typedef BasicStringPiece<std::string> StringPiece;
template <typename STRING_TYPE> class BasicStringPiece {
public:
// Standard STL container boilerplate.
typedef size_t size_type;
typedef typename STRING_TYPE::value_type value_type;
typedef const value_type* pointer;
typedef const value_type& reference;
typedef const value_type& const_reference;
typedef ptrdiff_t difference_type;
typedef const value_type* const_iterator;
typedef std::reverse_iterator<const_iterator> const_reverse_iterator;
static const size_type npos;
...
protected:
const value_type* ptr_;
size_type length_;
};
BasicStringPiece() : ptr_(NULL), length_(0) {}
BasicStringPiece(const value_type* str)
: ptr_(str),
length_((str == NULL) ? 0 : STRING_TYPE::traits_type::length(str)) {}
BasicStringPiece(const STRING_TYPE& str)
: ptr_(str.data()), length_(str.size()) {}
BasicStringPiece(const value_type* offset, size_type len)
: ptr_(offset), length_(len) {}
BasicStringPiece(const BasicStringPiece& str, size_type pos, size_type len = npos)
: ptr_(str.data() + pos), length_(std::min(len, str.length() - pos)) {}
BasicStringPiece(const typename STRING_TYPE::const_iterator& begin,
const typename STRING_TYPE::const_iterator& end)
: ptr_((end > begin) ? &(*begin) : NULL),
length_((end > begin) ? (size_type)(end - begin) : 0) {}
STRING_TYPE是模板参数类型,对StringPiece而言,STRING_TYPE也就是std::string
value_type是STRING_TYPE::value_type的别名,对StringPiece而言,也就是std::string::value_type ,即 char类型。
构造函数传入的如果是const char*,不指定长度的时候。StringPiece会调用string::traits_type::length()函数来计算长度,也就是std::char_traits::length
// char_traits::length
#include <iostream> // std::cout
#include <string> // std::char_traits
int main ()
{
const char * foo = "String literal";
std::cout << "foo has a length of ";
std::cout << std::char_traits<char>::length(foo);
std::cout << " characters.\n";
return 0;
}
返回的是'\0'结尾的字符串的长度。如果你传入的const char*类型的字符串中间有'\0'会被截断。
https://cplusplus.com/reference/string/char_traits/length/
容器相关函数
const value_type* data() const { return ptr_; }
size_type size() const { return length_; }
size_type length() const { return length_; }
bool empty() const { return length_ == 0; }
...
size_type max_size() const { return length_; }
size_type capacity() const { return length_; }
void clear() {
ptr_ = NULL;
length_ = 0;
}
size()、length()、max_size()、capacity()这4个函数返回的都是length_,他们的值是相同的。这点和std::string是不同的。
数据修改函数
BasicStringPiece& assign(const BasicStringPiece& str, size_type pos, size_type len = npos) {
ptr_ = str.data() + pos;
length_ = std::min(len, str.length() - pos);
return *this;
}
void set(const value_type* data, size_type len) {
ptr_ = data;
length_ = len;
}
void set(const value_type* str) {
ptr_ = str;
length_ = str ? STRING_TYPE::traits_type::length(str) : 0;
}
void remove_prefix(size_type n) {
ptr_ += n;
length_ -= n;
}
void remove_suffix(size_type n) {
length_ -= n;
}
void trim_spaces() {
size_t nsp = 0;
for (; nsp < size() && isspace(ptr_[nsp]); ++nsp) {}
remove_prefix(nsp);
nsp = 0;
for (; nsp < size() && isspace(ptr_[size()-1-nsp]); ++nsp) {}
remove_suffix(nsp);
}
修改其他字符串的函数
void CopyToString(STRING_TYPE* target) const {
internal::CopyToString(*this, target);
}
// 将StringPiece指向的这段字符串追加到target指向的位置
void AppendToString(STRING_TYPE* target) const {
internal::AppendToString(*this, target);
}
// 将StringPiece指向的这段字符串从pos位置开始复制n个字符到buf开始的位置中
size_type copy(value_type* buf, size_type n, size_type pos = 0) const {
return internal::copy(*this, buf, n, pos);
}
CopyToString()本质是调用的std::string的assign函数。AppendToString()本质是调用的std::string的append()函数。copy本质调用的是memcpy()。
数据访问函数
value_type operator[](size_type i) const { return ptr_[i]; }
char front() const { return *ptr_; }
char back() const { return *(ptr_ + length_ - 1); }
// Return the first/last character, 0 when StringPiece is empty.
char front_or_0() const { return length_ ? *ptr_ : '\0'; }
char back_or_0() const { return length_ ? *(ptr_ + length_ - 1) : '\0'; }
std::string支持const和非const两个重载,而StringPiece只有const版本。也就是无法通过[]来修改原类型的。其二就是std::string返回的都是引用(const版本是const引用),而StringPiece返回的值类型。这也不难理解,主要目的就是让Stringpiece的[]返回的单个字符和原字符串的生命周期解耦。(StringPiece内心OS:整个字符串的生命周期都是你的了,单个字符还是我说了算吧,不然也太不安全了)
支持front()、back()直接返回首尾元素,同样是const函数,且返回的是值。和普通std::string不同。比std::string多两个函数:front_or_0()和back_or_0(),他们是当字符串长度为0的时候,返回一个’\0’字符。
const_iterator begin() const { return ptr_; }
const_iterator end() const { return ptr_ + length_; }
const_reverse_iterator rbegin() const {
return const_reverse_iterator(ptr_ + length_);
}
const_reverse_iterator rend() const {
return const_reverse_iterator(ptr_);
}
比较函数
// Does "this" start with "x"
bool starts_with(const BasicStringPiece& x) const {
return ((this->length_ >= x.length_) &&
(wordmemcmp(this->ptr_, x.ptr_, x.length_) == 0));
}
// Does "this" end with "x"
bool ends_with(const BasicStringPiece& x) const {
return ((this->length_ >= x.length_) &&
(wordmemcmp(this->ptr_ + (this->length_-x.length_),
x.ptr_, x.length_) == 0));
}
static int wordmemcmp(const value_type* p,
const value_type* p2,
size_type N) {
return STRING_TYPE::traits_type::compare(p, p2, N);
}
static int compare (const char_type* p, const char_type* q, size_t n) {
while (n--) {if (!eq(*p,*q)) return lt(*p,*q)?-1:1; ++p; ++q;}
return 0;
}https://cplusplus.com/reference/string/char_traits/compare/
bool operator==(const StringPiece& x, const StringPiece& y) {
if (x.size() != y.size())
return false;
return StringPiece::wordmemcmp(x.data(), y.data(), x.size()) == 0;
}
inline bool operator<(const StringPiece& x, const StringPiece& y) {
const int r = StringPiece::wordmemcmp(
x.data(), y.data(), (x.size() < y.size() ? x.size() : y.size()));
return ((r < 0) || ((r == 0) && (x.size() < y.size())));
}
查找函数
// find: Search for a character or substring at a given offset.
size_type find(const BasicStringPiece<STRING_TYPE>& s,
size_type pos = 0) const {
return internal::find(*this, s, pos);
}
size_type find(value_type c, size_type pos = 0) const {
return internal::find(*this, c, pos);
}
// rfind: Reverse find.
size_type rfind(const BasicStringPiece& s,
size_type pos = BasicStringPiece::npos) const {
return internal::rfind(*this, s, pos);
}
size_type rfind(value_type c, size_type pos = BasicStringPiece::npos) const {
return internal::rfind(*this, c, pos);
}
size_t find(const StringPiece& self, const StringPiece& s, size_t pos) {
return findT(self, s, pos);
}
size_t find(const StringPiece& self, char c, size_t pos) {
return findT(self, c, pos);
}
template<typename STR>
size_t findT(const BasicStringPiece<STR>& self,
const BasicStringPiece<STR>& s,
size_t pos) {
if (pos > self.size())
return BasicStringPiece<STR>::npos;
typename BasicStringPiece<STR>::const_iterator result =
std::search(self.begin() + pos, self.end(), s.begin(), s.end());
const size_t xpos =
static_cast<size_t>(result - self.begin());
return xpos + s.size() <= self.size() ? xpos : BasicStringPiece<STR>::npos;
}
template<typename STR>
size_t findT(const BasicStringPiece<STR>& self,
typename STR::value_type c,
size_t pos) {
if (pos >= self.size())
return BasicStringPiece<STR>::npos;
typename BasicStringPiece<STR>::const_iterator result =
std::find(self.begin() + pos, self.end(), c);
return result != self.end() ?
static_cast<size_t>(result - self.begin()) : BasicStringPiece<STR>::npos;
}
可以看出对于查找字符串其本质是使用的STL的search函数,而查找单个字符的时候使用的是STL的find()函数。不过对于rfind()则并不相同。它的实现如下:
template<typename STR>
size_t rfindT(const BasicStringPiece<STR>& self,
const BasicStringPiece<STR>& s,
size_t pos) {
if (self.size() < s.size())
return BasicStringPiece<STR>::npos;
if (s.empty())
return std::min(self.size(), pos);
typename BasicStringPiece<STR>::const_iterator last =
self.begin() + std::min(self.size() - s.size(), pos) + s.size();
typename BasicStringPiece<STR>::const_iterator result =
std::find_end(self.begin(), last, s.begin(), s.end());
return result != last ?
static_cast<size_t>(result - self.begin()) : BasicStringPiece<STR>::npos;
}
template<typename STR>
size_t rfindT(const BasicStringPiece<STR>& self,
typename STR::value_type c,
size_t pos) {
if (self.size() == 0)
return BasicStringPiece<STR>::npos;
for (size_t i = std::min(pos, self.size() - 1); ;
--i) {
if (self.data()[i] == c)
return i;
if (i == 0)
break;
}
return BasicStringPiece<STR>::npos;
}
对于字符串的反向查找本质的调用的STL的find_end。而对于字符的反向查找是完全手写实现的。
另外StringPiece也支持std::string中的find_first_of()、find_first_not_of()、find_last_of()、find_last_not_of()。这里只展开一个:
size_t find_first_of(const StringPiece& self,
const StringPiece& s,
size_t pos) {
if (self.size() == 0 || s.size() == 0)
return StringPiece::npos;
// Avoid the cost of BuildLookupTable() for a single-character search.
if (s.size() == 1)
return find(self, s.data()[0], pos);
bool lookup[UCHAR_MAX + 1] = { false };
BuildLookupTable(s, lookup);
for (size_t i = pos; i < self.size(); ++i) {
if (lookup[static_cast<unsigned char>(self.data()[i])]) {
return i;
}
}
return StringPiece::npos;
}
inline void BuildLookupTable(const StringPiece& characters_wanted,
bool* table) {
const size_t length = characters_wanted.length();
const char* const data = characters_wanted.data();
for (size_t i = 0; i < length; ++i) {
table[static_cast<unsigned char>(data[i])] = true;
}
}
映射表用数组实现,数组下标就是对应字符的数值,value是是否在StringPiece中出现过。后面就是普通的遍历find了。其他的其他函数实现也是类似。
截取子串
StringPiece substr(const StringPiece& self,
size_t pos,
size_t n) {
return substrT(self, pos, n);
}
template<typename STR>
BasicStringPiece<STR> substrT(const BasicStringPiece<STR>& self,
size_t pos,
size_t n) {
if (pos > self.size()) pos = self.size();
if (n > self.size() - pos) n = self.size() - pos;
return BasicStringPiece<STR>(self.data() + pos, n);
}
返回string对象
STRING_TYPE as_string() const {
// std::string doesn't like to take a NULL pointer even with a 0 size.
return empty() ? STRING_TYPE() : STRING_TYPE(data(), size());
}
从StringPiece到string_view
在各个C++开源项目提供了不同版本StringPiece的许多年以后,事情开始有了变化。C++14开始,string_view作为实验功能被进入到C++标准。<experimental/string_view>头文件中有std::experimental::string_view 这一新增的字符串视图类型。C++17 开始string_view顺利转正。头文件<string_view>正式纳入标注,std::string_view类型进入大家的视野。
API的差异
比如string_view没有assign()、clear()以及as_string()函数。访问函数增加了at(),和一般的string类似。
string_view有front()和back(),但没有front_or_0()、back_or_0()。并且front()和back()返回的是 constexpr const_reference。当string_view为空的时候,此时调用front()和back()是未定义行为(UB)!这点要注意。
C++20开始string_view加入了starts_with()和ends_with()。C++23中多了一个contains()可以用来查找一个const char*、string_view、char是否在当前string_view中,返回一个bool类型。而StringPiece中没有contains()。
最主要的是string_view在C++的语法层面,增加了一个运算符sv
#include <string_view>
#include <iostream>
using namespace std::literals;
int main() {
auto print = [](std::string_view sv) {
std::cout << "size:" << sv.size() << std::endl;
for (char c: sv) {
std::cout << c << std::endl;
}
std::cout<<"----"<<std::endl;
};
std::string_view s1 = "abc\0\0def";
std::string_view s2 = "abc\0\0def"sv;
print(s1);
print(s2);
print(s1.substr(1, 5));
print(s2.substr(1, 5));
}
输出
size:3
a
b
c
----
size:8
a
b
c
d
e
f
----
size:2
b
c
----
size:5
b
c
d
----
在""双引号表达的C风格字符作为参数,且不指定长度来构造string_view对象的时候,其表现和StringPiece一致,会默认以’\0’结尾的字符作为string_view的有效字符串。而加了sv后缀以后则能将整个字符串视为有效字符串。
没有享受到C++14或17编译器的小伙伴们,可以使用boost库中的boost::string_view
https://zhuanlan.zhihu.com/p/98829229