This sample requires the Boost.Beast header-only library. If you are using vcpkg, it can install that component for you on-demand. Otherwise, you may need to install it separately if you have a selective installation of Boost.

The server executable listens for HTTP connections on http://127.0.0.1:8080/ (port 8080 of localhost). It is expecting POST requests with a JSON payload in the request body that looks like this:

{
	"query": "GraphQL query document goes here",
	"operationName": "(optional) GraphQL operation name goes here",
	"variables": "(optional) JSON object embedded in a string with GraphQL operation variables"
}

It has a single thread accepting requests, and it uses C++20 coroutines. The GraphQL service which actually resolves the requests is the same Star Wars learning sample in ../learn.

The client executable works like most of the other command line samples, it can take a single argument for the operation name and it will read the GraphQL query document from standard input. It packages the query parameters in variables for a pre-parsed query/query.graphql, and it implements an in-proc GraphQL service with the schema/schema.graphql schema using an asynchronous field resolver. The field resolver packages the parameters into a JSON message and sends it to the server process waiting on http://127.0.0.1:8080/.

When it gets a response, client uses the same pre-parsed query in query/query.graphql to parse the result and return the JSON document in the field returned from the server process.

As long as you start the server process first, the behavior of client should be almost identical to the learn_star_wars executable sample produced by ../learn.