Question

String literals are array objects:

typeid("hello").name()   // char [6]

This seems convenient because the size of the string literal (=array) is known at compile time. So why is there no constructor to std::string that takes a reference to an array?

// this would be great
template <int N>
std::string(const char (& array)[N]);

Instead there are constructors that take const char * s, const char * s, size_t n or two InputIterators (eg. const char * begin, const char * end). All of these have drawbacks; the array is implicitly converted to a pointer and the size information is lost, so various methods are used to get it back. Working, but increasingly crazy examples:

// std::string(const char * s) version:
std::string s1("hello");                      // calls std::strlen(s) internally

// std::string(const char * s, size_t n) version:
std::string s2("hello", 5);                   // programmer does std::strlen(s) mentally
std::string s3("hello", sizeof("hello"));     // have to repeat string literal

// macro helper to avoid repeating string literal (ugly and dangerous)
#define STRLIT(x) std::string(x, sizeof(x));  // could be a function I guess
std::string s4 = STRLIT("hello");             // not much improvement (and macros are evil)

// std::string(InputIterator begin, InputIterator end) version:
char tmp[] = "hello";                         // copy array
std::string s5(&tmp[0], &tmp[sizeof(tmp)]);   // so you can reference it twice

// or trust the compiler to return the same address for both literals
std::string s6(&"hello"[0], &"hello"[sizeof("hello")]);   // totally crazy

Answer 1

The key question when considering whether there should be a constructor taking a char const (&)[N] (for some statically determined N) is what the content of the resulting std::string should be. It might be obvious to some what it should be but I don’t think it is. Consider this std::string("abc\0def"):

1. It could be a string with 8 elements, including the two null bytes.
2. It could be a string with 7 elements, excluding the second null bytes.
3. It could be a string with 3 elements, excluding everything starting at the first null byte.

If you use

std::cout << "string='" << "abc\0def" << "'\n";

you get the third option. It seems, this is the only sane alternative…