Clementine-audio-player-Mac.../3rdparty/taglib/toolkit/tstring.h

581 lines
17 KiB
C++

/***************************************************************************
copyright : (C) 2002 - 2008 by Scott Wheeler
email : wheeler@kde.org
***************************************************************************/
/***************************************************************************
* This library is free software; you can redistribute it and/or modify *
* it under the terms of the GNU Lesser General Public License version *
* 2.1 as published by the Free Software Foundation. *
* *
* This library is distributed in the hope that it will be useful, but *
* WITHOUT ANY WARRANTY; without even the implied warranty of *
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU *
* Lesser General Public License for more details. *
* *
* You should have received a copy of the GNU Lesser General Public *
* License along with this library; if not, write to the Free Software *
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA *
* 02110-1301 USA *
* *
* Alternatively, this file is available under the Mozilla Public *
* License Version 1.1. You may obtain a copy of the License at *
* http://www.mozilla.org/MPL/ *
***************************************************************************/
#ifndef TAGLIB_STRING_H
#define TAGLIB_STRING_H
#include "taglib_export.h"
#include "taglib.h"
#include "tbytevector.h"
#include <string>
#include <iostream>
/*!
* \relates TagLib::String
*
* Converts a QString to a TagLib::String without a requirement to link to Qt.
*
* \note consider conversion via usual char-by-char for loop to avoid UTF16->UTF8->UTF16
* conversion happening in the background
*/
#if defined(QT_VERSION) && (QT_VERSION >= 0x040000)
#define QStringToTString(s) TagLib::String(s.toUtf8().data(), TagLib::String::UTF8)
#else
#define QStringToTString(s) TagLib::String(s.utf8().data(), TagLib::String::UTF8)
#endif
/*!
* \relates TagLib::String
*
* Converts a TagLib::String to a QString without a requirement to link to Qt.
*
* \note consider conversion via usual char-by-char for loop to avoid UTF16->UTF8->UTF16
* conversion happening in the background
*
*/
#define TStringToQString(s) QString::fromUtf8(s.toCString(true))
namespace TagLib {
class StringList;
//! A \e wide string class suitable for unicode.
/*!
* This is an implicitly shared \e wide string. For storage it uses
* TagLib::wstring, but as this is an <i>implementation detail</i> this of
* course could change. Strings are stored internally as UTF-16(without BOM/
* CPU byte order)
*
* The use of implicit sharing means that copying a string is cheap, the only
* \e cost comes into play when the copy is modified. Prior to that the string
* just has a pointer to the data of the \e parent String. This also makes
* this class suitable as a function return type.
*
* In addition to adding implicit sharing, this class keeps track of four
* possible encodings, which are the four supported by the ID3v2 standard.
*/
class TAGLIB_EXPORT String
{
public:
#ifndef DO_NOT_DOCUMENT
typedef TagLib::wstring::iterator Iterator;
typedef TagLib::wstring::const_iterator ConstIterator;
#endif
/**
* The four types of string encodings supported by the ID3v2 specification.
* ID3v1 is assumed to be Latin1 and Ogg Vorbis comments use UTF8.
*/
enum Type {
/*!
* IS08859-1, or <i>Latin1</i> encoding. 8 bit characters.
*/
Latin1 = 0,
/*!
* UTF16 with a <i>byte order mark</i>. 16 bit characters.
*/
UTF16 = 1,
/*!
* UTF16 <i>big endian</i>. 16 bit characters. This is the encoding used
* internally by TagLib.
*/
UTF16BE = 2,
/*!
* UTF8 encoding. Characters are usually 8 bits but can be up to 32.
*/
UTF8 = 3,
/*!
* UTF16 <i>little endian</i>. 16 bit characters.
*/
UTF16LE = 4
};
/*!
* Constructs an empty String.
*/
String();
/*!
* Make a shallow, implicitly shared, copy of \a s. Because this is
* implicitly shared, this method is lightweight and suitable for
* pass-by-value usage.
*/
String(const String &s);
/*!
* Makes a deep copy of the data in \a s.
*
* \note This should only be used with the 8-bit codecs Latin1 and UTF8, when
* used with other codecs it will simply print a warning and exit.
*/
String(const std::string &s, Type t = Latin1);
/*!
* Makes a deep copy of the data in \a s.
*
* /note If \a t is UTF16LE, the byte order of \a s will be swapped regardless
* of the CPU byte order. If UTF16BE, it will not be swapped. This behavior
* will be changed in TagLib2.0.
*/
String(const wstring &s, Type t = UTF16BE);
/*!
* Makes a deep copy of the data in \a s.
*
* /note If \a t is UTF16LE, the byte order of \a s will be swapped regardless
* of the CPU byte order. If UTF16BE, it will not be swapped. This behavior
* will be changed in TagLib2.0.
*/
String(const wchar_t *s, Type t = UTF16BE);
/*!
* Makes a deep copy of the data in \a c.
*
* \note This should only be used with the 8-bit codecs Latin1 and UTF8, when
* used with other codecs it will simply print a warning and exit.
*/
String(char c, Type t = Latin1);
/*!
* Makes a deep copy of the data in \a c.
*/
String(wchar_t c, Type t = Latin1);
/*!
* Makes a deep copy of the data in \a s.
*
* \note This should only be used with the 8-bit codecs Latin1 and UTF8, when
* used with other codecs it will simply print a warning and exit.
*/
String(const char *s, Type t = Latin1);
/*!
* Makes a deep copy of the data in \a v.
*/
String(const ByteVector &v, Type t = Latin1);
/*!
* Destroys this String instance.
*/
virtual ~String();
/*!
* Returns a deep copy of this String as an std::string. The returned string
* is encoded in UTF8 if \a unicode is true, otherwise Latin1.
*
* \see toCString()
*/
std::string to8Bit(bool unicode = false) const;
/*!
* Returns a deep copy of this String as a wstring. The returned string is
* encoded in UTF-16 (without BOM/CPU byte order), not UTF-32 even if wchar_t
* is 32-bit wide.
*
* \see toCWString()
*/
wstring toWString() const;
/*!
* Creates and returns a standard C-style (null-terminated) version of this
* String. The returned string is encoded in UTF8 if \a unicode is true,
* otherwise Latin1.
*
* The returned string is still owned by this String and should not be deleted
* by the user.
*
* The returned pointer remains valid until this String instance is destroyed
* or toCString() is called again.
*
* \warning This however has the side effect that the returned string will remain
* in memory <b>in addition to</b> other memory that is consumed by this
* String instance. So, this method should not be used on large strings or
* where memory is critical. Consider using to8Bit() instead to avoid it.
*
* \see to8Bit()
*/
const char *toCString(bool unicode = false) const;
/*!
* Returns a standard C-style (null-terminated) wide character version of
* this String. The returned string is encoded in UTF-16 (without BOM/CPU byte
* order), not UTF-32 even if wchar_t is 32-bit wide.
*
* The returned string is still owned by this String and should not be deleted
* by the user.
*
* The returned pointer remains valid until this String instance is destroyed
* or any other method of this String is called.
*
* \note This returns a pointer to the String's internal data without any
* conversions.
*
* \see toWString()
*/
const wchar_t *toCWString() const;
/*!
* Returns an iterator pointing to the beginning of the string.
*/
Iterator begin();
/*!
* Returns a const iterator pointing to the beginning of the string.
*/
ConstIterator begin() const;
/*!
* Returns an iterator pointing to the end of the string (the position
* after the last character).
*/
Iterator end();
/*!
* Returns a const iterator pointing to the end of the string (the position
* after the last character).
*/
ConstIterator end() const;
/*!
* Finds the first occurrence of pattern \a s in this string starting from
* \a offset. If the pattern is not found, -1 is returned.
*/
int find(const String &s, int offset = 0) const;
/*!
* Finds the last occurrence of pattern \a s in this string, searched backwards,
* either from the end of the string or starting from \a offset. If the pattern
* is not found, -1 is returned.
*/
int rfind(const String &s, int offset = -1) const;
/*!
* Splits the string on each occurrence of \a separator.
*/
StringList split(const String &separator = " ") const;
/*!
* Returns true if the strings starts with the substring \a s.
*/
bool startsWith(const String &s) const;
/*!
* Extract a substring from this string starting at \a position and
* continuing for \a n characters.
*/
String substr(unsigned int position, unsigned int n = 0xffffffff) const;
/*!
* Append \a s to the current string and return a reference to the current
* string.
*/
String &append(const String &s);
/*!
* Clears the string.
*/
String &clear();
/*!
* Returns an upper case version of the string.
*
* \warning This only works for the characters in US-ASCII, i.e. A-Z.
*/
String upper() const;
/*!
* Returns the size of the string.
*/
unsigned int size() const;
/*!
* Returns the length of the string. Equivalent to size().
*/
unsigned int length() const;
/*!
* Returns true if the string is empty.
*
* \see isNull()
*/
bool isEmpty() const;
/*!
* Returns true if this string is null -- i.e. it is a copy of the
* String::null string.
*
* \note A string can be empty and not null. So do not use this method to
* check if the string is empty.
*
* \see isEmpty()
*
* \deprecated
*/
// BIC: remove
bool isNull() const;
/*!
* Returns a ByteVector containing the string's data. If \a t is Latin1 or
* UTF8, this will return a vector of 8 bit characters, otherwise it will use
* 16 bit characters.
*
* \note If \a t is UTF16, the returned data is encoded in little-endian
* format and has a BOM.
*
* \note The returned data is not null terminated.
*/
ByteVector data(Type t) const;
/*!
* Convert the string to an integer.
*
* Returns the integer if the conversion was successful or 0 if the
* string does not represent a number.
*/
// BIC: merge with the method below
int toInt() const;
/*!
* Convert the string to an integer.
*
* If the conversion was successful, it sets the value of \a *ok to
* true and returns the integer. Otherwise it sets \a *ok to false
* and the result is undefined.
*/
int toInt(bool *ok) const;
/*!
* Returns a string with the leading and trailing whitespace stripped.
*/
String stripWhiteSpace() const;
/*!
* Returns true if the file only uses characters required by Latin1.
*/
bool isLatin1() const;
/*!
* Returns true if the file only uses characters required by (7-bit) ASCII.
*/
bool isAscii() const;
/*!
* Converts the base-10 integer \a n to a string.
*/
static String number(int n);
/*!
* Returns a reference to the character at position \a i.
*/
wchar_t &operator[](int i);
/*!
* Returns a const reference to the character at position \a i.
*/
const wchar_t &operator[](int i) const;
/*!
* Compares each character of the String with each character of \a s and
* returns true if the strings match.
*/
bool operator==(const String &s) const;
/*!
* Compares each character of the String with each character of \a s and
* returns false if the strings match.
*/
bool operator!=(const String &s) const;
/*!
* Compares each character of the String with each character of \a s and
* returns true if the strings match.
*/
bool operator==(const char *s) const;
/*!
* Compares each character of the String with each character of \a s and
* returns false if the strings match.
*/
bool operator!=(const char *s) const;
/*!
* Compares each character of the String with each character of \a s and
* returns true if the strings match.
*/
bool operator==(const wchar_t *s) const;
/*!
* Compares each character of the String with each character of \a s and
* returns false if the strings match.
*/
bool operator!=(const wchar_t *s) const;
/*!
* Appends \a s to the end of the String.
*/
String &operator+=(const String &s);
/*!
* Appends \a s to the end of the String.
*/
String &operator+=(const wchar_t* s);
/*!
* Appends \a s to the end of the String.
*/
String &operator+=(const char* s);
/*!
* Appends \a s to the end of the String.
*/
String &operator+=(wchar_t c);
/*!
* Appends \a c to the end of the String.
*/
String &operator+=(char c);
/*!
* Performs a shallow, implicitly shared, copy of \a s, overwriting the
* String's current data.
*/
String &operator=(const String &s);
/*!
* Performs a deep copy of the data in \a s.
*/
String &operator=(const std::string &s);
/*!
* Performs a deep copy of the data in \a s.
*/
String &operator=(const wstring &s);
/*!
* Performs a deep copy of the data in \a s.
*/
String &operator=(const wchar_t *s);
/*!
* Performs a deep copy of the data in \a s.
*/
String &operator=(char c);
/*!
* Performs a deep copy of the data in \a s.
*/
String &operator=(wchar_t c);
/*!
* Performs a deep copy of the data in \a s.
*/
String &operator=(const char *s);
/*!
* Performs a deep copy of the data in \a v.
*/
String &operator=(const ByteVector &v);
/*!
* Exchanges the content of the String by the content of \a s.
*/
void swap(String &s);
/*!
* To be able to use this class in a Map, this operator needed to be
* implemented. Returns true if \a s is less than this string in a byte-wise
* comparison.
*/
bool operator<(const String &s) const;
/*!
* A null string provided for convenience.
*
* \warning Do not modify this variable. It will mess up the internal state
* of TagLib.
*
* \deprecated
*/
// BIC: remove
static String null;
protected:
/*!
* If this String is being shared via implicit sharing, do a deep copy of the
* data and separate from the shared members. This should be called by all
* non-const subclass members.
*/
void detach();
private:
/*!
* \deprecated This variable is no longer used, but NEVER remove this. It
* may lead to a linkage error.
*/
// BIC: remove
static const Type WCharByteOrder;
class StringPrivate;
StringPrivate *d;
};
}
/*!
* \relates TagLib::String
*
* Concatenates \a s1 and \a s2 and returns the result as a string.
*/
TAGLIB_EXPORT const TagLib::String operator+(const TagLib::String &s1, const TagLib::String &s2);
/*!
* \relates TagLib::String
*
* Concatenates \a s1 and \a s2 and returns the result as a string.
*/
TAGLIB_EXPORT const TagLib::String operator+(const char *s1, const TagLib::String &s2);
/*!
* \relates TagLib::String
*
* Concatenates \a s1 and \a s2 and returns the result as a string.
*/
TAGLIB_EXPORT const TagLib::String operator+(const TagLib::String &s1, const char *s2);
/*!
* \relates TagLib::String
*
* Send the string to an output stream.
*/
TAGLIB_EXPORT std::ostream &operator<<(std::ostream &s, const TagLib::String &str);
#endif