Github   Telegram
No Matches
HTTP Flatbuf handler and requests

Before you start

Make sure that you can compile and run core tests and read a basic example Writing your first HTTP server.

Step by step guide

JSON is a nice format, but it does not suit well for high-load applications.

This tutorial shows you how to send and receive Flatbuffers over HTTP using userver.

In this sample we use the samples/flatbuf_service/flatbuffer_schema.fbs Flatbuffers scheme and compile it via the ‘flatc --cpp --gen-object-api --filename-suffix ’.fbs' flatbuffer_schema.fbs` command.

HTTP Flatbuffer handler component

There are two ways to write a handler that deals with Flatbuffers:

We are going to take the second approach. All the Flatbuffers related action happens in the HandleRequestFlatbufThrow method:

#include "flatbuffer_schema.fbs.h"
namespace samples::fbs_handle {
class FbsSumEcho final
: public server::handlers::HttpHandlerFlatbufBase<fbs::SampleRequest,
fbs::SampleResponse> {
// `kName` is used as the component name in static config
static constexpr std::string_view kName = "handler-fbs-sample";
// Component is valid after construction and is able to accept requests
FbsSumEcho(const components::ComponentConfig& config,
: HttpHandlerFlatbufBase(config, context) {}
fbs::SampleResponse::NativeTableType HandleRequestFlatbufThrow(
const server::http::HttpRequest& /*request*/,
const fbs::SampleRequest::NativeTableType& fbs_request,
fbs::SampleResponse::NativeTableType res;
res.sum = fbs_request.arg1 + fbs_request.arg2;
res.echo =;
return res;
} // namespace samples::fbs_handle

HTTP Flatbuffer request

A clients::http::Client is needed to make HTTP requests. It could be obtained from the components::HttpClient component.

class FbsRequest final : public components::LoggableComponentBase {
static constexpr auto kName = "fbs-request";
FbsRequest(const components::ComponentConfig& config,
: LoggableComponentBase(config, context),
task_{utils::Async("requests", [this]() { KeepRequesting(); })} {}
void KeepRequesting() const;
clients::http::Client& http_client_;

After that, we just send the data and validate the response:

void FbsRequest::KeepRequesting() const {
// Fill the payload data
fbs::SampleRequest::NativeTableType payload;
payload.arg1 = 20;
payload.arg2 = 22; = "Hello word";
// Serialize the payload into a std::string
flatbuffers::FlatBufferBuilder fbb;
auto ret_fbb = fbs::SampleRequest::Pack(fbb, &payload);
std::string data(reinterpret_cast<const char*>(fbb.GetBufferPointer()),
// Send it
const auto response = http_client_.CreateRequest()
->post("http://localhost:8084/fbs", std::move(data))
// Response code should be 200 (use proper error handling in real code!)
UASSERT_MSG(response->IsOk(), "Sample should work well in tests");
// Verify and deserialize response
const auto body = response->body_view();
const auto* response_fb =
flatbuffers::Verifier verifier(reinterpret_cast<const uint8_t*>(,
UASSERT_MSG(response_fb->Verify(verifier), "Broken flatbuf in sample");
fbs::SampleResponse::NativeTableType result;
// Make sure that the response is the expected one for sample
UASSERT_MSG(result.sum == 42, "Sample should work well in tests");
UASSERT_MSG(result.echo ==, "Sample should work well in tests");

Build and Run

To build the sample, execute the following build steps at the userver root directory:

mkdir build_release
cd build_release
cmake -DCMAKE_BUILD_TYPE=Release ..
make userver-samples-flatbuf_service

The sample could be started by running make start-userver-samples-flatbuf_service. The command would invoke testsuite start target that sets proper paths in the configuration files and starts the service.

To start the service manually run ./samples/flatbuf_service/userver-samples-flatbuf_service -c </path/to/static_config.yaml> (do not forget to prepare the configuration files!).

Now you can send a request to your server from another terminal:

$ echo "100000000c00180000000800100004000c00000014000000140000000000000016000000000000000a00000048656c6c6f20776f72640000" \
| xxd -r -p | curl --data-binary "@-" http://localhost:8084/fbs -v --output /dev/null
* Connected to localhost (::1) port 8084 (#0)
> POST /fbs HTTP/1.1
> Host: localhost:8084
> User-Agent: curl/7.58.0
> Accept: */*
> Content-Length: 56
> Content-Type: application/x-www-form-urlencoded
* upload completely sent off: 56 out of 56 bytes
< HTTP/1.1 200 OK
< Date: Wed, 16 Jun 2021 12:52:22 UTC
< Content-Type: text/html
< X-YaRequestId: c8b2aee7ca5f4165ad25119b1850e778
< Server: userver/1.0.0 (20210616074040; rv:2c78282ea)
< X-YaTraceId: 8f4765a7176e41d28c3c6a677f00193e
< Connection: keep-alive
< Content-Length: 48
* Connection #0 to host localhost left intact

Functional testing

Naive functional tests for the service could be implemented using the testsuite in the following way:

async def test_flatbuf(service_client):
body = bytearray.fromhex(
response = await'/fbs', data=body)
assert response.status == 200

Full sources

See the full example: