C++: How do null characters work in std::string?

Question

I come from C background, while learning C++ I came across the <string> header file. In C strings would be an array of characters terminated by '\0'.

However, in std::string I found that this is not the case, and on inserting/replacing a null character at any valid index does not trim the string as I would have expected.

string s;

getline(cin, s);

// remove all punctuation 
for(string::size_type i = 0, n = s.size(); i < n; i++)
{
     if(ispunct(s[i]))
         s[i] = '\0';
}

input: Hello, World!!!!

output: Hello World

expected output: Hello

On observing the above behaviour I assumed that strings in C++ are not null terminated. Then I found this question on SO Use of null character in strings (C++) This got me confused.

string s = "Hello\0, World";

cout << s << endl;

output: Hello

expected output: Hello, World

It would be helpful if anyone could explain the reason behind this behaviour.

Possible duplicate of [Can a std::string contain embedded nulls?](http://stackoverflow.com/questions/2845769/can-a-stdstring-contain-embedded-nulls) — Algirdas Preidžius, Mar 07 '17 at 16:45
Think of `std::string` like a `std::vector` that has a hidden extra element which is '\0'. That is an implementation detail though. The only guaranteed way to get a null-terminated array is through `std::string::c_str()`. — zett42, Mar 07 '17 at 16:47
There is more than one kind of string in C++. There is the kind that is fully compatible with C, including null termination. There is another kind which is `std::string` from the `` header. Strings of this second kind are not null terminated arrays of characters. — n. m. could be an AI, Mar 07 '17 at 16:52

IInspectable · Accepted Answer · 2017-03-07T16:55:24.257

10

std::string supports embedded NUL characters^*. The fact that your example code doesn't produce the expected result is, because you are constructing a std::string from a pointer to a zero-terminated string. There is no length information, and the c'tor stops at the first NUL character. s contains Hello, hence the output.

If you want to construct a std::string with an embedded NUL character, you have to use a c'tor that takes an explicit length argument:

std::string s("Hello\0, World", 13);
std::cout << s << std::endl;

produces this output:

Hello, World

* std::string maintains an explicit length member, so it doesn't need to reserve a character to act as the end-of-string sentinel.

edited Mar 07 '17 at 16:55

answered Mar 07 '17 at 16:45

IInspectable

46,945
8
85
181

_you have to use a c'tor that takes an explicit length argument_... [or add the NUL character as an individual element](http://coliru.stacked-crooked.com/a/0cdaff448394c7d8). – zett42 Mar 07 '17 at 16:54
1

@zett42: *"If you want to **construct** a `std::string` with an embedded NUL character"*. You are constructing a string and then modify it. That's a different predicate. Using a c'tor taking an explicit length argument, an iterator pair, or an initializer list are the only ways to construct a `std::string` with embedded NUL characters. – IInspectable Mar 07 '17 at 16:57
Correct. I just gave another example how to stuff a NUL character into an std::string. – zett42 Mar 07 '17 at 17:12

C++: How do null characters work in std::string?

1 Answers1

Linked