目录
1.引言
2.原理分析
2.1.结构
2.2.构造函数
2.3.成员函数
2.4.std::string_view字面量
3.实例
3.1.std::string_view和std::string的运算符操作
3.2.查找函数使用
3.3.std::string_view和临时字符串
4.总结
1.引言
在C/C++日常编程中,我们常进行数据的传递操作,比如,将数据传给函数。当数据占用的内存较大时,减少数据的拷贝可以有效提高程序的性能。在C++17之前,为了接收只读字符串的函数选择形参一直是一件进退两难的事情,它应该是const char*吗?那样的话,如果客户用std::string,这必须使用c_str()或data()来获取const char*。更糟糕的是,函数讲失去std::string良好的面向对象的方法及其良好的辅助方法。或许,形参应改用const std::string&?这种情况下,始终需要std::string。例如,如果传递一个字符串字面量,编译器将默认创建一个临时字符串对象(其中包括字符串字面量的副本),并将该对象传递给函数,因此会增加一点开销。有时,人们会编写同一函数的多个重载版本,一个接收const char*,另一个接收const std::string&,单显然,这并不是一个优雅的解决方案,而且std::string的substr函数,每次都要返回一个新生成的子串,很容易引起性能问题。实际上我们本意不是要改变原字符串,为什么不在原字符串基础上返回呢?
在C++17中引入了std::string_view,就很好的解决了上面这些问题。
2.原理分析
std::string_view是字符串的视图版本,它能让我们像处理字符串一样处理字符序列,而不需要为它们分配内存空间。也就是说,std::string_view类型的对象只是引用一个外部的字符序列,而不需要持有它们。因此,一个字符串视图对象可以被看作字符串序列的引用 。
使用字符串视图的开销很小,速度却很快(以值传递一个std::string_view的开销总是很小)。然而,它也有一些潜在的危险,就和原生指针一样,在使用string_view时也必须由程序员自己来保证引用的字符串序列是有效的。那么,std::string_view为什么能做到开销小,速度很快呢?我们接着往下看,本文以VS2019平台展开讲解std::string_view的原理和深层次用法。
2.1.结构
std::string_view的类UML图如下:
std::string_view是std::basic_string_view<char>的特化版本,std::basic_string_view的所有接口都适用于std::std::string_view。对于使用宽字符集,例如Unicode
或者某些亚洲字符集的字符串,还定义了其它几个版本:
#ifdef __cpp_lib_char8_t
using u8string_view = basic_string_view<char8_t>;
#endif // __cpp_lib_char8_t
using u16string_view = basic_string_view<char16_t>;
using u32string_view = basic_string_view<char32_t>;
using wstring_view = basic_string_view<wchar_t>;
std::basic_string_view的内部结构:
template <class _Elem, class _Traits>
class basic_string_view { // wrapper for any kind of contiguous character buffer
public:
using traits_type = _Traits;
using value_type = _Elem;
using pointer = _Elem*;
using const_pointer = const _Elem*;
using reference = _Elem&;
using const_reference = const _Elem&;
using const_iterator = _String_view_iterator<_Traits>;
using iterator = const_iterator;
using const_reverse_iterator = _STD reverse_iterator<const_iterator>;
using reverse_iterator = const_reverse_iterator;
using size_type = size_t;
using difference_type = ptrdiff_t;
//...
private:
const_pointer _Mydata;
size_type _Mysize;
};
从中可以看出std::basic_string_view只存储了{_Mydata,_Mysize}两个元素,不会具体存储原数据,仅仅存储指向的数据的起始指针和长度,所以这个开销是非常小的。
2.2.构造函数
std::basic_string_view的构造函数如下:
//构造函数
constexpr basic_string_view() noexcept : _Mydata(), _Mysize(0) {} //1
constexpr basic_string_view(const basic_string_view&) noexcept = default; //2
constexpr basic_string_view& operator=(const basic_string_view&) noexcept = default; //3
/* implicit */ constexpr basic_string_view(_In_z_ const const_pointer _Ntcts) noexcept // strengthened
: _Mydata(_Ntcts), _Mysize(_Traits::length(_Ntcts)) {} //4
constexpr basic_string_view(
_In_reads_(_Count) const const_pointer _Cts, const size_type _Count) noexcept // strengthened
: _Mydata(_Cts), _Mysize(_Count) {
#if _CONTAINER_DEBUG_LEVEL > 0
_STL_VERIFY(_Count == 0 || _Cts, "non-zero size null string_view");
#endif // _CONTAINER_DEBUG_LEVEL > 0
} //5
#ifdef __cpp_lib_concepts
// clang-format off
template <contiguous_iterator _It, sized_sentinel_for<_It> _Se>
requires (is_same_v<iter_value_t<_It>, _Elem> && !is_convertible_v<_Se, size_type>)
constexpr basic_string_view(_It _First, _Se _Last) noexcept(noexcept(_Last - _First)) // strengthened
: _Mydata(_STD to_address(_First)), _Mysize(static_cast<size_type>(_Last - _First)) {} //6
// clang-format on
#endif // __cpp_lib_concepts
// 从迭代器中获取地址函数to_address
从上面的远源码可以看出,td::basic_string_view的构造函数有:
1)std::string_view a; 调用第1个构造函数
2)std::string_view a("123"); 调用第4个构造函数,改函数里面调用了_Traits::length函数
_NODISCARD static _CONSTEXPR17 size_t length(_In_z_ const _Elem* const _First) noexcept /* strengthened */ {
// find length of null-terminated string
#if _HAS_CXX17
#ifdef __cpp_char8_t
if constexpr (is_same_v<_Elem, char8_t>) {
#if _HAS_U8_INTRINSICS
return __builtin_u8strlen(_First);
#else // ^^^ use u8 intrinsics / no u8 intrinsics vvv
return _Primary_char_traits::length(_First);
#endif // _HAS_U8_INTRINSICS
} else
#endif // __cpp_char8_t
{
return __builtin_strlen(_First);
}
#else // _HAS_CXX17
return _CSTD strlen(reinterpret_cast<const char*>(_First));
#endif // _HAS_CXX17
}
_Traits::length函数又调用了strlen计算了字符串的长度;构造了一个3长度的std::string_view。
3)std::string_view a("122323423", 5); 调用第5个构造函数,把内容和长度构造函数,构建了一个内容为"12232",长度为5的std::string_view。
4)构造函数还可以接收迭代器,如下:
char b[] = "121212e124124";
std::string_view e(std::begin(b), std::end(b));
调用第6个构造函数,里面调用to_address通过迭代器获取指针:
template <class _Ty, class = void>
inline constexpr bool _Has_to_address_v = false; // determines whether _Ptr has pointer_traits<_Ptr>::to_address(p)
template <class _Ty>
inline constexpr bool
_Has_to_address_v<_Ty, void_t<decltype(pointer_traits<_Ty>::to_address(_STD declval<const _Ty&>()))>> = true;
template <class _Ty>
_NODISCARD constexpr _Ty* to_address(_Ty* const _Val) noexcept {
static_assert(!is_function_v<_Ty>,
"N4810 20.10.4 [pointer.conversion]/2: The program is ill-formed if T is a function type.");
return _Val;
}
template <class _Ptr>
_NODISCARD constexpr auto to_address(const _Ptr& _Val) noexcept {
if constexpr (_Has_to_address_v<_Ptr>) {
return pointer_traits<_Ptr>::to_address(_Val);
} else {
return _STD to_address(_Val.operator->()); // plain pointer overload must come first
}
}
5)拷贝构造函数和赋值构造函数都是使用的系统默认函数,类似memcpy,直接把_Mydata,_Mysize的值拷贝到另外对象,比较简单,就不赘述了。
上面的都好理解,唯一需要说明的是:为什么我们代码string_view test(string("123"))可以编译通过,但为什么没有对应的构造函数?
实际上这是因为std::string类重载了std::string到std::string_view的转换操作符:
operator std::basic_string_view<CharT, Traits>() const noexcept;
所以,std::string_view test(std::string("123"))实际执行了两步操作:
a.std::string("abc")转换为std::string_view对象
b.test调用了第2个构造函数生成std::string_view对象
2.3.成员函数
1)获取容量的函数
_NODISCARD constexpr size_type size() const noexcept {
return _Mysize;
}
_NODISCARD constexpr size_type length() const noexcept {
return _Mysize;
}
_NODISCARD constexpr bool empty() const noexcept {
return _Mysize == 0;
}
_NODISCARD constexpr size_type max_size() const noexcept {
// bound to PTRDIFF_MAX to make end() - begin() well defined (also makes room for npos)
// bound to static_cast<size_t>(-1) / sizeof(_Elem) by address space limits
return (_STD min)(static_cast<size_t>(PTRDIFF_MAX), static_cast<size_t>(-1) / sizeof(_Elem));
}
2) 迭代器相关的函数
_NODISCARD constexpr const_iterator begin() const noexcept {
#if _ITERATOR_DEBUG_LEVEL >= 1
return const_iterator(_Mydata, _Mysize, 0);
#else // ^^^ _ITERATOR_DEBUG_LEVEL >= 1 ^^^ // vvv _ITERATOR_DEBUG_LEVEL == 0 vvv
return const_iterator(_Mydata);
#endif // _ITERATOR_DEBUG_LEVEL
}
_NODISCARD constexpr const_iterator end() const noexcept {
#if _ITERATOR_DEBUG_LEVEL >= 1
return const_iterator(_Mydata, _Mysize, _Mysize);
#else // ^^^ _ITERATOR_DEBUG_LEVEL >= 1 ^^^ // vvv _ITERATOR_DEBUG_LEVEL == 0 vvv
return const_iterator(_Mydata + _Mysize);
#endif // _ITERATOR_DEBUG_LEVEL
}
_NODISCARD constexpr const_iterator cbegin() const noexcept {
return begin();
}
_NODISCARD constexpr const_iterator cend() const noexcept {
return end();
}
_NODISCARD constexpr const_reverse_iterator rbegin() const noexcept {
return const_reverse_iterator{end()};
}
_NODISCARD constexpr const_reverse_iterator rend() const noexcept {
return const_reverse_iterator{begin()};
}
_NODISCARD constexpr const_reverse_iterator crbegin() const noexcept {
return rbegin();
}
_NODISCARD constexpr const_reverse_iterator crend() const noexcept {
return rend();
}
3)元素访问函数
_NODISCARD constexpr const_pointer data() const noexcept {
return _Mydata;
}
_NODISCARD constexpr const_reference operator[](const size_type _Off) const noexcept /* strengthened */ {
#if _CONTAINER_DEBUG_LEVEL > 0
_STL_VERIFY(_Off < _Mysize, "string_view subscript out of range");
#endif // _CONTAINER_DEBUG_LEVEL > 0
return _Mydata[_Off];
}
_NODISCARD constexpr const_reference at(const size_type _Off) const {
// get the character at _Off or throw if that is out of range
_Check_offset_exclusive(_Off);
return _Mydata[_Off];
}
_NODISCARD constexpr const_reference front() const noexcept /* strengthened */ {
#if _CONTAINER_DEBUG_LEVEL > 0
_STL_VERIFY(_Mysize != 0, "cannot call front on empty string_view");
#endif // _CONTAINER_DEBUG_LEVEL > 0
return _Mydata[0];
}
_NODISCARD constexpr const_reference back() const noexcept /* strengthened */ {
#if _CONTAINER_DEBUG_LEVEL > 0
_STL_VERIFY(_Mysize != 0, "cannot call back on empty string_view");
#endif // _CONTAINER_DEBUG_LEVEL > 0
return _Mydata[_Mysize - 1];
}
4)修改器函数
constexpr void remove_prefix(const size_type _Count) noexcept /* strengthened */ {
#if _CONTAINER_DEBUG_LEVEL > 0
_STL_VERIFY(_Mysize >= _Count, "cannot remove prefix longer than total size");
#endif // _CONTAINER_DEBUG_LEVEL > 0
_Mydata += _Count;
_Mysize -= _Count;
}
constexpr void remove_suffix(const size_type _Count) noexcept /* strengthened */ {
#if _CONTAINER_DEBUG_LEVEL > 0
_STL_VERIFY(_Mysize >= _Count, "cannot remove suffix longer than total size");
#endif // _CONTAINER_DEBUG_LEVEL > 0
_Mysize -= _Count;
}
constexpr void swap(basic_string_view& _Other) noexcept {
const basic_string_view _Tmp{_Other}; // note: std::swap is not constexpr before C++20
_Other = *this;
*this = _Tmp;
}
remove_prefix(size_type)和remove_suffix(size_type)方法,前者是将起始指针前移给定的偏移量来收缩字符串,后者是将结尾指针倒退给定的偏移量来收缩字符串,三个函数仅会修改std::string_view的数据指向,不会修改指向的数据。
5)查找比较函数
_CONSTEXPR20 size_type copy(
_Out_writes_(_Count) _Elem* const _Ptr, size_type _Count, const size_type _Off = 0) const {
// copy [_Off, _Off + Count) to [_Ptr, _Ptr + _Count)
_Check_offset(_Off);
_Count = _Clamp_suffix_size(_Off, _Count);
_Traits::copy(_Ptr, _Mydata + _Off, _Count);
return _Count;
}
_Pre_satisfies_(_Dest_size >= _Count) _CONSTEXPR20 size_type
_Copy_s(_Out_writes_all_(_Dest_size) _Elem* const _Dest, const size_type _Dest_size, size_type _Count,
const size_type _Off = 0) const {
// copy [_Off, _Off + _Count) to [_Dest, _Dest + _Count)
_Check_offset(_Off);
_Count = _Clamp_suffix_size(_Off, _Count);
_Traits::_Copy_s(_Dest, _Dest_size, _Mydata + _Off, _Count);
return _Count;
}
_NODISCARD constexpr basic_string_view substr(const size_type _Off = 0, size_type _Count = npos) const {
// return a new basic_string_view moved forward by _Off and trimmed to _Count elements
_Check_offset(_Off);
_Count = _Clamp_suffix_size(_Off, _Count);
return basic_string_view(_Mydata + _Off, _Count);
}
constexpr bool _Equal(const basic_string_view _Right) const noexcept {
return _Traits_equal<_Traits>(_Mydata, _Mysize, _Right._Mydata, _Right._Mysize);
}
_NODISCARD constexpr int compare(const basic_string_view _Right) const noexcept {
return _Traits_compare<_Traits>(_Mydata, _Mysize, _Right._Mydata, _Right._Mysize);
}
_NODISCARD constexpr int compare(const size_type _Off, const size_type _Nx, const basic_string_view _Right) const {
// compare [_Off, _Off + _Nx) with _Right
return substr(_Off, _Nx).compare(_Right);
}
_NODISCARD constexpr int compare(const size_type _Off, const size_type _Nx, const basic_string_view _Right,
const size_type _Roff, const size_type _Count) const {
// compare [_Off, _Off + _Nx) with _Right [_Roff, _Roff + _Count)
return substr(_Off, _Nx).compare(_Right.substr(_Roff, _Count));
}
_NODISCARD constexpr int compare(_In_z_ const _Elem* const _Ptr) const { // compare [0, _Mysize) with [_Ptr, <null>)
return compare(basic_string_view(_Ptr));
}
_NODISCARD constexpr int compare(const size_type _Off, const size_type _Nx, _In_z_ const _Elem* const _Ptr) const {
// compare [_Off, _Off + _Nx) with [_Ptr, <null>)
return substr(_Off, _Nx).compare(basic_string_view(_Ptr));
}
_NODISCARD constexpr int compare(const size_type _Off, const size_type _Nx,
_In_reads_(_Count) const _Elem* const _Ptr, const size_type _Count) const {
// compare [_Off, _Off + _Nx) with [_Ptr, _Ptr + _Count)
return substr(_Off, _Nx).compare(basic_string_view(_Ptr, _Count));
}
#if _HAS_CXX20
_NODISCARD constexpr bool starts_with(const basic_string_view _Right) const noexcept {
const auto _Rightsize = _Right._Mysize;
if (_Mysize < _Rightsize) {
return false;
}
return _Traits::compare(_Mydata, _Right._Mydata, _Rightsize) == 0;
}
_NODISCARD constexpr bool starts_with(const _Elem _Right) const noexcept {
return !empty() && _Traits::eq(front(), _Right);
}
_NODISCARD constexpr bool starts_with(const _Elem* const _Right) const noexcept /* strengthened */ {
return starts_with(basic_string_view(_Right));
}
_NODISCARD constexpr bool ends_with(const basic_string_view _Right) const noexcept {
const auto _Rightsize = _Right._Mysize;
if (_Mysize < _Rightsize) {
return false;
}
return _Traits::compare(_Mydata + (_Mysize - _Rightsize), _Right._Mydata, _Rightsize) == 0;
}
_NODISCARD constexpr bool ends_with(const _Elem _Right) const noexcept {
return !empty() && _Traits::eq(back(), _Right);
}
_NODISCARD constexpr bool ends_with(const _Elem* const _Right) const noexcept /* strengthened */ {
return ends_with(basic_string_view(_Right));
}
#endif // _HAS_CXX20
_NODISCARD constexpr size_type find(const basic_string_view _Right, const size_type _Off = 0) const noexcept {
// look for _Right beginning at or after _Off
return _Traits_find<_Traits>(_Mydata, _Mysize, _Off, _Right._Mydata, _Right._Mysize);
}
_NODISCARD constexpr size_type find(const _Elem _Ch, const size_type _Off = 0) const noexcept {
// look for _Ch at or after _Off
return _Traits_find_ch<_Traits>(_Mydata, _Mysize, _Off, _Ch);
}
_NODISCARD constexpr size_type find(_In_reads_(_Count) const _Elem* const _Ptr, const size_type _Off,
const size_type _Count) const noexcept /* strengthened */ {
// look for [_Ptr, _Ptr + _Count) beginning at or after _Off
return _Traits_find<_Traits>(_Mydata, _Mysize, _Off, _Ptr, _Count);
}
_NODISCARD constexpr size_type find(_In_z_ const _Elem* const _Ptr, const size_type _Off = 0) const noexcept
/* strengthened */ {
// look for [_Ptr, <null>) beginning at or after _Off
return _Traits_find<_Traits>(_Mydata, _Mysize, _Off, _Ptr, _Traits::length(_Ptr));
}
_NODISCARD constexpr size_type rfind(const basic_string_view _Right, const size_type _Off = npos) const noexcept {
// look for _Right beginning before _Off
return _Traits_rfind<_Traits>(_Mydata, _Mysize, _Off, _Right._Mydata, _Right._Mysize);
}
_NODISCARD constexpr size_type rfind(const _Elem _Ch, const size_type _Off = npos) const noexcept {
// look for _Ch before _Off
return _Traits_rfind_ch<_Traits>(_Mydata, _Mysize, _Off, _Ch);
}
_NODISCARD constexpr size_type rfind(_In_reads_(_Count) const _Elem* const _Ptr, const size_type _Off,
const size_type _Count) const noexcept /* strengthened */ {
// look for [_Ptr, _Ptr + _Count) beginning before _Off
return _Traits_rfind<_Traits>(_Mydata, _Mysize, _Off, _Ptr, _Count);
}
_NODISCARD constexpr size_type rfind(_In_z_ const _Elem* const _Ptr, const size_type _Off = npos) const noexcept
/* strengthened */ {
// look for [_Ptr, <null>) beginning before _Off
return _Traits_rfind<_Traits>(_Mydata, _Mysize, _Off, _Ptr, _Traits::length(_Ptr));
}
_NODISCARD constexpr size_type find_first_of(const basic_string_view _Right,
const size_type _Off = 0) const noexcept { // look for one of _Right at or after _Off
return _Traits_find_first_of<_Traits>(
_Mydata, _Mysize, _Off, _Right._Mydata, _Right._Mysize, _Is_specialization<_Traits, char_traits>{});
}
_NODISCARD constexpr size_type find_first_of(const _Elem _Ch, const size_type _Off = 0) const noexcept {
// look for _Ch at or after _Off
return _Traits_find_ch<_Traits>(_Mydata, _Mysize, _Off, _Ch);
}
_NODISCARD constexpr size_type find_first_of(_In_reads_(_Count) const _Elem* const _Ptr, const size_type _Off,
const size_type _Count) const noexcept /* strengthened */ {
// look for one of [_Ptr, _Ptr + _Count) at or after _Off
return _Traits_find_first_of<_Traits>(
_Mydata, _Mysize, _Off, _Ptr, _Count, _Is_specialization<_Traits, char_traits>{});
}
_NODISCARD constexpr size_type find_first_of(
_In_z_ const _Elem* const _Ptr, const size_type _Off = 0) const noexcept /* strengthened */ {
// look for one of [_Ptr, <null>) at or after _Off
return _Traits_find_first_of<_Traits>(
_Mydata, _Mysize, _Off, _Ptr, _Traits::length(_Ptr), _Is_specialization<_Traits, char_traits>{});
}
_NODISCARD constexpr size_type find_last_of(const basic_string_view _Right,
const size_type _Off = npos) const noexcept { // look for one of _Right before _Off
return _Traits_find_last_of<_Traits>(
_Mydata, _Mysize, _Off, _Right._Mydata, _Right._Mysize, _Is_specialization<_Traits, char_traits>{});
}
_NODISCARD constexpr size_type find_last_of(const _Elem _Ch, const size_type _Off = npos) const noexcept {
// look for _Ch before _Off
return _Traits_rfind_ch<_Traits>(_Mydata, _Mysize, _Off, _Ch);
}
_NODISCARD constexpr size_type find_last_of(_In_reads_(_Count) const _Elem* const _Ptr, const size_type _Off,
const size_type _Count) const noexcept /* strengthened */ {
// look for one of [_Ptr, _Ptr + _Count) before _Off
return _Traits_find_last_of<_Traits>(
_Mydata, _Mysize, _Off, _Ptr, _Count, _Is_specialization<_Traits, char_traits>{});
}
_NODISCARD constexpr size_type find_last_of(
_In_z_ const _Elem* const _Ptr, const size_type _Off = npos) const noexcept /* strengthened */ {
// look for one of [_Ptr, <null>) before _Off
return _Traits_find_last_of<_Traits>(
_Mydata, _Mysize, _Off, _Ptr, _Traits::length(_Ptr), _Is_specialization<_Traits, char_traits>{});
}
_NODISCARD constexpr size_type find_first_not_of(const basic_string_view _Right,
const size_type _Off = 0) const noexcept { // look for none of _Right at or after _Off
return _Traits_find_first_not_of<_Traits>(
_Mydata, _Mysize, _Off, _Right._Mydata, _Right._Mysize, _Is_specialization<_Traits, char_traits>{});
}
_NODISCARD constexpr size_type find_first_not_of(const _Elem _Ch, const size_type _Off = 0) const noexcept {
// look for any value other than _Ch at or after _Off
return _Traits_find_not_ch<_Traits>(_Mydata, _Mysize, _Off, _Ch);
}
_NODISCARD constexpr size_type find_first_not_of(_In_reads_(_Count) const _Elem* const _Ptr, const size_type _Off,
const size_type _Count) const noexcept /* strengthened */ {
// look for none of [_Ptr, _Ptr + _Count) at or after _Off
return _Traits_find_first_not_of<_Traits>(
_Mydata, _Mysize, _Off, _Ptr, _Count, _Is_specialization<_Traits, char_traits>{});
}
_NODISCARD constexpr size_type find_first_not_of(
_In_z_ const _Elem* const _Ptr, const size_type _Off = 0) const noexcept /* strengthened */ {
// look for none of [_Ptr, <null>) at or after _Off
return _Traits_find_first_not_of<_Traits>(
_Mydata, _Mysize, _Off, _Ptr, _Traits::length(_Ptr), _Is_specialization<_Traits, char_traits>{});
}
_NODISCARD constexpr size_type find_last_not_of(const basic_string_view _Right,
const size_type _Off = npos) const noexcept { // look for none of _Right before _Off
return _Traits_find_last_not_of<_Traits>(
_Mydata, _Mysize, _Off, _Right._Mydata, _Right._Mysize, _Is_specialization<_Traits, char_traits>{});
}
_NODISCARD constexpr size_type find_last_not_of(const _Elem _Ch, const size_type _Off = npos) const noexcept {
// look for any value other than _Ch before _Off
return _Traits_rfind_not_ch<_Traits>(_Mydata, _Mysize, _Off, _Ch);
}
_NODISCARD constexpr size_type find_last_not_of(_In_reads_(_Count) const _Elem* const _Ptr, const size_type _Off,
const size_type _Count) const noexcept /* strengthened */ {
// look for none of [_Ptr, _Ptr + _Count) before _Off
return _Traits_find_last_not_of<_Traits>(
_Mydata, _Mysize, _Off, _Ptr, _Count, _Is_specialization<_Traits, char_traits>{});
}
_NODISCARD constexpr size_type find_last_not_of(
_In_z_ const _Elem* const _Ptr, const size_type _Off = npos) const noexcept /* strengthened */ {
// look for none of [_Ptr, <null>) before _Off
return _Traits_find_last_not_of<_Traits>(
_Mydata, _Mysize, _Off, _Ptr, _Traits::length(_Ptr), _Is_specialization<_Traits, char_traits>{});
}
_NODISCARD constexpr bool _Starts_with(const basic_string_view _View) const noexcept {
return _Mysize >= _View._Mysize && _Traits::compare(_Mydata, _View._Mydata, _View._Mysize) == 0;
}
这些函数大致std::string的功能一致,在这里我就不赘述了。
2.4.std::string_view字面量
标准的用户自定义字面量sv,将字符串字面量解释为std::string_view,类似Qt的QString中的QStringLiteral。看源码是怎么实现的:
// basic_string_view LITERALS
inline namespace literals {
inline namespace string_view_literals {
_NODISCARD constexpr string_view operator"" sv(const char* _Str, size_t _Len) noexcept {
return string_view(_Str, _Len);
}
_NODISCARD constexpr wstring_view operator"" sv(const wchar_t* _Str, size_t _Len) noexcept {
return wstring_view(_Str, _Len);
}
#ifdef __cpp_char8_t
_NODISCARD constexpr basic_string_view<char8_t> operator"" sv(const char8_t* _Str, size_t _Len) noexcept {
return basic_string_view<char8_t>(_Str, _Len);
}
#endif // __cpp_char8_t
_NODISCARD constexpr u16string_view operator"" sv(const char16_t* _Str, size_t _Len) noexcept {
return u16string_view(_Str, _Len);
}
_NODISCARD constexpr u32string_view operator"" sv(const char32_t* _Str, size_t _Len) noexcept {
return u32string_view(_Str, _Len);
}
} // namespace string_view_literals
} // namespace literals
用户就可以这样定义:
using namespace std::literals::string_view_literals;
using namespace std::string_view_literals;
using namespace std::literals;
using namespace std;
std::string_view sv { "My Hello world"sv };
3.实例
3.1.std::string_view和std::string的运算符操作
当我们将std::string_view类型的常量弱引用类型的字符串和std::string类型的字符串进行相加(运算符+)操作时会出错,必须要先将string_view转化为const char*,也就是调用data()接口,测试代码如下:
#include<iostream>
#include<string>
#include<string_view>
int main()
{
std::string str1 = "hello";
std::string_view sv1 = " world";
//使用+号运算符时,必须将string_view转化为const char*
auto it = str1 + sv1.data();
//使用append追加字符串不会出错
auto it2 = str1.append(sv1);
std::cout << it2 << std::endl;
return 0;
}
警告:返回字符串的函数应该返回const std::string&或std::string,但不应该返回std::string_view。返回std::string_view会带来使返回的std::string_view无效的风险,例如当它指向的字符串需要重新分配时。
警告:将const std:string&或std::string_view存储为类的数据成员需要确保它们指向的字符串在对象的生命周期内保持有效状态,存储std::string更安全。
3.2.查找函数使用
先看一个例子:
string replace_post(string_view src, string_view new_post)
{
// 找到点的位置
auto pos = src.find(".") + 1;
// 取出点及点之前的全部字符,string_view的substr会返回一个
// string_view对象,所以要取data()赋值给string对象
string s1 = src.substr(0, pos).data();
// 加上新的后缀
return s1 + new_post.data();
}
int main()
{
string_view sv = "abcdefg.xxx";
string s = replace_post(sv, "yyy");
cout << sv << " replaced post by yyy result is:" << s << endl;
return 0;
}
输出:
abcdefg.xxxyyy
为什么输出 "abcdefg.xxxyyy" 了呢?那是因为在这一步string s1 = src.substr(0, pos).data();返回后s1还是 "abcdefg.xxx",std::string_view内部只是简单地封装原始字符串的起始位置和结束位置, 相当于给字符串设置了一个观察窗口,用户只能看到通过窗口能看到的那部分数据. data()成员返回的是char*的指针, 是std::string_view内部字符串的起始位置. 所以其表现再来的行为跟C字符串一样了, 直到遇到空字符串才结束。
3.3.std::string_view和临时字符串
std::string_view并不拥有其指向内容的所有权,用Rust的术语来说,它仅仅是暂时borrow(借用)了它。如果拥有者提前释放了,你还在使用这些内容,那会出现内存问题,这跟悬挂指针(dangling pointer)或悬挂引用(dangling references)很像。Rust专门有套机制在编译时分析变量的生命期,保证borrow的资源在使用期间不会被释放,但C++没有这样的检查,需要人工保证。下面列出一些典型的问题情况:
std::string_view sv = std::string{"hello world"};
string_view foo() {
std::string s{"hello world"};
return string_view{s};
}
auto id(std::string_view sv) { return sv; }
int main() {
std::string s = "hello";
auto sv = id(s + " world");
}
警告:永远不要使用std::string_view保存临时字符串的视图。
4.总结
std::string_view的优点:
1)高效性:std:string_view主要用于提供字符串的视图(view),使std::string_view拷贝字符串的过程非常高效,永远不会创建字符串的任何副本,不像std::string会效率低下且导致内存开销。std::string_view不拥有字符串数据,它仅提供对现有字符串的视图或引用(view or reference)。这使得它适合需要访问或处理字符串而无需内存分配或重新分配开销的场景,特别是在处理大量字符串时非常有用。
2)安全性:由于stdstring_view是只读的,因此它不能被用来修改字符串。这使得它成为一个安全的工具,可以防止由于修改字符串而导致的错误。
3) 灵活性:stdstring_view可以轻松地与各种字符串类型一起使用包括std::string、字符数组和字符指针。这使得它成为处理字符串的灵活工具。
当然任何事物都有它的两面性,它也有些不足,在一些复杂的场景的,人工是很难保证指向的内容的生命周期足够长。所以,推荐的使用方式:仅仅作为函数参数,因为如果该参数仅仅在函数体内使用而不传递出去,这样使用是安全的。
参考:std::basic_string_view - cppreference.com