datasette-graphql 0.14

Datasette plugin providing an automatic GraphQL API for your SQLite databases

datasette-graphql

Datasette plugin providing an automatic GraphQL API for your SQLite databases

Read more about this project: GraphQL in Datasette with the new datasette-graphql plugin

Try out a live demo at datasette-graphql-demo.datasette.io/graphql

• Installation
• Usage
  • Querying for tables and columns
  • Fetching a single record
  • Accessing nested objects
  • Accessing related objects
  • Filtering tables
  • Sorting
  • Pagination
  • Search
  • Columns containing JSON strings
  • Auto camelCase
  • CORS
• The graphql() template function
• Still to come
• Development

Installation

Install this plugin in the same environment as Datasette.

$ pip install datasette-graphql

Usage

This plugin sets up /graphql as a GraphQL endpoint for the first attached database.

If you have multiple attached databases each will get its own endpoint at /graphql/name_of_database.

The automatically generated GraphQL schema is available at /graphql/name_of_database.graphql - here's an example.

Querying for tables and columns

Individual tables (and SQL views) can be queried like this:

{
  repos {
    nodes {
      id
      full_name
      description
    }
  }
}

In this example query the underlying database table is called repos and its columns include id, full_name and description.

Fetching a single record

If you only want to fetch a single record - for example if you want to fetch a row by its primary key - you can use the tablename_row field:

{
  repos_row(id: 107914493) {
    id
    full_name
    description
  }
}

The tablename_row field accepts the primary key column (or columns) as arguments. It also supports the same filter:, search:, sort: and sort_desc: arguments as the tablename field, described below.

Accessing nested objects

If a column is a foreign key to another table, you can request columns from the table pointed to by that foreign key using a nested query like this:

{
  repos {
    nodes {
      id
      full_name
      owner {
        id
        login
      }
    }
  }
}

Accessing related objects

If another table has a foreign key back to the table you are accessing, you can fetch rows from that related table.

Consider a users table which is related to repos - a repo has a foreign key back to the user that owns the repository. The users object type will have a repos_list field which can be used to access those related repos:

{
  users(first: 1, search:"simonw") {
    nodes {
      name
      repos_list(first: 5) {
        totalCount
        nodes {
          full_name
        }
      }
    }
  }
}

Filtering tables

You can filter the rows returned for a specific table using the filter: argument. This accepts a filter object mapping columns to operations. For example, to return just repositories with the Apache 2 license and more than 10 stars:

{
  repos(filter: {license: {eq: "apache-2.0"}, stargazers_count: {gt: 10}}) {
    nodes {
      full_name
      stargazers_count
      license {
        key
      }
    }
  }
}

See table filters examples for more operations, and column filter arguments in the Datasette documentation for details of how those operations work.

These same filters can be used on nested relationships, like so:

{
  users_row(id: 9599) {
    name
    repos_list(filter: {name: {startswith: "datasette-"}}) {
      totalCount
      nodes {
        full_name
      }
    }
  }
}

The where: argument can be used as an alternative to filter: when the thing you are expressing is too complex to be modeled using a filter expression. It accepts a string fragment of SQL that will be included in the WHERE clause of the SQL query.

{
  repos(where: "name='sqlite-utils' or name like 'datasette-%'") {
    totalCount
    nodes {
      full_name
    }
  }
}

Sorting

You can set a sort order for results from a table using the sort: or sort_desc: arguments. The value for this argument should be the name of the column you wish to sort (or sort-descending) by.

{
  repos(sort_desc: stargazers_count) {
    nodes {
      full_name
      stargazers_count
    }
  }
}

Pagination

By default the first 10 rows will be returned. You can control this using the first: argument.

{
  repos(first: 20) {
    totalCount
    pageInfo {
      hasNextPage
      endCursor
    }
    nodes {
      full_name
      stargazers_count
      license {
        key
      }
    }
  }
}

The totalCount field returns the total number of records that match the query.

Requesting the pageInfo.endCursor field provides you with the value you need to request the next page. You can pass this to the after: argument to request the next page.

{
  repos(first: 20, after: "134874019") {
    totalCount
    pageInfo {
      hasNextPage
      endCursor
    }
    nodes {
      full_name
      stargazers_count
      license {
        key
      }
    }
  }
}

The hasNextPage field tells you if there are any more records.

Search

If a table has been configured to use SQLite full-text search you can execute searches against it using the search: argument:

{
  repos(search: "datasette") {
    totalCount
    pageInfo {
      hasNextPage
      endCursor
    }
    nodes {
      full_name
      description
    }
  }
}

The sqlite-utils Python library and CLI tool can be used to add full-text search to an existing database table.

Columns containing JSON strings

If your table has a column that contains data encoded as JSON, datasette-graphql will make that column available as an encoded JSON string. Clients calling your API will need to parse the string as JSON in order to access the data.

You can return the data as a nested structure by configuring that column to be treated as a JSON column. The plugin configuration for that looks like this:

{
    "databases": {
        "test": {
            "tables": {
                "repos": {
                    "plugins": {
                        "datasette-graphql": {
                            "json_columns": [
                                "tags"
                            ]
                        }
                    }
                }
            }
        }
    }
}

Auto camelCase

The names of your columns and tables default to being matched by their representations in GraphQL.

If you have tables with names_like_this you may want to work with them in GraphQL using namesLikeThis, for consistency with GraphQL and JavaScript conventions.

You can turn on automatic camelCase using the "auto_camelcase" plugin configuration setting in metadata.json, like this:

{
    "plugins": {
        "datasette-graphql": {
            "auto_camelcase": true
        }
    }
}

CORS

This plugin obeys the --cors option passed to the datasette command-line tool. If you pass --cors it adds the following CORS HTTP headers to allow JavaScript running on other domains to access the GraphQL API:

access-control-allow-headers: content-type
access-control-allow-method: POST
access-control-allow-origin: *

The graphql() template function

The plugin also makes a Jinja template function available called graphql(). You can use that function in your Datasette custom templates like so:

{% set users = graphql("""
{
    users {
        nodes {
            name
            points
            score
        }
    }
}
""")["users"] %}
{% for user in users.nodes %}
    <p>{{ user.name }} - points: {{ user.points }}, score = {{ user.score }}</p>
{% endfor %}

The function executes a GraphQL query against the generated schema and returns the results. You can assign those results to a variable in your template and then loop through and display them.

By default the query will be run against the first attached database. You can use the optional second argument to the function to specify a different database - for example, to run against an attached github.db database you would do this:

{% set user = graphql("""
{
    users_row(id:9599) {
        name
        login
        avatar_url
    }
}
""", "github")["users_row"] %}

<h1>Hello, {{ user.name }}</h1>

Still to come

See issues for a full list. Planned improvements include:

• Canned query support
• Ability to allowlist specific tables, views and canned queries

Development

To set up this plugin locally, first checkout the code. Then create a new virtual environment:

cd datasette-graphql
python3 -mvenv venv
source venv/bin/activate

Or if you are using pipenv:

pipenv shell

Now install the dependencies and tests:

pip install -e '.[test]'

To run the tests:

pytest