GraphQL

Characteristics

Read-only API
Public access, no authentication needed
Uses GraphQL technology:
- Maximum page size (and the default) is 25 records
- Maximum query depth is set at 8 levels
- A maximum of two collections can be requested within the same query
- Support for GET requests (query must be inside the query string) and POST requests (query must be within the body, encoded as application/json or application/graphql)

GraphQL

The Consul Democracy API uses GraphQL, specifically the Ruby implementation. If you're not familiar with this kind of APIs, we recommended you to check the GraphQL official documentation.

One of the characteristics that differentiates a REST API from a GraphQL one is that with the latter one it's possible for the client to build its own custom queries, so the server will only return information in which we're interested.

GraphQL queries are written following a format which resembles JSON. For example:

{
  proposal(id: 1) {
    id,
    title,
    public_author {
      id,
      username
    }
  }
}

Responses are formatted in JSON:

{
  "data": {
    "proposal": {
      "id": 1,
      "title": "Increase the amount of green spaces",
      "public_author": {
        "id": 2,
        "username": "abetterworld"
      }
    }
  }
}

Making API requests

Following the official recommendations, the Consul Democracy API supports the following kind of requests:

GET requests, with the query inside the query string.
POST requests
- With the query inside the body, with Content-Type: application/json
- With the query inside the body, with Content-Type: application/graphql

Supported clients

Since this is an API that works through HTTP, any tool capable of making this kind of requests is capable of querying the API.

This section presents a few examples about how to make requests using:

GraphiQL
Chrome extensions like Postman
Any HTTP library

GraphiQL

GraphiQL is a browser interface for making queries against a GraphQL API. It's also an additional source of documentation. Consul Democracy uses the graphiql-rails to access this interface at /graphiql; it's the best way to get familiar with GraphQL-based APIs.

It's got three main panels:

The left panel is used to write the query.
The central panel shows the result of the request.
The right panel (hideable) shows a documentation autogenerated from the models and fields exposed in the API.

Postman

Here's an example of a GET request, with the query as part of the query string:

And here's an example of a POST request, with the query as part of the body and encoded as application/json:

The query must be located inside a valid JSON document, as the value of the "query" key:

HTTP libraries

You can use any of the HTTP libraries available for most programming languages.

IMPORTANT: Some servers might use security protocols that will make it necessary to include a User Agent header from a web browser so the request is not rejected. For example:

User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:109.0) Gecko/20100101 Firefox/116.0

Available information

The app/graphql/types/ folder contains a complete list of all the models (and their attributes) which are currently being exposed in the API.

The models are the following:

Model

Description

Examples of queries

Request a single record from a collection

{
  proposal(id: 2) {
    id,
    title,
    comments_count
  }
}

Response:

{
  "data": {
    "proposal": {
      "id": 2,
      "title": "Create a dog-friendly area near the beach",
      "comments_count": 10
    }
  }
}

Request a complete collection

{
  proposals {
    edges {
      node {
        title
      }
    }
  }
}

Response:

{
  "data": {
    "proposals": {
      "edges": [
        {
          "node": {
            "title": "Bus stop near the school"
          }
        },
        {
          "node": {
            "title": "Improve the lighting in the football stadium"
          }
        }
      ]
    }
  }
}

Pagination

The maximum (and default) number of records that each page contains is set to 25. In order to navigate through the different pages, it's necessary to request endCursor information:

{
  proposals(first: 25) {
    pageInfo {
      hasNextPage
      endCursor
    }
    edges {
      node {
        title
      }
    }
  }
}

The response:

{
  "data": {
    "proposals": {
      "pageInfo": {
        "hasNextPage": true,
        "endCursor": "NQ=="
      },
      "edges": [
        # ...
      ]
    }
  }
}

To retrieve the next page, pass the cursor received in the previous request as a parameter:

{
  proposals(first: 25, after: "NQ==") {
    pageInfo {
      hasNextPage
      endCursor
    }
    edges {
      node {
        title
      }
    }
  }
}

Accessing several resources in a single request

This query requests information about several models in a single request: Proposal, User, Geozone and Comment:

{
  proposal(id: 15262) {
    id,
    title,
    public_author {
      username
    },
    geozone {
      name
    },
    comments(first: 2) {
      edges {
        node {
          body
        }
      }
    }
  }
}

Security limitations

Allowing a client to customize queries is a major risk factor. If queries that are too complex were allowed, it would be possible to perform a DoS attack against the server.

There are three main mechanisms to prevent such abuses:

Pagination of results
Limit the maximum depth of the queries
Limit the amount of information that is possible to request in a query

Example of a query which is too deep

The maximum depth of queries is currently set at 8. Deeper queries (such as the following) will be rejected:

{
  user(id: 1) {
    public_proposals {
      edges {
        node {
          id,
          title,
          comments {
            edges {
              node {
                body,
                public_author {
                  username
                }
              }
            }
          }
        }
      }
    }
  }
}

The response will look something like this:

{
  "errors": [
    {
      "message": "Query has depth of 9, which exceeds max depth of 8"
    }
  ]
}

Example of a query which is too complex

The main risk factor is the option to request multiple collections of resources in the same query. The maximum number of collections that can appear in the same query is limited to 2. The following query requests information from the users, debates and proposals collections, so it will be rejected:

{
  users {
    edges {
      node {
        public_debates {
          edges {
            node {
              title
            }
          }
        },
        public_proposals {
          edges {
            node {
              title
            }
          }
        }
      }
    }
  }
}

The response will look something like this:

{
  "errors": [
    {
      "message": "Query has complexity of 3008, which exceeds max complexity of 2500"
    }
  ]
}

However, it is possible to request information belonging to more than two models in a single query, as long as you do not try to access the entire collection. For example, the following query that accesses the User, Proposal and Geozone models is valid:

{
  user(id: 468501) {
    id
    public_proposals {
      edges {
        node {
          title
          geozone {
            name
          }
        }
      }
    }
  }
}

The response:

{
  "data": {
    "user": {
      "id": 468501,
      "public_proposals": {
        "edges": [
          {
            "node": {
              "title": "Make a discount to locals in sports centers",
              "geozone": {
                "name": "South area"
              }
            }
          }
        ]
      }
    }
  }
}

Code examples

Simple example

Here's a simple example of code accessing the Consul Democracy demo API:

require "net/http"

API_ENDPOINT = "https://demo.consuldemocracy.org/graphql".freeze

def make_request(query_string)
  uri = URI(API_ENDPOINT)
  uri.query = URI.encode_www_form(query: query_string)
  request = Net::HTTP::Get.new(uri)
  request[:accept] = "application/json"

  Net::HTTP.start(uri.hostname, uri.port, use_ssl: true) do |https|
    https.request(request)
  end
end

query = <<-GRAPHQL
  {
    proposal(id: 1) {
        id,
        title,
        public_created_at
    }
  }
GRAPHQL

response = make_request(query)

puts "Response code: #{response.code}"
puts "Response body: #{response.body}"

Example with pagination

And here is a more complex example using pagination, once again accessing the Consul Democracy demo API:

require "net/http"
require "json"

API_ENDPOINT = "https://demo.consuldemocracy.org/graphql".freeze

def make_request(query_string)
  uri = URI(API_ENDPOINT)
  uri.query = URI.encode_www_form(query: query_string)
  request = Net::HTTP::Get.new(uri)
  request[:accept] = "application/json"

  Net::HTTP.start(uri.hostname, uri.port, use_ssl: true) do |https|
    https.request(request)
  end
end

def build_query(options = {})
  page_size = options[:page_size] || 25
  page_size_parameter = "first: #{page_size}"

  page_number = options[:page_number] || 0
  after_parameter = page_number.positive? ? ", after: \"#{options[:next_cursor]}\"" : ""

  <<-GRAPHQL
  {
    proposals(#{page_size_parameter}#{after_parameter}) {
      pageInfo {
        endCursor,
        hasNextPage
      },
      edges {
        node {
          id,
          title,
          public_created_at
        }
      }
    }
  }
  GRAPHQL
end

page_number = 0
next_cursor = nil
proposals   = []

loop do
  puts "> Requesting page #{page_number}"

  query = build_query(page_size: 25, page_number: page_number, next_cursor: next_cursor)
  response = make_request(query)

  response_hash  = JSON.parse(response.body)
  page_info      = response_hash["data"]["proposals"]["pageInfo"]
  has_next_page  = page_info["hasNextPage"]
  next_cursor    = page_info["endCursor"]
  proposal_edges = response_hash["data"]["proposals"]["edges"]

  puts "\tHTTP code: #{response.code}"

  proposal_edges.each do |edge|
    proposals << edge["node"]
  end

  page_number += 1

  break unless has_next_page
end

PreviousOAuth NextDebates and proposals recommendations

Last updated 1 month ago