// Copyright (C) 2020 Klarälvdalens Datakonsult AB, a KDAB Group company, info@kdab.com, author Marc Mutz // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only #include "qanystringview.h" #include "qdebug.h" #include "qttypetraits.h" QT_BEGIN_NAMESPACE /*! \class QAnyStringView \inmodule QtCore \since 6.0 \brief The QAnyStringView class provides a unified view on Latin-1, UTF-8, or UTF-16 strings with a read-only subset of the QString API. \reentrant \ingroup tools \ingroup string-processing \compares strong \compareswith strong char16_t QChar {const char16_t *} {const char *} \ QByteArray QByteArrayView QString QStringView QUtf8StringView \ QLatin1StringView \endcompareswith A QAnyStringView references a contiguous portion of a string it does not own. It acts as an interface type to all kinds of strings, without the need to construct a QString first. Unlike QStringView and QUtf8StringView, QAnyStringView can hold strings of any of the following encodings: UTF-8, UTF-16, and Latin-1. The latter is supported because Latin-1, unlike UTF-8, can be efficiently compared to UTF-16 data: a length mismatch already means the strings cannot be equal. This is not true for UTF-8/UTF-16 comparisons, because UTF-8 is a variable-length encoding. The string may be represented as an array (or an array-compatible data-structure such as QString, std::basic_string, etc.) of \c char, \c char8_t, QChar, \c ushort, \c char16_t or (on platforms, such as Windows, where it is a 16-bit type) \c wchar_t. QAnyStringView is designed as an interface type; its main use-case is as a function parameter type. When QAnyStringViews are used as automatic variables or data members, care must be taken to ensure that the referenced string data (for example, owned by a QString) outlives the QAnyStringView on all code paths, lest the string view ends up referencing deleted data. When used as an interface type, QAnyStringView allows a single function to accept a wide variety of string data sources. One function accepting QAnyStringView thus replaces five function overloads (taking QString, \c{(const QChar*, qsizetype)}, QUtf8StringView, QLatin1StringView (but see above), and QChar), while at the same time enabling even more string data sources to be passed to the function, such as \c{u8"Hello World"}, a \c char8_t string literal. Like elsewhere in Qt, QAnyStringView assumes \c char data is encoded in UTF-8, unless it is presented as a QLatin1StringView. Since Qt 6.4, however, UTF-8 string literals that are pure US-ASCII are automatically stored as Latin-1. This is a compile-time check with no runtime overhead. The feature requires compiling in C++20, or with a recent GCC. QAnyStringViews should be passed by value, not by reference-to-const: \snippet code/src_corelib_text_qanystringview.cpp 0 QAnyStringView can also be used as the return value of a function, but this is not recommended. QUtf8StringView or QStringView are better suited as function return values. If you call a function returning QAnyStringView, take extra care to not keep the QAnyStringView around longer than the function promises to keep the referenced string data alive. If in doubt, obtain a strong reference to the data by calling toString() to convert the QAnyStringView into a QString. QAnyStringView is a \e{Literal Type}. \section2 Compatible Character Types QAnyStringView accepts strings over a variety of character types: \list \li \c char (both signed and unsigned) \li \c char8_t (C++20 only) \li \c char16_t \li \c wchar_t (where it's a 16-bit type, e.g. Windows) \li \c ushort \li \c QChar \endlist The 8-bit character types are interpreted as UTF-8 data (except when presented as a QLatin1StringView) while the 16-bit character types are interpreted as UTF-16 data in host byte order (the same as QString). \section2 Sizes and Sub-Strings All sizes and positions in QAnyStringView functions are in the encoding's code units (that is, UTF-16 surrogate pairs count as two for the purposes of these functions, the same as in QString, and UTF-8 multibyte sequences count as two, three or four, depending on their length). \sa QUtf8StringView, QStringView */ /*! \typedef QAnyStringView::difference_type Alias for \c{std::ptrdiff_t}. Provided for compatibility with the STL. */ /*! \typedef QAnyStringView::size_type Alias for qsizetype. Provided for compatibility with the STL. */ /*! \fn QAnyStringView::QAnyStringView() Constructs a null string view. \sa isNull() */ /*! \fn QAnyStringView::QAnyStringView(std::nullptr_t) Constructs a null string view. \sa isNull() */ /*! \fn template = true> QAnyStringView::QAnyStringView(const Char *str, qsizetype len) Constructs a string view on \a str with length \a len. The range \c{[str,len)} must remain valid for the lifetime of this string view object. Passing \nullptr as \a str is safe if \a len is 0, too, and results in a null string view. The behavior is undefined if \a len is negative or, when positive, if \a str is \nullptr. This constructor only participates in overload resolution if \c Char is a compatible character type. \sa isNull(), {Compatible Character Types} */ /*! \fn template = true> QAnyStringView::QAnyStringView(const Char *first, const Char *last) Constructs a string view on \a first with length (\a last - \a first). The range \c{[first,last)} must remain valid for the lifetime of this string view object. Passing \nullptr as \a first is safe if \a last is \nullptr, too, and results in a null string view. The behavior is undefined if \a last precedes \a first, or \a first is \nullptr and \a last is not. This constructor only participates in overload resolution if \c Char is a compatible character type. \sa isNull(), {Compatible Character Types} */ /*! \fn template QAnyStringView::QAnyStringView(const Char *str) Constructs a string view on \a str. The length is determined by scanning for the first \c{Char(0)}. \a str must remain valid for the lifetime of this string view object. Passing \nullptr as \a str is safe and results in a null string view. This constructor only participates in overload resolution if \a str is not an array and if \c Char is a compatible character type. \sa isNull(), {Compatible Character Types} */ /*! \fn template QAnyStringView::QAnyStringView(const Char (&string)[N]) Constructs a string view on the character string literal \a string. The view covers the array until the first \c{Char(0)} is encountered, or \c N, whichever comes first. If you need the full array, use fromArray() instead. \a string must remain valid for the lifetime of this string view object. This constructor only participates in overload resolution if \a string is an actual array and \c Char is a compatible character type. \sa {Compatible Character Types} */ /*! \fn QAnyStringView::QAnyStringView(const QString &str) Constructs a string view on \a str. \c{str.data()} must remain valid for the lifetime of this string view object. The string view will be null if and only if \c{str.isNull()}. */ /*! \fn QAnyStringView::QAnyStringView(const QByteArray &str) Constructs a string view on \a str. The data in \a str is interpreted as UTF-8. \c{str.data()} must remain valid for the lifetime of this string view object. The string view will be null if and only if \c{str.isNull()}. */ /*! \fn template > QAnyStringView::QAnyStringView(const Container &str) Constructs a string view on \a str. The length is taken from \c{std::size(str)}. \c{std::data(str)} must remain valid for the lifetime of this string view object. This constructor only participates in overload resolution if \c Container is a container with a compatible character type as \c{value_type}. The string view will be empty if and only if \c{std::size(str) == 0}. It is unspecified whether this constructor can result in a null string view (\c{std::data(str)} would have to return \nullptr for this). \sa isNull(), isEmpty() */ /*! \fn template static QAnyStringView fromArray(const Char (&string)[Size]) noexcept Constructs a string view on the full character string literal \a string, including any trailing \c{Char(0)}. If you don't want the null-terminator included in the view then you can use the constructor overload taking a pointer and a size: \snippet code/src_corelib_text_qanystringview.cpp 2 Alternatively you can use the constructor overload taking an array literal which will create a view up to, but not including, the first null-terminator in the data. \a string must remain valid for the lifetime of this string view object. This function will work with any array literal if \c Char is a compatible character type. */ /*! \fn QString QAnyStringView::toString() const Returns a deep copy of this string view's data as a QString. The return value will be a null QString if and only if this string view is null. */ /*! \fn const void *QAnyStringView::data() const Returns a const pointer to the first character in the string view. \note The character array represented by the return value is \e not null-terminated. \sa size_bytes() */ /*! \fn bool QAnyStringView::empty() const Returns whether this string view is empty - that is, whether \c{size() == 0}. This function is provided for STL compatibility. \sa isEmpty(), isNull(), size() */ /*! \fn bool QAnyStringView::isEmpty() const Returns whether this string view is empty - that is, whether \c{size() == 0}. This function is provided for compatibility with other Qt containers. \sa empty(), isNull(), size() */ /*! \fn bool QAnyStringView::isNull() const Returns whether this string view is null - that is, whether \c{data() == nullptr}. This functions is provided for compatibility with other Qt containers. \sa empty(), isEmpty(), size() */ /*! \fn qsizetype QAnyStringView::size() const Returns the size of this string view, in the encoding's code points. \sa empty(), isEmpty(), isNull(), size_bytes(), {Sizes and Sub-Strings} */ /*! \fn QAnyStringView::size_bytes() const Returns the size of this string view, but in bytes, not code-points. You can use this function together with data() for hashing or serialization. This function is provided for STL compatibility. \sa size(), data() */ /*! \fn QAnyStringView::length() const Same as size(). This function is provided for compatibility with other Qt containers. \sa size() */ /*! \fn QChar QAnyStringView::front() const Returns the first character in the string view. This function is provided for STL compatibility. \warning Calling this function on an empty string view constitutes undefined behavior. \sa back(), {Sizes and Sub-Strings} */ /*! \fn QChar QAnyStringView::back() const Returns the last character in the string view. This function is provided for STL compatibility. \warning Calling this function on an empty string view constitutes undefined behavior. \sa front(), {Sizes and Sub-Strings} */ /*! \fn QAnyStringView::mid(qsizetype pos, qsizetype n) const \since 6.5 Returns the substring of length \a n starting at position \a pos in this object. \deprecated Use sliced() instead in new code. Returns an empty string view if \a n exceeds the length of the string view. If there are less than \a n code points available in the string view starting at \a pos, or if \a n is negative (default), the function returns all code points that are available from \a pos. \sa first(), last(), sliced(), chopped(), chop(), truncate(), {Sizes and Sub-Strings} */ /*! \fn QAnyStringView::left(qsizetype n) const \since 6.5 \deprecated Use first() instead in new code. Returns the substring of length \a n starting at position 0 in this object. The entire string view is returned if \a n is greater than or equal to size(), or less than zero. \sa first(), last(), sliced(), chopped(), chop(), truncate(), {Sizes and Sub-Strings} */ /*! \fn QAnyStringView::right(qsizetype n) const \since 6.5 \deprecated Use last() instead in new code. Returns the substring of length \a n starting at position size() - \a n in this object. The entire string view is returned if \a n is greater than or equal to size(), or less than zero. \sa first(), last(), sliced(), chopped(), chop(), truncate(), {Sizes and Sub-Strings} */ /*! \fn QAnyStringView::first(qsizetype n) const \since 6.5 Returns a string view that contains the first \a n code points of this string view. \note The behavior is undefined when \a n < 0 or \a n > size(). \sa last(), sliced(), chopped(), chop(), truncate(), {Sizes and Sub-Strings} */ /*! \fn QAnyStringView::last(qsizetype n) const \since 6.5 Returns a string view that contains the last \a n code points of this string view. \note The behavior is undefined when \a n < 0 or \a n > size(). \sa first(), sliced(), chopped(), chop(), truncate(), {Sizes and Sub-Strings} */ /*! \fn QAnyStringView::sliced(qsizetype pos, qsizetype n) const \since 6.5 Returns a string view containing \a n code points of this string view, starting at position \a pos. //! [UB-sliced-index-length] \note The behavior is undefined when \a pos < 0, \a n < 0, or \a pos + \a n > size(). //! [UB-sliced-index-length] \sa first(), last(), chopped(), chop(), truncate(), {Sizes and Sub-Strings} */ /*! \fn QAnyStringView::sliced(qsizetype pos) const \since 6.5 Returns a string view starting at position \a pos in this object, and extending to its end. //! [UB-sliced-index-only] \note The behavior is undefined when \a pos < 0 or \a pos > size(). //! [UB-sliced-index-only] \sa first(), last(), chopped(), chop(), truncate(), {Sizes and Sub-Strings} */ /*! \fn QAnyStringView::chopped(qsizetype n) const \since 6.5 Returns the substring of length size() - \a n starting at the beginning of this object. Same as \c{first(size() - n)}. \note The behavior is undefined when \a n < 0 or \a n > size(). \sa sliced(), first(), last(), chop(), truncate(), {Sizes and Sub-Strings} */ /*! \fn QAnyStringView::truncate(qsizetype n) \since 6.5 Truncates this string view to \a n code points. Same as \c{*this = first(n)}. \note The behavior is undefined when \a n < 0 or \a n > size(). \sa sliced(), first(), last(), chopped(), chop(), {Sizes and Sub-Strings} */ /*! \fn QAnyStringView::chop(qsizetype n) \since 6.5 Truncates this string view by \a n code points. Same as \c{*this = first(size() - n)}. \note The behavior is undefined when \a n < 0 or \a n > size(). \sa sliced(), first(), last(), chopped(), truncate(), {Sizes and Sub-Strings} */ /*! \fn template decltype(auto) QAnyStringView::visit(Visitor &&v) const Calls \a v with either a QUtf8StringView, QLatin1String, or QStringView, depending on the encoding of the string data this string-view references. This is how most functions taking QAnyStringView fork off into per-encoding functions: \code void processImpl(QLatin1String s) { ~~~ } void processImpl(QUtf8StringView s) { ~~~ } void processImpl(QStringView s) { ~~~ } void process(QAnyStringView s) { s.visit([](auto s) { processImpl(s); }); } \endcode Here, we're reusing the same name, \c s, for both the QAnyStringView object, as well as the lambda's parameter. This is idiomatic code and helps track the identity of the objects through visit() calls, for example in more complex situations such as \code bool equal(QAnyStringView lhs, QAnyStringView rhs) { // assuming operator==(QAnyStringView, QAnyStringView) didn't, yet, exist: return lhs.visit([rhs](auto lhs) { rhs.visit([lhs](auto rhs) { return lhs == rhs; }); }); } \endcode visit() requires that all lambda instantiations have the same return type. If they differ, you get a compile error, even if there is a common type. To fix, you can use explicit return types on the lambda, or cast in the return statements: \code // wrong: QAnyStringView firstHalf(QAnyStringView input) { return input.visit([](auto input) { // ERROR: lambdas return different types return input.sliced(0, input.size() / 2); }); } // correct: QAnyStringView firstHalf(QAnyStringView input) { return input.visit([](auto input) -> QAnyStringView { // OK, explicit return type return input.sliced(0, input.size() / 2); }); } // also correct: QAnyStringView firstHalf(QAnyStringView input) { return input.visit([](auto input) { return QAnyStringView(input.sliced(0, input.size() / 2)); // OK, cast to common type }); } \endcode */ /*! \fn QAnyStringView::compare(QAnyStringView lhs, QAnyStringView rhs, Qt::CaseSensitivity cs) Compares the string view \a lhs with the string view \a rhs and returns a negative integer if \a lhs is less than \a rhs, a positive integer if it is greater than \a rhs, and zero if they are equal. If \a cs is Qt::CaseSensitive (the default), the comparison is case sensitive; otherwise the comparison is case-insensitive. \sa operator==(), operator<(), operator>() */ /*! \fn bool QAnyStringView::operator==(const QAnyStringView &lhs, const QAnyStringView & rhs) \fn bool QAnyStringView::operator!=(const QAnyStringView & lhs, const QAnyStringView & rhs) \fn bool QAnyStringView::operator<=(const QAnyStringView & lhs, const QAnyStringView & rhs) \fn bool QAnyStringView::operator>=(const QAnyStringView & lhs, const QAnyStringView & rhs) \fn bool QAnyStringView::operator<(const QAnyStringView & lhs, const QAnyStringView & rhs) \fn bool QAnyStringView::operator>(const QAnyStringView & lhs, const QAnyStringView & rhs) Operators that compare \a lhs to \a rhs. \sa compare() */ /*! \fn template qToAnyStringViewIgnoringNull(const QStringLike &s); \since 6.0 \internal Convert \a s to a QAnyStringView ignoring \c{s.isNull()}. Returns a string view that references \a{s}'s data, but is never null. This is a faster way to convert a QString or QByteArray to a QAnyStringView, if null QStrings or QByteArrays can legitimately be treated as empty ones. \sa QString::isNull(), QAnyStringView */ /*! \fn QAnyStringView::operator<<(QDebug d, QAnyStringView s) \since 6.7 \relates QDebug Outputs \a s to debug stream \a d. If \c{d.quotedString()} is \c true, indicates which encoding the string is in. If you just want the string data, use visit() like this: \code s.visit([&d) (auto s) { d << s; }); \endcode \sa QAnyStringView::visit() */ QDebug operator<<(QDebug d, QAnyStringView s) { struct S { const char *prefix, *suffix; }; const auto affixes = s.visit([](auto s) { using View = decltype(s); if constexpr (std::is_same_v) { return S{"", "_L1"}; } else if constexpr (std::is_same_v) { return S{"u8", ""}; } else if constexpr (std::is_same_v) { return S{"u", ""}; } else { static_assert(QtPrivate::type_dependent_false()); } }); const QDebugStateSaver saver(d); d.nospace(); if (d.quoteStrings()) d << affixes.prefix; s.visit([&d](auto s) { d << s; }); if (d.quoteStrings()) d << affixes.suffix; return d; } QT_END_NAMESPACE