

tmp[i] = tolower(c);

q++;

}

return tmp;

return true;

#endif /* PEGASUS_HAS_ICU */

}

int CompareNoCase(const char* s1, const char* s2)

#ifdef PEGASUS_USE_STRING_EXTENSIONS

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

{

while (*s1 && *s2)

_check_null_pointer(s2);

#if defined(PEGASUS_HAS_ICU)

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

#elif defined(PEGASUS_STRING_NO_UTF8)

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

const char* p2 = s2;

size_t n = s1._rep->size;

while (n--)

{

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

if (!*p2)

return false;

if (r)

if (_to_upper(*p1++) != CharSet::to_upper(int(*p2++)))

return r;

return false;

}

if (*s2)

return true;

return -1;

else if (*s1)

return 1;

return 0;

#else /* PEGASUS_HAS_ICU */

// ATTN: optimize this!

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

#endif /* PEGASUS_HAS_ICU */

}

#endif /* PEGASUS_USE_STRING_EXTENSIONS */

Boolean GetLine(PEGASUS_STD(istream)& is, String& line)

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

{

line.clear();

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

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

Boolean gotChar = false;

}

char c;

while (is.get(c))

#ifdef PEGASUS_USE_STRING_EXTENSIONS

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

{

gotChar = true;

#ifdef PEGASUS_STRING_NO_UTF8

if (c == '\n')

_check_null_pointer(s2);

break;

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

const char* q = s2;

line.append(c);

while (*p && *q)

{

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

return false;

}

return gotChar;

return !(*p || *q);

#else /* PEGASUS_STRING_NO_UTF8 */

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

#endif /* PEGASUS_STRING_NO_UTF8 */

}

#endif /* PEGASUS_USE_STRING_EXTENSIONS */

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)

String::~String()

if (InitializeICU::initICUSuccessful())

{

char *buf = NULL;

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

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;

}

String& String::assign(const String& x)

#endif // PEGASUS_HAS_ICU

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

{

_rep = x._rep;

Uint16 code = str[i];

return *this;

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

os << char(code);

else

{

// Print in hex format:

char buffer[8];

sprintf(buffer, "\\x%04X", code);

os << buffer;

}

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

return os;

#endif // PEGASUS_OS_OS400

}

void String::_append_char_aux()

{

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

StringRep* tmp;

return *this;

if (_rep->cap)

{

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;

}

void String::clear()

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(&_empty_rep) { }

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

For example:

static const char _upper[] =

{

0,1,2,...255

};

inline Uint16 _to_upper(Uint16 x)

{

_rep.clear();

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

_rep.append('\0');

}

void String::reserve(Uint32 capacity)

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 _next_pow_2(), used in rounding the capacity to the next

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

static Uint32 _next_pow_2(Uint32 x)

{

_rep.reserve(capacity + 1);

if (x < 8)

return 8;

x--;

x |= (x >> 1);

x |= (x >> 2);

x |= (x >> 4);

x |= (x >> 8);

x |= (x >> 16);

x++;

return x;

}

const Array<String>& EmptyStringArray()

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)

{

static Array<String> tmp;

return String(s1, s2, 0);

return tmp;

}

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.

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

Legend:

Removed from v.1.20
changed lines
	Added in v.1.111.2.15

No CVS admin address has been configured