Read into std::string using scanf

this can work

char tmp[101];
scanf("%100s", tmp);
string a = tmp;

There is no situation under which gets() is to be used! It is always wrong to use gets() and it is removed from C11 and being removed from C++14.

scanf() doens't support any C++ classes. However, you can store the result from scanf() into a std::string:

Editor's note: The following code is wrong, as explained in the comments. See the answers by Patato, tom, and Daniel Trugman for correct approaches.

std::string str(100, ' ');
if (1 == scanf("%*s", &str[0], str.size())) {
    // ...
}

I'm not entirely sure about the way to specify that buffer length in scanf() and in which order the parameters go (there is a chance that the parameters &str[0] and str.size() need to be reversed and I may be missing a . in the format string). Note that the resulting std::string will contain a terminating null character and it won't have changed its size.

Of course, I would just use if (std::cin >> str) { ... } but that's a different question.

Problem explained:

You CAN populate the underlying buffer of an std::string using scanf, but(!) the managed std::string object will NOT be aware of the change.

const char *line="Daniel 1337"; // The line we're gonna parse

std::string token;
token.reserve(64); // You should always make sure the buffer is big enough

sscanf(line, "%s %*u", token.data());
std::cout << "Managed string: '" << token
          << " (size = " << token.size() << ")" << std::endl;
std::cout << "Underlying buffer: " << token.data()
          << " (size = " << strlen(token.data()) << ")" << std::endl;

Outputs:

Managed string:  (size = 0)
Underlying buffer: Daniel (size = 6)

So, what happened here? The object std::string is not aware of changes not performed through the exported, official, API.

When we write to the object through the underlying buffer, the data changes, but the string object is not aware of that.

If we were to replace the original call: token.reseve(64) with token.resize(64), a call that changes the size of the managed string, the results would've been different:

const char *line="Daniel 1337"; // The line we're gonna parse

std::string token;
token.resize(64); // You should always make sure the buffer is big enough

sscanf(line, "%s %*u", token.data());
std::cout << "Managed string: " << token
          << " (size = " << token.size() << ")" << std::endl;
std::cout << "Underlying buffer: " << token.data()
          << " (size = " << strlen(token.data()) << ")" << std::endl;

Outputs:

Managed string: Daniel (size = 64)
Underlying buffer: Daniel (size = 6)

Once again, the result is sub-optimal. The output is correct, but the size isn't.

Solution:

If you really want to make do this, follow these steps:

1. Call resize to make sure your buffer is big enough. Use a #define for the maximal length (see step 2 to understand why):

std::string buffer;
buffer.resize(MAX_TOKEN_LENGTH);

2. Use scanf while limiting the size of the scanned string using "width modifiers" and check the return value (return value is the number of tokens scanned):

#define XSTR(__x) STR(__x)
#define STR(__x) #x
...
int rv = scanf("%" XSTR(MAX_TOKEN_LENGTH) "s", &buffer[0]);

3. Reset the managed string size to the actual size in a safe manner:

buffer.resize(strnlen(buffer.data(), MAX_TOKEN_LENGTH));