Implement a C++ FIDL client

Prerequisites

This tutorial builds on the FIDL server tutorial. For the full set of FIDL tutorials, refer to the overview.

Overview

This tutorial implements a client for a FIDL protocol and runs it against the server created in the previous tutorial. The client in this tutorial is asynchronous. There is an alternate tutorial for synchronous clients.

Structure of the client example

The example code accompanying this tutorial is located in your Fuchsia checkout at //examples/fidl/cpp/client. It consists of a client component and its containing package. For more information about building components, see Build components.

To get the client component up and running, there are three targets that are defined in //examples/fidl/cpp/client/BUILD.gn:

  1. The raw executable file for the client. This produces a binary with the specified output name that can run on Fuchsia:

    executable("bin") {
      output_name = "fidl_echo_cpp_client"
      sources = [ "main.cc" ]
    
      deps = [
        "//examples/fidl/fuchsia.examples:fuchsia.examples_cpp",
    
        # This library is used to log messages.
        "//sdk/lib/syslog/cpp",
    
        # This library provides an asynchronous event loop implementation.
        "//zircon/system/ulib/async-loop:async-loop-cpp",
    
        # This library is used to consume capabilities, e.g. protocols,
        # from the component's incoming directory.
        "//zircon/system/ulib/service:service-llcpp",
      ]
    }
    
    
  2. A component that is set up to run the client executable. Components are the units of software execution on Fuchsia. A component is described by its manifest file. In this case meta/client.cml configures echo-client as an executable component which runs fidl_echo_cpp_client in :bin.

    fuchsia_component("echo-client") {
      component_name = "echo_client"
      manifest = "meta/client.cml"
      deps = [ ":bin" ]
    }
    
    

    The server component manifest is located at //examples/fidl/cpp/client/meta/client.cml. The binary name in the manifest must match the output name of the executable defined in BUILD.gn.

    {
        include: [ "syslog/client.shard.cml" ],
    
        // Information about the program to run.
        program: {
            // Use the built-in ELF runner.
            runner: "elf",
    
            // The binary to run for this component.
            binary: "bin/fidl_echo_cpp_client",
        },
    
        // Capabilities used by this component.
        use: [
            { protocol: "fuchsia.examples.Echo" },
        ],
    }
    
    
  3. The component is then put into a package, which is the unit of software distribution on Fuchsia. In this case, the package contains a client and a server component, and realm component to to declare the appropriate capabilities and routes.

    # C++ async client and server example package
    fuchsia_package("echo-cpp-client") {
      deps = [
        ":echo-client",
        "//examples/fidl/cpp/server:echo-server",
        "//examples/fidl/echo-realm:echo_realm",
      ]
    }
    
    

Building the client

  1. Add the client to your build configuration. This only needs to be done once:

    fx set core.qemu-x64 --with //examples/fidl/cpp/client
    
  2. Build the client:

    fx build examples/fidl/cpp/client
    

Connect to the protocol

In its main function, the client component connects to the fuchsia.examples/Echo protocol in its namespace .

int main(int argc, const char** argv) {
  // Connect to the |fuchsia.examples/Echo| protocol inside the component's
  // namespace. This can fail so it's wrapped in a |zx::status| and it must be
  // checked for errors.
  zx::status client_end = service::Connect<fuchsia_examples::Echo>();
  if (!client_end.is_ok()) {
    FX_LOGS(ERROR) << "Synchronous error when connecting to the |Echo| protocol: "
                   << client_end.status_string();
    return -1;
  }

  // ...

In parallel, the component manager will route the request to the server component. The server handler implemented in the server tutorial will be called with the server endpoint, binding the channel to the server implementation.

An important point to note here is that this code assumes that the component's namespace already contains an instance of the Echo protocol. When running the example at the end of the tutorial, a realm component is used to route the protocol from the server and offer it to the client component.

Initialize the event loop

An asynchronous client needs an async_dispatcher_t* to asynchronously monitor messages from a channel. The async::Loop provides a dispatcher implementation backed by an event loop.

  // As in the server, the code sets up an async loop so that the client can
  // listen for incoming responses from the server without blocking.
  async::Loop loop(&kAsyncLoopConfigNeverAttachToThread);
  async_dispatcher_t* dispatcher = loop.dispatcher();

The dispatcher is used to run pieces of asynchronous code. It is first used to run the EchoString method, and quits when the response is received. It is then run after calling the SendString in order to listen for events, and quits when an OnString event is received. The call to ResetQuit() in between these two instances allows the client to reuse the loop.

Initialize the client

In order to make Echo requests to the server, initialize a client using the client endpoint from the previous step, the loop dispatcher, as well as an event handler delegate:

  // Define the event handler implementation for the client.
  //
  // The event handler delegate should be an object that implements the
  // |fidl::AsyncEventHandler<Echo>| virtual class, which has methods
  // corresponding to the events offered by the protocol. By default those
  // methods are no-ops.
  class EventHandler : public fidl::AsyncEventHandler<fuchsia_examples::Echo> {
   public:
    void OnString(fidl::Event<::fuchsia_examples::Echo::OnString>& event) override {
      FX_LOGS(INFO) << "(Natural types) got event: " << event.response();
      loop_.Quit();
    }

    // One may also override the |on_fidl_error| method, which is called
    // when the client encounters an error and is going to teardown.
    void on_fidl_error(fidl::UnbindInfo error) override { FX_LOGS(ERROR) << error; }

    explicit EventHandler(async::Loop& loop) : loop_(loop) {}

   private:
    async::Loop& loop_;
  };

  // Create an instance of the event handler.
  EventHandler event_handler{loop};

  // Create a client to the Echo protocol, passing our |event_handler| created
  // earlier. The |event_handler| must live for at least as long as the
  // |client|. The |client| holds a reference to the |event_handler| so it is a
  // bug for the |event_handler| to be destroyed before the |client|.
  fidl::Client client(std::move(*client_end), dispatcher, &event_handler);

Make FIDL calls

The methods to make FIDL calls are exposed behind a dereference operator, such that FIDL calls look like client->EchoString(...).

An asynchronous EchoString call takes a request object, and accepts a callback which is invoked with the result of the call, indicating success or failure:

  client->EchoString({"hello"}).ThenExactlyOnce(
      [&](fidl::Result<fuchsia_examples::Echo::EchoString>& result) {
        // Check if the FIDL call succeeded or not.
        if (!result.is_ok()) {
          // If the call failed, log the error, and crash the program.
          // Production code should do more graceful error handling depending
          // on the situation.
          FX_LOGS(ERROR) << "EchoString failed: " << result.error_value();
          ZX_PANIC("%s", result.error_value().FormatDescription().c_str());
        }
        // Dereference (->) the result object to access the response payload.
        FX_LOGS(INFO) << "(Natural types) got response: " << result->response();
        loop.Quit();
      });
  // Run the dispatch loop, until we receive a reply, in which case the callback
  // above will quit the loop, breaking us out of the |Run| function.
  loop.Run();
  loop.ResetQuit();

You may also use the designated initialization style double braces syntax supported by natural structs and tables:

  client->EchoString({{.value = "hello"}})
      .ThenExactlyOnce(
        // ... callback ...

A one way SendString call doesn't have a reply, so callbacks are not needed. The returned result represents any errors occurred when sending the request.

  // Make a SendString one way call with natural types.
  fitx::result<::fidl::Error> result = client->SendString({"hello"});
  if (!result.is_ok()) {
    FX_LOGS(ERROR) << "SendString failed: " << result.error_value();
    ZX_PANIC("%s", result.error_value().FormatDescription().c_str());
  }

Make calls using wire domain objects

The above tutorial makes client calls with natural domain objects: each call consumes request messages represented using natural domain objects, and returns back replies also in natural domain objects. When optimizing for performance and heap allocation, one may make calls using wire domain objects. To do that, insert a .wire() before the dereference operator used when making calls, i.e. client.wire()->EchoString(...).

Make a EchoString two way call with wire types:

  client.wire()->EchoString("hello").ThenExactlyOnce(
      [&](fidl::WireUnownedResult<fuchsia_examples::Echo::EchoString>& result) {
        if (!result.ok()) {
          FX_LOGS(ERROR) << "EchoString failed: " << result.error();
          ZX_PANIC("%s", result.error().FormatDescription().c_str());
          return;
        }
        fidl::WireResponse<fuchsia_examples::Echo::EchoString>& response = result.value();
        std::string reply(response.response.data(), response.response.size());
        FX_LOGS(INFO) << "(Wire types) got response: " << reply;
        loop.Quit();
      });

Make a SendString one way call with wire types:

  fidl::Status wire_status = client.wire()->SendString("hello");
  if (!wire_status.ok()) {
    FX_LOGS(ERROR) << "SendString failed: " << result.error_value();
    ZX_PANIC("%s", result.error_value().FormatDescription().c_str());
  }

The relevant classes and functions used in a wire client call have similar shapes to those used in a natural client call. When a different class or function is called for, the wire counterpart is usually prefixed with Wire. There are also differences in pointers vs references and argument structure:

  • The EchoString method taking natural domain objects accepts a single argument that is the request domain object:

    client->EchoString({{.value = "hello"}})
        .ThenExactlyOnce(
    

    When the request payload is a struct, the EchoString method taking wire domain objects flattens the list of struct fields in the request body into separate arguments (here, a single fidl::StringView argument):

    client.wire()->EchoString("hello").ThenExactlyOnce(
    
  • The callback in async natural calls accepts a fidl::Result<Method>&:

    client->EchoString({"hello"}).ThenExactlyOnce(
        [&](fidl::Result<fuchsia_examples::Echo::EchoString>& result) {
          // Check if the FIDL call succeeded or not.
          if (!result.is_ok()) {
            // If the call failed, log the error, and crash the program.
            // Production code should do more graceful error handling depending
            // on the situation.
            FX_LOGS(ERROR) << "EchoString failed: " << result.error_value();
            ZX_PANIC("%s", result.error_value().FormatDescription().c_str());
          }
          // Dereference (->) the result object to access the response payload.
          FX_LOGS(INFO) << "(Natural types) got response: " << result->response();
          loop.Quit();
        });
    // Run the dispatch loop, until we receive a reply, in which case the callback
    // above will quit the loop, breaking us out of the |Run| function.
    loop.Run();
    loop.ResetQuit();
    
    • To check for success or error, use the is_ok() or is_error() method.
    • To access the response payload afterwards, use value() or ->.
    • You may move out the result or the payload since these types all implement hierarchical object ownership.

    The callback in async wire calls accepts a fidl::WireUnownedResult<Method>&:

    client.wire()->EchoString("hello").ThenExactlyOnce(
        [&](fidl::WireUnownedResult<fuchsia_examples::Echo::EchoString>& result) {
          if (!result.ok()) {
            FX_LOGS(ERROR) << "EchoString failed: " << result.error();
            ZX_PANIC("%s", result.error().FormatDescription().c_str());
            return;
          }
          fidl::WireResponse<fuchsia_examples::Echo::EchoString>& response = result.value();
          std::string reply(response.response.data(), response.response.size());
          FX_LOGS(INFO) << "(Wire types) got response: " << reply;
          loop.Quit();
        });
    
    • To check for success, use the ok() method.
    • To access the response payload afterwards, use value() or ->.
    • You must synchronously use the result within the callback. The result type is unowned, meaning it only borrows the response allocated somewhere else by the FIDL runtime.
  • One way calls also take the whole request domain object in the natural case, and flatten request struct fields into separate arguments in the wire case:

    // Make a SendString one way call with natural types.
    fitx::result<::fidl::Error> result = client->SendString({"hello"});
    
    fidl::Status wire_status = client.wire()->SendString("hello");
    

Run the client

In order for the client and server to communicate using the Echo protocol, component framework must route the fuchsia.examples.Echo capability from the server to the client. For this tutorial, a realm component is provided to declare the appropriate capabilities and routes.

  1. Configure your build to include the provided package that includes the echo realm, server, and client:

    fx set core.qemu-x64 --with //examples/fidl/cpp/client
    
  2. Build the Fuchsia image:

    fx build
    
  3. Run the echo_realm component. This creates the client and server component instances and routes the capabilities:

    ffx component run fuchsia-pkg://fuchsia.com/echo-cpp-client#meta/echo_realm.cm
    
  4. Start the echo_client instance:

    ffx component start /core/ffx-laboratory:echo_realm/echo_client
    

The server component starts when the client attempts to connect to the Echo protocol. You should see output similar to the following in the device logs (ffx log):

[echo_server][][I] Running C++ echo server with natural types
[echo_server][][I] Incoming connection for fuchsia.examples.Echo
[echo_client][][I] (Natural types) got response: hello
[echo_client][][I] (Natural types) got response: hello
[echo_client][][I] (Natural types) got response: hello
[echo_client][][I] (Natural types) got event: hello
[echo_client][][I] (Wire types) got response: hello
[echo_client][][I] (Natural types) got event: hello
[echo_server][][I] Client disconnected

Terminate the realm component to stop execution and clean up the component instances:

ffx component destroy /core/ffx-laboratory:echo_realm

Wire domain objects only client

fidl::Client supports making calls with both natural domain objects and wire domain objects. If you only need to use wire domain objects, you may create a WireClient that exposes the equivalent method call interface as the subset obtained from calling client.wire() on a fidl::Client.

A WireClient is created the same way as a Client:

fidl::WireClient client(std::move(*client_end), dispatcher, &event_handler);

fidl::Client always exposes received events to the user in the form of natural domain objects. On the other hand, fidl::WireClient will expose received events in the form of wire domain objects. To do that, the event handler passed to a WireClient needs to implement fidl::WireAsyncEventHandler<Protocol>.

  • Implementing a natural event handler:

    class EventHandler : public fidl::AsyncEventHandler<fuchsia_examples::Echo> {
     public:
      void OnString(fidl::Event<::fuchsia_examples::Echo::OnString>& event) override {
        FX_LOGS(INFO) << "(Natural types) got event: " << event.response();
        loop_.Quit();
      }
    
      // ...
    };
    
  • Implementing a wire event handler:

    class EventHandler : public fidl::WireAsyncEventHandler<fuchsia_examples::Echo> {
     public:
      void OnString(fidl::WireEvent<fuchsia_examples::Echo::OnString>* event) override {
        FX_LOGS(INFO) << "(Natural types) got event: " << event->response.get();
        loop_.Quit();
      }
    
      // ...
    };
    

Synchronous call

WireClient objects also allows synchronous calls, which will block until the response is received and return the response object. These may be selected using the .sync() accessor. (e.g. client.sync()->EchoString()).

  // Make a synchronous EchoString call, which blocks until it receives the response,
  // then returns a WireResult object for the response.
  fidl::WireResult result_sync = client.sync()->EchoString("hello");
  ZX_ASSERT(result_sync.ok());
  std::string_view response = result_sync->response.get();
  FX_LOGS(INFO) << "Got synchronous response: " << response;

In synchronous calls, a result object is returned, synchronously communicating the success or failure of the call.

The full example code for using a wire client is located in your Fuchsia checkout at //examples/fidl/cpp/client/wire.