How to read trailer headers in boost::beast::http

Question

As I understand it, in HTTP 1.1 in case the response comes in the form of Transfer-Encoding: chunked, the response may also contain trailers (additional headers) after the body.

I have two questions about this in boost::beast context:

1. When reading the response using the functions boost::beast::http::read or boost::beast::http::async_read, will the trailers also be read from the network buffer?
2. How to access the trailers, after read/async_read if they exist?

Answer 1

Reading trailers is uncommon. MDN:

Warning: Developers cannot access HTTP trailers via the Fetch API or XHR. Additionally, browsers ignore HTTP trailers, with the exception of Server-Timing. See Browser compatibility for more information.

That said, Beast message parsers appear to simply treat trailers as regular fields:

case state::fields:
    parse_fields(p, n, ec);
    if(ec)
        goto done;

// ...
case state::trailer_fields:
    parse_fields(p, n, ec);
    if(ec)
        goto done;

Live demo:

#include <boost/asio.hpp>
#include <boost/beast.hpp>
#include <iostream>
#include <thread>

namespace beast = boost::beast;
namespace http  = beast::http;
namespace net   = boost::asio;
using tcp       = net::ip::tcp;

static std::string_view constexpr g_request = "GET / HTTP/1.1\r\n"
                                              "Host: localhost:7878\r\n"
                                              "TE: trailers\r\n"
                                              "Connection: close\r\n"
                                              "\r\n";

static std::string_view constexpr g_response = "HTTP/1.1 200 OK\r\n"
                                               "Trailer: SeheStackoverflow\r\n"
                                               "Transfer-Encoding: chunked\r\n"
                                               "\r\n"
                                               "4\r\n"
                                               "Wiki\r\n"
                                               "7\r\n"
                                               "pedia i\r\n"
                                               "B\r\n"
                                               "n \r\nchunks.\r\n"
                                               "0\r\n"
                                               "SeheStackoverflow: was here\r\n"
                                               "\r\n";

static void dummy_server() {
    net::io_context ioc;
    tcp::acceptor   acceptor(ioc, {{}, 7878});
    auto            s = acceptor.accept();

    http::request<http::empty_body> req;
    beast::flat_buffer buffer;
    http::read(s, buffer, req);

    write(s, net::buffer(g_response));
}

int main() try {
    std::jthread srv(dummy_server);
    std::this_thread::sleep_for(std::chrono::seconds(1));

    net::io_context   ioc;
    beast::tcp_stream stream(ioc);
    stream.connect({{}, 7878});

    // send request
    write(stream, net::buffer(g_request));

    std::cout << "Request written" << std::endl;

    // read response
    beast::flat_buffer buffer;
    http::response_parser<http::string_body> parser;
    auto cb = [](std::uint64_t n, beast::string_view extensions, beast::error_code& ec) {
        std::cout << "Chunk header received: <" << n << "> bytes, extensions: <" << extensions << ">, ec: <"
                  << ec.message() << ">\n";
    };
    parser.on_chunk_header(cb);
    http::read(stream, buffer, parser);

    auto res = parser.release();
    std::cout << "\n --- Response Headers: " << res.base() << std::endl;
    std::cout << "\n --- Response body: " << res.body() << std::endl;
    std::cout << "\n --- Response status: " << res.result_int() << " " << res.reason()
              << " parser.is_done() = " << parser.is_done() << "\n";

    //for (auto const& field : res)
        //std::cout << field.name_string() << ": " << field.value() << "\n";
} catch (std::exception const& e) {
    std::cerr << "Error: " << e.what() << "\n";
    return 1;
}

Prints

Request written
Chunk header received: <4> bytes, extensions: <>, ec: <Success>
Chunk header received: <7> bytes, extensions: <>, ec: <Success>
Chunk header received: <11> bytes, extensions: <>, ec: <Success>
Chunk header received: <0> bytes, extensions: <>, ec: <Success>

 --- Response Headers: HTTP/1.1 200 OK
Trailer: SeheStackoverflow
Transfer-Encoding: chunked
SeheStackoverflow: was here



 --- Response body: Wikipedia in 
chunks.

 --- Response status: 200 OK parser.is_done() = 1

Recap

1. yes
2. by accessing the header fields as normal