Upload uint8_t* buffer to AWS S3 without going via filesystem

问题

Disclaimers: I am not a c++ programmer, please save me from myself.

I'm trying to create a PutObjectRequest in c++ using the AWS SDK.

I have a 'uint8_t*' (in Java-land from whence I hail we call this a byte[], I believe on planet c++ this is a buffer), and I need to get it into an Aws::IOStream somehow.

All of the examples show the data coming directly from the filesystem.

I've seen a couple of similar-ish (but not really) questions with answers that point to another third party library called Boost, but surely this is a common usecase? Why would I need a third party library to do something that should just be possible using the AWS SDK?:

"I have data, I want to put it up on S3. No it's not in the filesystem, yes I created it in memory."

uint8_t* buf; //<-- How do I get this...
...
Aws::S3::Model::PutObjectRequest object_request;
object_request.WithBucket(output_bucket).WithKey(key_name);

object_request.SetBody(data); //<-- ...into here

I really appreciate any help or pointers (no pun intended) here.

Update
I've tried everything in the comments, and this:

std::shared_ptr<Aws::IOStream> objectStream = Aws::MakeShared<Aws::StringStream>("PutObjectInputStream");        
*objectStream << data;
objectStream->flush();
object_request.SetBody(objectStream);

and this:

std::shared_ptr<Aws::IOStream> objectStream = Aws::MakeShared<Aws::StringStream>("PutObjectInputStream");        
std::istringstream is((char*) data);
*objectStream << is.rdbuf();
objectStream->flush();
object_request.SetBody(objectStream);

which compile, but each only uploads 2 bytes of data.

Other thing I've tried that don't compile are:

auto input_data = Aws::MakeShared<Aws::IOStream>("PutObjectInputStream", std::istringstream((char*) data), std::ios_base::in | std::ios_base::binary);
object_request.SetBody(input_data);

and

object_request.SetBody(std::make_shared<std::istringstream>( std::istringstream( (char*) spn ) ));

and these ones creates the object on S3, but with 0 bytes:

    std::shared_ptr<Aws::IOStream> objectStream = Aws::MakeShared<Aws::StringStream>("PutObjectInputStream");
    objectStream->rdbuf()->pubsetbuf(static_cast<char*>(reinterpret_cast<char*>(data)), length);
    objectStream->rdbuf()->pubseekpos(length);
    objectStream->seekg(0);
    object_request.SetBody(objectStream);


    std::shared_ptr<Aws::IOStream> objectStream = Aws::MakeShared<Aws::StringStream>("PutObjectInputStream");
    objectStream->rdbuf()->pubsetbuf(reinterpret_cast<char*>(data), length);
    objectStream->rdbuf()->pubseekpos(length);
    objectStream->seekg(0);
    object_request.SetBody(objectStream);

回答1:

And after hours of hacking, here is the answer:

    Aws::S3::Model::PutObjectRequest object_request;
    object_request.WithBucket(output_bucket).WithKey(key_name);
    auto data = Aws::MakeShared<Aws::StringStream>("PutObjectInputStream", std::stringstream::in | std::stringstream::out | std::stringstream::binary);
    data->write(reinterpret_cast<char*>(buffer), length);
    object_request.SetBody(data);

Thanks to Ben Voigt for pointing out that when you make_shared (or MakeShared in my case), you're not actually passing it the data at that point. You're just telling it what T you're making shared.

Also helped by nbubis' answer to this question: const char * to std::basic_iostream

来源：https://stackoverflow.com/questions/48666549/upload-uint8-t-buffer-to-aws-s3-without-going-via-filesystem