Setting up a GraphQL schema using an existing Postgres database

Introduction

When you have an existing database with a schema already present, you don’t need to create tables or views or run DDL queries through the Hasura console.

All you need to do is indicate to Hasura GraphQL engine which tables and views you want to expose over GraphQL and how they are connected to each other so that you can query them as a “graph”.

Step 1: Track tables/views

Tracking a table or a view means telling Hasura GraphQL engine that you want to expose that table/view over GraphQL.

To track a table or a view

  1. Head to the Data -> Schema section of the console.
  2. Under the heading Untracked Tables/Views, click on the Track button next to the table/view name.

To track the table and expose it over the GraphQL API, add it to the tables.yaml file in the metadata directory as follows:

- table:
   schema: public
   name: <table name>

Apply the metadata by running:

hasura metadata apply

To track a table and expose it over the GraphQL API, use the pg_track_table metadata API:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
   "type": "pg_track_table",
   "args": {
      "source": "<db_name>",
      "schema": "public",
      "name": "<table_name>"
   }
}

To track all tables and views present in the database

  1. Head to the Data -> Schema section of the console.
  2. Under the heading Untracked Tables/Views, click the Track All button.

To track all tables and expose them over the GraphQL API, add them to the tables.yaml file in the metadata directory as follows:

- table:
   schema: public
   name: <table-name-1>
- table:
   schema: public
   name: <table-name-2>

To automate this, you could add the tables in a loop through a script.

Apply the metadata by running:

hasura metadata apply

To track all tables and expose them over the GraphQL API, use the pg_track_table metadata API:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
  "type": "bulk",
  "args": [
    {
       "type": "pg_track_table",
       "args": {
          "source": "<db_name>",
          "schema": "public",
          "name": "<table-name-1>"
       }
    },
    {
       "type": "pg_track_table",
       "args": {
          "source": "<db_name>",
          "schema": "public",
          "name": "<table-name-2>"
       }
    }
  ]
}

To automate this, you could add the pg_track_table requests to the bulk request in a loop through a script.

Step 2: Track foreign-keys

Tracking a foreign-key means creating a relationship between the tables involved in the foreign-key.

To track a foreign-key between two tables in the database

  1. Head to the Data -> Schema section of the console.
  2. Click on a table involved in the foreign-key and head to the Relationships tab.
  3. You should see a suggested relationship based on the foreign-key. Click Add, give a name to your relationship (this will be the name of the nested object in the GraphQL query), and hit Save to create the relationship.
  4. Repeat with the other table involved in the foreign-key.

To track a relationship and expose it over the GraphQL API, add it to the tables.yaml file in the metadata directory as follows:

Object relationship

- table:
    schema: public
    name: <table name>
  object_relationships:
  - name: <relationship name>
    using:
      foreign_key_constraint_on: <reference column>

Array relationship

- table:
      schema: public
      name: <table name>
   array_relationships:
   - name: <relationship name>
      using:
      foreign_key_constraint_on:
         column: <reference column>
         table:
            schema: public
            name: <reference table name>

Apply the metadata by running:

hasura metadata apply

Object relationship

You can create an object relationship by using the pg_create_object_relationship metadata API:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
   "type": "pg_create_object_relationship",
   "args": {
      "source": "<db_name>",
      "table": "<table name>",
      "name": "<relationship name>",
      "using": {
         "foreign_key_constraint_on": "<reference column>"
      }
   }
}

Array relationship

You can create an array relationship by using the pg_create_array_relationship metadata API:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
   "type": "pg_create_array_relationship",
   "args": {
      "source": "<db_name>",
      "table": "<table name>",
      "name": "<relationship name>",
      "using": {
         "foreign_key_constraint_on" : {
            "table" : "<reference table name>",
            "column" : "<reference column>"
         }
      }
   }
}

To track all the foreign-keys of all tables in the database

  1. Head to the Data -> Schema section of the console.
  2. Under the heading Untracked foreign-key relations, click the Track All button to automatically create relationships based on the foreign-keys.

To track all relationships and expose them over the GraphQL API, add them to the tables.yaml file in the metadata directory as follows:

Object relationship

- table:
    schema: public
    name: <table name>
  object_relationships:
  - name: <relationship name>
    using:
      foreign_key_constraint_on: <reference column>

Array relationship

- table:
    schema: public
    name: <table name>
  array_relationships:
  - name: <relationship name>
    using:
      foreign_key_constraint_on:
        column: <reference column>
        table:
          schema: public
          name: <reference table name>

To automate this, you could add the relationships in a loop through a script.

Apply the metadata by running:

hasura metadata apply

You can create multiple relationships by using the pg_create_object_relationship and the pg_create_array_relationship metadata APIs:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
  "type": "bulk",
  "args": [
    {
      "type": "pg_create_object_relationship",
      "args": {
        "source": "<db_name>",
        "table": "<table name>",
        "name": "<relationship name>",
        "using": {
          "foreign_key_constraint_on": "<reference column>"
        }
      }
    },
    {
      "type": "pg_create_array_relationship",
      "args": {
        "source": "<db_name>",
        "table": "<table name>",
        "name": "<relationship name>",
        "using": {
          "foreign_key_constraint_on" : {
            "table" : "<reference table name>",
            "column" : "<reference column>"
          }
        }
      }
    }
  ]
}

To automate this, you could add the create relationships requests to the bulk request in a loop through a script.

Relationship nomenclature

In this case, Hasura GraphQL engine will automatically generate relationship names (the names of the nested objects in the GraphQL query) based on the table names and the foreign-key names.

The name is generated in the following format:

  • For object relationships: singular of foreignTableName
  • For array relationships: plural of foreignTableName

For example, for the foreign-key article.author_id -> author.id, the relationship names will be author for article table and articles for author table.

In case a field with the generated name already exists, a new name will be generated of the form: camel case of (singular/plural of foreignTableName + _by_ + foreignKeyColumnName)

Note that, this is just an arbitrary naming convention chosen by Hasura to ensure the generation of unique relationship names. You can choose to rename your relationships to anything you wish. You can change the relationship names with a name of your choice as shown in renaming relationships.