GraphQL

Schemathesis provides capabilities for testing GraphQL-based applications:

  • Generating conforming queries & mutations

  • Using default values in generated queries

  • Customizing scalars generation

  • Validating responses based on the presence of the errors field

Usage

This test will load GraphQL schema from https://bahnql.herokuapp.com/graphql, generate queries and mutations for it, send them to the server, and verify that responses are not 5xx.

st run --hypothesis-deadline=None https://bahnql.herokuapp.com/graphql

Or in Python tests:

import schemathesis
from hypothesis import settings

schema = schemathesis.graphql.from_url("https://bahnql.herokuapp.com/graphql")


@schema.parametrize()
@settings(deadline=None)
def test(case):
    case.call_and_validate()

Custom scalars

Standard scalars work out of the box, for custom ones you need to pass custom strategies that generate proper AST nodes:

from hypothesis import strategies as st
import schemathesis
from schemathesis.graphql import nodes

schemathesis.graphql.scalar("Date", st.dates().map(nodes.String))

Such a strategy generates valid dates as strings, for example:

{ getByDate(created: "2000-01-01") }

To simplify AST node generation, use schemathesis.graphql.nodes to generate AST nodes of the desired type. There are ready-to-use factories for common node types. They correspond to the following nodes in the graphql library:

  • String -> graphql.StringValueNode

  • Float -> graphql.FloatValueNode

  • Int -> graphql.IntValueNode

  • Object -> graphql.ObjectValueNode

  • List -> graphql.ListValueNode

  • Boolean -> graphql.BooleanValueNode

  • Enum -> graphql.EnumValueNode

  • Null -> graphql.NullValueNode (a constant, not a function)

They exist because classes like graphql.StringValueNode can’t be directly used in Hypothesis’ map calls due to kwarg-only arguments.

Limitations

At the moment, in CLI it is only possible to select API operations to test via hooks.