How do you clear a stringstream variable?

You can clear the error state and empty the stringstream all in one line

std::stringstream().swap(m); // swap m with a default constructed stringstream

This effectively resets m to a default constructed state, meaning that it actually deletes the buffers allocated by the string stream and resets the error state. Here's an experimental proof:

int main ()
{
    std::string payload(16, 'x');
    
    std::stringstream *ss = new std::stringstream; // Create a memory leak
    (*ss) << payload;                              // Leak more memory
    
    // Now choose a way to "clear" a string stream
    //std::stringstream().swap(*ss); // Method 1
    //ss->str(std::string());        // Method 2
    
    std::cout << "end" << std::endl;
}

Demo

When the demo is compiled with address sanitizer, memory usage is revealed:

=================================================================
==10415==ERROR: LeakSanitizer: detected memory leaks

Direct leak of 392 byte(s) in 1 object(s) allocated from:
    #0 0x510ae8 in operator new(unsigned long) (/tmp/1637178326.0089633/a.out+0x510ae8)
    #1 0x514e80 in main (/tmp/1637178326.0089633/a.out+0x514e80)
    #2 0x7f3079ffb82f in __libc_start_main /build/glibc-Cl5G7W/glibc-2.23/csu/../csu/libc-start.c:291

Indirect leak of 513 byte(s) in 1 object(s) allocated from:
    #0 0x510ae8 in operator new(unsigned long) (/tmp/1637178326.0089633/a.out+0x510ae8)
    #1 0x7f307b03a25c in std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::reserve(unsigned long) (/usr/local/lib64/libstdc++.so.6+0x13725c)
    #2 0x603000000010  (<unknown module>)

SUMMARY: AddressSanitizer: 905 byte(s) leaked in 2 allocation(s).

Pretty steep if you ask me. To hold just 16bytes of payload, we spent 905 bytes ... string streams are no toy. Memory is allocated in two parts:

• The constructed string stream (392 bytes)
• The extra buffer needed for the payload (513 bytes). The extraneous size has to do with the allocation strategy chosen by the stream and for payloads <= 8 bytes, blocks inside the initial object can be used.

If you enable method 1 (the one shown in this answer) the extra 513 (payload) bytes are reclaimed, because the stream is actually cleared.

If you enable method2 as suggested in the comments or other answers, you can see that all 905 bytes are in use by the time we exit.

In terms of program semantics, one may only care that the stream "appears" and "behaves" as empty, similar to how a vector::clear may leave the capacity untouched but render the vector empty to the user (of course vector would spend just 16 bytes here). Given the memory allocation that string stream requires, I can imagine this approach being often faster. This answer's primary goal is to actually clear the string stream, given that memory consumption that comes with it is no joke. Depending on your use case (number of streams, data they hold, frequency of clearing) you may choose the best approach.

Finally note that it's rarely useful to clear the stream without clearing the error state and all inherited state. The one liner in this answer does both.

This should be the most reliable way regardless of the compiler:

m=std::stringstream();

m.str("");

seems to work.

I am always scoping it:

{
    std::stringstream ss;
    ss << "what";
}

{
    std::stringstream ss;
    ss << "the";
}

{
    std::stringstream ss;
    ss << "heck";
}

my 2 cents:

this seemed to work for me in xcode and dev-c++, I had a program in the form of a menu that if executed iteratively as per the request of a user will fill up a stringstream variable which would work ok the first time the code would run but would not clear the stringstream the next time the user will run the same code. but the two lines of code below finally cleared up the stringstream variable everytime before filling up the string variable. (2 hours of trial and error and google searches), btw, using each line on their own would not do the trick.

//clear the stringstream variable

sstm.str("");
sstm.clear();

//fill up the streamstream variable
sstm << "crap" << "morecrap";

There are many other answers that "work", but they often do unnecessary copies or reallocate memory.

1. Swapping streams means that you need to discard one of them, wasting the memory allocation. Same goes for assigning a default-constructed stream,

2. Assigning to the string in the string buffer (via stringstream::str or stringbuf::str) may lose the buffer already allocated by the string.

The canonical way to clear the string stream would be:

void clear(std::stringstream &stream)
{
   if (stream.rdbuf()) stream.rdbuf()->pubseekpos(0);
}

The canonical way to get the size of the data in the stream's buffer is:

std::size_t availSize() (const std::stringstream& stream)
{
   if (stream.rdbuf())
      return std::size_t(
         stream.rdbuf()->pubseekoff(0, std::ios_base::cur, std::ios_base::out));
   else
      return 0;
}

The canonical way to copy the data from the stream to some other preallocated buffer and then clear it would then be:

std::size_t readAndClear(std::stringstream &stream, void* outBuf, std::size_t outSize)
{
   auto const copySize = std::min(availSize(stream), outSize);
   if (!copySize) return 0; // takes care of null stream.rdbuf()

   stream.rdbuf()->sgetn(outBuf, copySize);
   stream.rdbuf()->pubseekpos(0); // clear the buffer

   return copySize;
}

I intend this to be a canonical answer. Language lawyers, feel free to pitch in.

@Azeem's answer should be considered the best solution here. On the one hand his answer doesn't address the specific words of the OP's question, but on the other hand I think it addresses the intent of the OP's question. Excerpting @Azeem's code:

{
    std::stringstream ss;
    ss << "what";
}

Benefits of using scoping contexts:

• Reusing a variable is a code smell, so you should avoid doing it whenever possible. https://maximilianocontieri.com/code-smell-107-variables-reuse
• When the context scope is entered you're guaranteed to have a new instance that's correctly initialized.
• When the context scope is exited the instance is guaranteed to be destroyed correctly and all the memory is reclaimed.
• It doesn't require knowing how a stringstream is implemented (buffer, pointer, index, etc). If you find that you need to know how a class is implemented, then either the class interface has been poorly designed, or you're just using the class wrong -- and that makes your code poorly designed.
• It relies on the stringstream's constructor and destructor to do the right thing (code that's been thoroughly tested and debugged) instead of relying on code that you just wrote to do the right thing.
• It's easy to type, easy to read, easy to understand: just two braces for each context.

Drawbacks:

• You have to learn about contexts / blocks / scopes. https://en.cppreference.com/w/cpp/language/scope. There, you're finished.
• It seems like it would require extra CPU cycles in order to create a new stringstream and then destroy it once for each context. That's fine, though. Your whole program uses extra CPU cycles. Don't count cycles (or even think about them) until you've proven empirically that you're running out of them.

These 2 functions generate identical code for me on compiler explorer when using clang but with gcc the swap version results in much less assembly. Either way I think this empty initializer method should be considered for being the easiest to read.

void ClearEmptyBraces(std::stringstream& ss) {
    ss = { };
}

void ClearSwap(std::stringstream& ss) {
    std::stringstream().swap(ss);
}

The shortest and the most intuitive one.

m = {};

It's a conceptual problem.

Stringstream is a stream, so its iterators are forward, cannot return. In an output stringstream, you need a flush() to reinitialize it, as in any other output stream.