Introduction ============ Welcome to Ariadne! This guide will introduce you to the basic concepts behind creating GraphQL APIs, and show how Ariadne helps you to implement them with just a little Python code. At the end of this guide you will have your own simple GraphQL API accessible through the browser, implementing a single field that returns a "Hello" message along with a client's user agent. Make sure that you've installed Ariadne using ``pip install ariadne``, and that you have your favorite code editor open and ready. Defining schema --------------- First, we will describe what data can be obtained from our API. In Ariadne this is achieved by defining Python strings with content written in `Schema Definition Language `_ (SDL), a special language for declaring GraphQL schemas. We will start by defining the special type ``Query`` that GraphQL services use as entry point for all reading operations. Next, we will specify a single field on it, named ``hello``, and define that it will return a value of type ``String``, and that it will never return ``null``. Using the SDL, our ``Query`` type definition will look like this:: type_defs = """ type Query { hello: String! } """ The ``type Query { }`` block declares the type, ``hello`` is the field definition, ``String`` is the return value type, and the exclamation mark following it means that the returned value will never be ``null``. Validating schema ----------------- Ariadne provides tiny ``gql`` utility function that takes single argument: GraphQL string, validates it and raises descriptive ``GraphQLSyntaxError``, or returns the original unmodified string if its correct:: from ariadne import gql type_defs = gql(""" type Query { hello String! } """) If we try to run the above code now, we will get an error pointing to our incorrect syntax within our ``type_defs`` declaration:: graphql.error.syntax_error.GraphQLSyntaxError: Syntax Error: Expected :, found Name GraphQL request (3:19) type Query { hello String! ^ } Using ``gql`` is optional; however, without it, the above error would occur during your server's initialization and point to somewhere inside Ariadne's GraphQL initialization logic, making tracking down the error tricky if your API is large and spread across many modules. Resolvers --------- The resolvers are functions mediating between API consumers and the application's business logic. In Ariadne every GraphQL type has fields, and every field has a resolver function that takes care of returning the value that the client has requested. We want our API to greet clients with a "Hello (user agent)!" string. This means that the ``hello`` field has to have a resolver that somehow finds the client's user agent, and returns a greeting message from it. At its simplest, resolver is a function that returns a value:: def resolve_hello(*_): return "Hello..." # What's next? The above code is perfectly valid, with a minimal resolver meeting the requirements of our schema. It takes any arguments, does nothing with them and returns a blank greeting string. Real-world resolvers are rarely that simple: they usually read data from some source such as a database, process inputs, or resolve value in the context of a parent object. How should our basic resolver look to resolve a client's user agent? In Ariadne every field resolver is called with at least two arguments: ``obj`` parent object, and the query's execution ``info`` that usually contains the ``context`` attribute that is GraphQL's way of passing additional information from the application to its query resolvers. The default GraphQL server implementation provided by Ariadne defines ``info.context`` as Python ``dict`` containing a single key named ``request`` containing a request object. We can use this in our resolver:: def resolve_hello(_, info): request = info.context["request"] user_agent = request.headers.get("user-agent", "guest") return "Hello, %s!" % user_agent Notice that we are discarding the first argument in our resolver. This is because ``resolve_hello`` is a special type of resolver: it belongs to a field defined on a root type (`Query`), and such fields, by default, have no parent that could be passed to their resolvers. This type of resolver is called a *root resolver*. Now we need to set our resolver on the ``hello`` field of type ``Query``. To do this, we will use the ``QueryType`` class that sets resolver functions to the ``Query`` type in the schema. First, we will update our imports:: from ariadne import QueryType, gql Next, we will instantiate the ``QueryType`` and set our function as resolver for ``hello`` field using it's field decorator:: # Create QueryType instance for Query type defined in our schema... query = QueryType() # ...and assign our resolver function to its "hello" field. @query.field("hello") def resolve_hello(_, info): request = info.context["request"] user_agent = request.headers.get("user-agent", "guest") return "Hello, %s!" % user_agent Making executable schema ------------------------ Before we can run our server, we need to combine our textual representation of the API's shape with the resolvers we've defined above into what is called an "executable schema". Ariadne provides a function that does this for you:: from ariadne import make_executable_schema You pass it your type definitions and resolvers that you want to use:: schema = make_executable_schema(type_defs, query) In Ariadne the process of adding the Python logic to GraphQL schema is called *binding to schema*, and special types that can be passed to the ``make_executable_schema`` second argument are called *bindables*. ``QueryType`` introduced earlier is one of many *bindables* provided by Ariadne that developers will use when creating their GraphQL APIs. Next chapters will In our first API we are passing only single instance to the ``make_executable_schema``, but most of your future APIs will likely pass list of bindables instead, for example:: make_executable_schema(type_defs, [query, user, mutations, fallback_resolvers]) .. note:: Passing bindables to ``make_executable_schema`` is not required, but will result in your API handling very limited number of use cases: browsing schema types and, if you've defined root resolver, accessing root type's fields. Testing the API --------------- Now we have everything we need to finish our API, with the missing only piece being the http server that would receive the HTTP requests, execute GraphQL queries and return responses. Use an ASGI server like `uvicorn `_, `daphne `_, or `hypercorn `_ to serve your application:: $ pip install uvicorn Create a ``ariadne.asgi.GraphQL`` instance for your schema:: from ariadne.asgi import GraphQL app = GraphQL(schema, debug=True) Run your script with ``uvicorn myscript:app`` (remember to replace ``myscript.py`` with the name of your file!). If all is well, you will see a message telling you that the simple GraphQL server is running on the http://127.0.0.1:8000. Open this link in your web browser. You will see the GraphQL Playground, the open source API explorer for GraphQL APIs. You can enter ``{ hello }`` query on the left, press the big, bright "run" button, and see the result on the right: .. image:: _static/hello-world.png :alt: Your first Ariadne GraphQL in action! :target: _static/hello-world.png Your first GraphQL API build with Ariadne is now complete. Congratulations! Completed code -------------- For reference here is complete code of the API from this guide:: from ariadne import QueryType, gql, make_executable_schema from ariadne.asgi import GraphQL type_defs = gql(""" type Query { hello: String! } """) # Create type instance for Query type defined in our schema... query = QueryType() # ...and assign our resolver function to its "hello" field. @query.field("hello") def resolve_hello(_, info): request = info.context["request"] user_agent = request.headers.get("user-agent", "guest") return "Hello, %s!" % user_agent schema = make_executable_schema(type_defs, query) app = GraphQL(schema, debug=True)