GraphQL API

Understanding the core concepts of the GraphQL API.

In our API, each SQL table is reflected as a set of GraphQL types. At a high level, tables become types and columns/foreign keys become fields on those types.

By default, PostgreSQL table and column names are not inflected when reflecting GraphQL names. For example, an account_holder table has GraphQL type name account_holder. In cases where SQL entities are named using snake_case, enable inflection to match GraphQL/Javascript conventions e.g. account_holder -> AccountHolder.

Individual table, column, and relationship names may also be manually overridden.

Primary Keys (Required)

Every table must have a primary key for it to be exposed in the GraphQL schema. For example, the following Blog table will be available in the GraphQL schema as blogCollection since it has a primary key named id:

1
create table "Blog"(
2
  id serial primary key,
3
  name varchar(255) not null,
4
);

But the following table will not be exposed because it doesn't have a primary key:

1
create table "Blog"(
2
  id int,
3
  name varchar(255) not null,
4
);

QueryType

The Query type is the entrypoint for all read access into the graph.

Node

The node interface allows for retrieving records that are uniquely identifiable by a globally unique nodeId: ID! field. For more information about nodeId, see nodeId.

SQL Setup

1
create table "Blog"(
2
  id serial primary key,
3
  name varchar(255) not null,
4
  description varchar(255),
5
  "createdAt" timestamp not null,
6
  "updatedAt" timestamp not null
7
);

GraphQL Types

1
"""The root type for querying data"""
2
type Query {
3
4
  """Retrieve a record by its `ID`"""
5
  node(nodeId: ID!): Node
6
7
}

To query the node interface effectively, use inline fragments to specify which fields to return for each type.

Example

1
{
2
  node(
3
    nodeId: "WyJwdWJsaWMiLCAiYmxvZyIsIDFd"
4
  ) {
5
    nodeId
6
    # Inline fragment for `Blog` type
7
    ... on Blog {
8
      name
9
      description
10
    }
11
  }
12
}

1
{
2
  "data": {
3
    "node": {
4
      "name": "Some Blog",
5
      "nodeId": "WyJwdWJsaWMiLCAiYmxvZyIsIDFd",
6
      "description": "Description of Some Blog"
7
    }
8
  }
9
}

Collections

Each table has top level entry in the Query type for selecting records from that table. Collections return a connection type and can be paginated, filtered, and sorted using the available arguments.

SQL Setup

1
create table "Blog"(
2
  id serial primary key,
3
  name varchar(255) not null,
4
  description varchar(255),
5
  "createdAt" timestamp not null,
6
  "updatedAt" timestamp not null
7
);

GraphQL Types

1
"""The root type for querying data"""
2
type Query {
3
4
  """A pagable collection of type `Blog`"""
5
  blogCollection(
6
7
    """Query the first `n` records in the collection"""
8
    first: Int
9
10
    """Query the last `n` records in the collection"""
11
    last: Int
12
13
    """Query values in the collection before the provided cursor"""
14
    before: Cursor
15
16
    """Query values in the collection after the provided cursor"""
17
    after: Cursor
18
19
    """
20
    Skip n values from the after cursor. Alternative to cursor pagination. Backward pagination not supported.
21
    """
22
    offset: Int
23
24
    """Filters to apply to the results set when querying from the collection"""
25
    filter: BlogFilter
26
27
    """Sort order to apply to the collection"""
28
    orderBy: [BlogOrderBy!]
29
  ): BlogConnection!
30
}

Connection types are the primary interface to returning records from a collection.

Connections wrap a result set with some additional metadata.

1
type BlogConnection {
2
3
  # Count of all records matching the *filter* criteria
4
  totalCount: Int!
5
6
  # Pagination metadata
7
  pageInfo: PageInfo!
8
9
  # Result set
10
  edges: [BlogEdge!]!
11
12
  # Aggregate functions
13
  aggregate: BlogAggregate
14
15
}

1
type BlogEdge {
2
3
  # Unique identifier of the record within the query
4
  cursor: String!
5
6
  # Contents of a record/row in the results set
7
  node: Blog
8
9
}

1
type PageInfo {
2
3
  # unique identifier of the first record within the query
4
  startCursor: String
5
6
  # unique identifier of the last record within the query
7
  endCursor: String
8
9
  # is another page of content available
10
  hasNextPage: Boolean!
11
12
  # is another page of content available
13
  hasPreviousPage: Boolean!
14
}

1
# A record from the `blog` table
2
type Blog {
3
4
  # globally unique identifier
5
  nodeId: ID!
6
7
  # Value from `id` column
8
  id: Int!
9
10
  # Value from `name` column
11
  name: String!
12
13
  # Value from `description` column
14
  description: String
15
16
  # Value from `createdAt` column
17
  createdAt: Datetime!
18
19
  # Value from `updatedAt` column
20
  updatedAt: Datetime!
21
22
}

1
input BlogOrderBy {
2
  id: OrderByDirection
3
  name: OrderByDirection
4
  description: OrderByDirection
5
  createdAt: OrderByDirection
6
  updatedAt: OrderByDirection
7
}

1
input BlogFilter {
2
  nodeId: IDFilter
3
  id: IntFilter
4
  name: StringFilter
5
  description: StringFilter
6
  createdAt: DatetimeFilter
7
  updatedAt: DatetimeFilter
8
  and: [BlogFilter!]
9
  or: [BlogFilter!]
10
  not: BlogFilter
11
}

The totalCount field is disabled by default because it can be expensive on large tables. To enable it use a comment directive

Aggregates

Aggregate functions are available on the collection's aggregate field when enabled via comment directive. These allow you to perform calculations on the collection of records that match your filter criteria.

The supported aggregate operations are:

count: Always available, returns the number of records matching the query
sum: Available for numeric fields, returns the sum of values
avg: Available for numeric fields, returns the average (mean) of values
min: Available for numeric, string, boolean, and date/time fields, returns the minimum value
max: Available for numeric, string, boolean, and date/time fields, returns the maximum value

Example

1
{
2
  blogCollection(
3
    filter: { rating: { gt: 3 } }
4
  ) {
5
    aggregate {
6
      count
7
      sum {
8
        rating
9
        visits
10
      }
11
      avg {
12
        rating
13
      }
14
      min {
15
        createdAt
16
        title
17
      }
18
      max {
19
        rating
20
        updatedAt
21
      }
22
    }
23
  }
24
}

1
{
2
  "data": {
3
    "blogCollection": {
4
      "aggregate": {
5
        "count": 5,
6
        "sum": {
7
          "rating": 23,
8
          "visits": 1250
9
        },
10
        "avg": {
11
          "rating": 4.6
12
        },
13
        "min": {
14
          "createdAt": "2022-01-15T08:30:00Z",
15
          "title": "A Blog Post"
16
        },
17
        "max": {
18
          "rating": 5,
19
          "updatedAt": "2023-04-22T14:15:30Z"
20
        }
21
      }
22
    }
23
  }
24
}

1
"""Aggregate results for `Blog`"""
2
type BlogAggregate {
3
  """The number of records matching the query"""
4
  count: Int!
5
6
  """Summation aggregates for `Blog`"""
7
  sum: BlogSumAggregateResult
8
9
  """Average aggregates for `Blog`"""
10
  avg: BlogAvgAggregateResult
11
12
  """Minimum aggregates for comparable fields"""
13
  min: BlogMinAggregateResult
14
15
  """Maximum aggregates for comparable fields"""
16
  max: BlogMaxAggregateResult
17
}

GraphQL Types

1
"""Result of summation aggregation for `Blog`"""
2
type BlogSumAggregateResult {
3
  """Sum of rating values"""
4
  rating: BigFloat
5
6
  """Sum of visits values"""
7
  visits: BigInt
8
9
  # Other numeric fields...
10
}

1
"""Result of average aggregation for `Blog`"""
2
type BlogAvgAggregateResult {
3
  """Average of rating values"""
4
  rating: BigFloat
5
6
  """Average of visits values"""
7
  visits: BigFloat
8
9
  # Other numeric fields...
10
}

1
"""Result of minimum aggregation for `Blog`"""
2
type BlogMinAggregateResult {
3
  """Minimum rating value"""
4
  rating: Float
5
6
  """Minimum title value"""
7
  title: String
8
9
  """Minimum createdAt value"""
10
  createdAt: Datetime
11
12
  # Other comparable fields...
13
}

1
"""Result of maximum aggregation for `Blog`"""
2
type BlogMaxAggregateResult {
3
  """Maximum rating value"""
4
  rating: Float
5
6
  """Maximum title value"""
7
  title: String
8
9
  """Maximum updatedAt value"""
10
  updatedAt: Datetime
11
12
  # Other comparable fields...
13
}

The return type for sum depends on the input type: integer fields return BigInt, while other numeric fields return BigFloat.
The return type for avg is always BigFloat.
The return types for min and max match the original field types.

The aggregate field is disabled by default because it can be expensive on large tables. To enable it use a comment directive

Pagination

Keyset Pagination

Paginating forwards and backwards through collections is handled using the first, last, before, and after parameters, following the relay spec.

1
type Query {
2
3
  blogCollection(
4
5
    """Query the first `n` records in the collection"""
6
    first: Int
7
8
    """Query the last `n` records in the collection"""
9
    last: Int
10
11
    """Query values in the collection before the provided cursor"""
12
    before: Cursor
13
14
    """Query values in the collection after the provided cursor"""
15
    after: Cursor
16
17
    ...truncated...
18
19
  ): BlogConnection!
20
}

Metadata relating to the current page of a result set is available on the pageInfo field of the connection type returned from a collection.

1
type PageInfo {
2
3
  # unique identifier of the first record within the query
4
  startCursor: String
5
6
  # unique identifier of the last record within the query
7
  endCursor: String
8
9
  # is another page of content available
10
  hasNextPage: Boolean!
11
12
  # is another page of content available
13
  hasPreviousPage: Boolean!
14
}

1
type BlogConnection {
2
3
  # Pagination metadata
4
  pageInfo: PageInfo!
5
6
  # Result set
7
  edges: [BlogEdge!]!
8
9
}

To paginate forward in the collection, use the first and after arguments. To retrieve the first page, the after argument should be null or absent.

Example

1
{
2
  blogCollection(
3
    first: 2,
4
    after: null
5
  ) {
6
    pageInfo {
7
      startCursor
8
      endCursor
9
      hasPreviousPage
10
      hasNextPage
11
    }
12
    edges {
13
      cursor
14
      node {
15
        id
16
      }
17
    }
18
  }
19
}

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 1
8
          },
9
          "cursor": "WzFd"
10
        },
11
        {
12
          "node": {
13
            "id": 2
14
          },
15
          "cursor": "WzJd"
16
        }
17
      ],
18
      "pageInfo": {
19
        "startCursor": "WzFd",
20
        "endCursor": "WzJd",
21
        "hasNextPage": true,
22
        "hasPreviousPage": false
23
      }
24
    }
25
  }
26
}

To retrieve the next page, provide the cursor value from data.blogCollection.pageInfo.endCursor to the after argument of another query.

1
{
2
  blogCollection(
3
    first: 2,
4
    after: "WzJd"
5
  ) {
6
  ...truncated...
7
}

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 3
8
          },
9
          "cursor": "WzNd"
10
        },
11
        {
12
          "node": {
13
            "id": 4
14
          },
15
          "cursor": "WzRd"
16
        }
17
      ],
18
      "pageInfo": {
19
        "startCursor": "WzNd",
20
        "endCursor": "WzRd",
21
        "hasNextPage": false,
22
        "hasPreviousPage": true
23
      }
24
    }
25
  }
26
}

once the collection has been fully enumerated, data.blogConnection.pageInfo.hasNextPage returns false.

To paginate backwards through a collection, repeat the process substituting first -> last, after -> before, hasNextPage -> hasPreviousPage

Offset Pagination

In addition to keyset pagination, collections may also be paged using first and offset, which operates like SQL's limit and offset to skip offset number of records in the results.

offset based pagination becomes inefficient the offset value increases. For this reason, prefer cursor based pagination where possible.

1
{
2
  blogCollection(
3
    first: 2,
4
    offset: 2
5
  ) {
6
  ...truncated...
7
}

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 3
8
          },
9
          "cursor": "WzNd"
10
        },
11
        {
12
          "node": {
13
            "id": 4
14
          },
15
          "cursor": "WzRd"
16
        }
17
      ],
18
      "pageInfo": {
19
        "startCursor": "WzNd",
20
        "endCursor": "WzRd",
21
        "hasNextPage": false,
22
        "hasPreviousPage": true
23
      }
24
    }
25
  }
26
}

Filtering

To filter the result set, use the filter argument.

1
type Query {
2
3
  blogCollection(
4
5
    """Filters to apply to the results set when querying from the collection"""
6
    filter: BlogFilter
7
8
    ...truncated...
9
10
  ): BlogConnection!
11
}

Where the <Table>Filter type enumerates filterable fields and their associated <Type>Filter.

1
input BlogFilter {
2
  nodeId: IDFilter
3
  id: IntFilter
4
  name: StringFilter
5
  description: StringFilter
6
  tags: StringListFilter
7
  createdAt: DatetimeFilter
8
  updatedAt: DatetimeFilter
9
  and: [BlogFilter!]
10
  or: [BlogFilter!]
11
  not: BlogFilter
12
}

1
"""
2
Boolean expression comparing fields on type "Int"
3
"""
4
input IntFilter {
5
  eq: Int
6
  gt: Int
7
  gte: Int
8
  in: [Int!]
9
  lt: Int
10
  lte: Int
11
  neq: Int
12
  is: FilterIs
13
}

1
"""
2
Boolean expression comparing fields on type "String"
3
"""
4
input StringFilter {
5
  eq: String
6
  gt: String
7
  gte: String
8
  in: [String!]
9
  lt: String
10
  lte: String
11
  neq: String
12
  is: FilterIs
13
  startsWith: String
14
  like: String
15
  ilike: String
16
  regex: String
17
  iregex: String
18
}

1
"""
2
Boolean expression comparing fields on type "StringList"
3
"""
4
input StringListFilter {
5
  contains: [String!]
6
  containedBy: [String!]
7
  eq: [String!]
8
  overlaps: [String!]
9
  is: FilterIs
10
}

1
enum FilterIs {
2
  NULL
3
  NOT_NULL
4
}

The following list shows the operators that may be available on <Type>Filter types.

Operator	Description
eq	Equal To
neq	Not Equal To
gt	Greater Than
gte	Greater Than Or Equal To
in	Contained by Value List
lt	Less Than
lte	Less Than Or Equal To
is	Null or Not Null
startsWith	Starts with prefix
like	Pattern Match. '%' as wildcard
ilike	Pattern Match. '%' as wildcard. Case Insensitive
regex	POSIX Regular Expression Match
iregex	POSIX Regular Expression Match. Case Insensitive
contains	Contains. Applies to array columns only.
containedBy	Contained in. Applies to array columns only.
overlaps	Overlap (have points in common). Applies to array columns only.

Not all operators are available on every <Type>Filter type. For example, UUIDFilter only supports eq and neq because UUIDs are not ordered.

Example: simple

1
{
2
  blogCollection(
3
    filter: {id: {lt: 3}},
4
  ) {
5
    edges {
6
      cursor
7
      node {
8
        id
9
      }
10
    }
11
  }
12
}

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 1
8
          },
9
          "cursor": "WzFd"
10
        },
11
        {
12
          "node": {
13
            "id": 2
14
          },
15
          "cursor": "WzJd"
16
        }
17
      ]
18
    }
19
  }
20
}

Example: array column

The contains filter is used to return results where all the elements in the input array appear in the array column.

contains

Filter Query"

1
{
2
  blogCollection(
3
    filter: {tags: {contains: ["tech", "innovation"]}},
4
  ) {
5
    edges {
6
      cursor
7
      node {
8
        id
9
        name
10
        tags
11
        createdAt
12
      }
13
    }
14
  }
15
}

contains

Filter Result"

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 1,
8
            "name": "A: Blog 1",
9
            "createdAt": "2023-07-24T04:01:09.882781",
10
            "tags": ["tech", "innovation"]
11
          },
12
          "cursor": "WzFd"
13
        },
14
        {
15
          "node": {
16
            "id": 2,
17
            "name": "A: Blog 2",
18
            "createdAt": "2023-07-24T04:01:09.882781",
19
            "tags": ["tech", "innovation", "entrepreneurship"]
20
          },
21
          "cursor": "WzJd"
22
        }
23
      ]
24
    }
25
  }
26
}

The contains filter can also accept a single scalar.

contains

Filter with Scalar Query"

1
{
2
  blogCollection(
3
    filter: {tags: {contains: "tech"}},
4
  ) {
5
    edges {
6
      cursor
7
      node {
8
        id
9
        name
10
        tags
11
        createdAt
12
      }
13
    }
14
  }
15
}

contains

Filter with Scalar Result"

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 1,
8
            "name": "A: Blog 1",
9
            "createdAt": "2023-07-24T04:01:09.882781",
10
            "tags": ["tech", "innovation"]
11
          },
12
          "cursor": "WzFd"
13
        },
14
        {
15
          "node": {
16
            "id": 2,
17
            "name": "A: Blog 2",
18
            "createdAt": "2023-07-24T04:01:09.882781",
19
            "tags": ["tech", "innovation", "entrepreneurship"]
20
          },
21
          "cursor": "WzJd"
22
        }
23
      ]
24
    }
25
  }
26
}

The containedBy filter is used to return results where every element of the array column appears in the input array.

containedBy

Filter Query"

1
{
2
  blogCollection(
3
    filter: {tags: {containedBy: ["entrepreneurship", "innovation", "tech"]}},
4
  ) {
5
    edges {
6
      cursor
7
      node {
8
        id
9
        name
10
        tags
11
        createdAt
12
      }
13
    }
14
  }
15
}

containedBy

Filter Result"

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 1,
8
            "name": "A: Blog 1",
9
            "createdAt": "2023-07-24T04:01:09.882781",
10
            "tags": ["tech", "innovation"]
11
          },
12
          "cursor": "WzFd"
13
        },
14
        {
15
          "node": {
16
            "id": 3,
17
            "name": "A: Blog 3",
18
            "createdAt": "2023-07-24T04:01:09.882781",
19
            "tags": ["innovation", "entrepreneurship"]
20
          },
21
          "cursor": "WzNd"
22
        }
23
      ]
24
    }
25
  }
26
}

The containedBy filter can also accept a single scalar. In this case, only results where the only element in the array column is the input scalar are returned.

containedBy

Filter with Scalar Query"

1
{
2
  blogCollection(
3
    filter: {tags: {containedBy: "travel"}},
4
  ) {
5
    edges {
6
      cursor
7
      node {
8
        id
9
        name
10
        tags
11
        createdAt
12
      }
13
    }
14
  }
15
}

containedBy

Filter with Scalar Result"

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 4,
8
            "name": "A: Blog 4",
9
            "createdAt": "2023-07-24T04:01:09.882781",
10
            "tags": ["travel"]
11
          },
12
          "cursor": "WzPd"
13
        }
14
      ]
15
    }
16
  }
17
}

The overlaps filter is used to return results where the array column and the input array have at least one element in common.

overlaps

Filter Query"

1
{
2
  blogCollection(
3
    filter: {tags: {overlaps: ["tech", "travel"]}},
4
  ) {
5
    edges {
6
      cursor
7
      node {
8
        id
9
        name
10
        tags
11
        createdAt
12
      }
13
    }
14
  }
15
}

overlaps

Filter Result"

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 1,
8
            "name": "A: Blog 1",
9
            "createdAt": "2023-07-24T04:01:09.882781",
10
            "tags": ["tech", "innovation"]
11
          },
12
          "cursor": "WzFd"
13
        },
14
        {
15
          "node": {
16
            "id": 2,
17
            "name": "A: Blog 2",
18
            "createdAt": "2023-07-24T04:01:09.882781",
19
            "tags": ["tech", "innovation", "entrepreneurship"]
20
          },
21
          "cursor": "WzJd"
22
        },
23
        {
24
          "node": {
25
            "id": 4,
26
            "name": "A: Blog 4",
27
            "createdAt": "2023-07-24T04:01:09.882781",
28
            "tags": ["travel"]
29
          },
30
          "cursor": "WzPd"
31
        }
32
      ]
33
    }
34
  }
35
}

Example: and/or

Multiple filters can be combined with and, or and not operators. The and and or operators accept a list of <Type>Filter.

and

Filter Query"

1
{
2
  blogCollection(
3
    filter: {
4
      and: [
5
        {id: {eq: 1}}
6
        {name: {eq: "A: Blog 1"}}
7
      ]
8
    }
9
  ) {
10
    edges {
11
      cursor
12
      node {
13
        id
14
        name
15
        description
16
        createdAt
17
      }
18
    }
19
  }
20
}

and

Filter Result"

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 1,
8
            "name": "A: Blog 1",
9
            "createdAt": "2023-07-24T04:01:09.882781",
10
            "description": "a desc1"
11
          },
12
          "cursor": "WzFd"
13
        }
14
      ]
15
    }
16
  }
17
}

or

Filter Query"

1
{
2
  blogCollection(
3
    filter: {
4
      or: [
5
        {id: {eq: 1}}
6
        {name: {eq: "A: Blog 2"}}
7
      ]
8
    }
9
  ) {
10
    edges {
11
      cursor
12
      node {
13
        id
14
        name
15
        description
16
        createdAt
17
      }
18
    }
19
  }
20
}

or

Filter Result"

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 1,
8
            "name": "A: Blog 1",
9
            "createdAt": "2023-07-24T04:01:09.882781",
10
            "description": "a desc1"
11
          },
12
          "cursor": "WzFd"
13
        },
14
        {
15
          "node": {
16
            "id": 2,
17
            "name": "A: Blog 2",
18
            "createdAt": "2023-07-24T04:01:09.882781",
19
            "description": "a desc2"
20
          },
21
          "cursor": "WzJd"
22
        }
23
      ]
24
    }
25
  }
26
}

Example: not

not accepts a single <Type>Filter.

not

Filter Query"

1
{
2
  blogCollection(
3
    filter: {
4
      not: {id: {eq: 1}}
5
    }
6
  ) {
7
    edges {
8
      cursor
9
      node {
10
        id
11
        name
12
        description
13
        createdAt
14
      }
15
    }
16
  }
17
}

not

Filter Result"

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 2,
8
            "name": "A: Blog 2",
9
            "createdAt": "2023-07-24T04:01:09.882781",
10
            "description": "a desc2"
11
          },
12
          "cursor": "WzJd"
13
        },
14
        {
15
          "node": {
16
            "id": 3,
17
            "name": "A: Blog 3",
18
            "createdAt": "2023-07-24T04:01:09.882781",
19
            "description": "a desc3"
20
          },
21
          "cursor": "WzNd"
22
        },
23
        {
24
          "node": {
25
            "id": 4,
26
            "name": "B: Blog 3",
27
            "createdAt": "2023-07-24T04:01:09.882781",
28
            "description": "b desc1"
29
          },
30
          "cursor": "WzRd"
31
        }
32
      ]
33
    }
34
  }
35
}

Example: nested composition

The and, or and not operators can be arbitrarily nested inside each other.

1
{
2
  blogCollection(
3
    filter: {
4
      or: [
5
        { id: { eq: 1 } }
6
        { id: { eq: 2 } }
7
        { and: [{ id: { eq: 3 }, not: { name: { eq: "A: Blog 2" } } }] }
8
      ]
9
    }
10
  ) {
11
    edges {
12
      cursor
13
      node {
14
        id
15
        name
16
        description
17
        createdAt
18
      }
19
    }
20
  }
21
}

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 1,
8
            "name": "A: Blog 1",
9
            "createdAt": "2023-07-24T04:01:09.882781",
10
            "description": "a desc1"
11
          },
12
          "cursor": "WzFd"
13
        },
14
        {
15
          "node": {
16
            "id": 2,
17
            "name": "A: Blog 2",
18
            "createdAt": "2023-07-24T04:01:09.882781",
19
            "description": "a desc2"
20
          },
21
          "cursor": "WzJd"
22
        },
23
        {
24
          "node": {
25
            "id": 3,
26
            "name": "A: Blog 3",
27
            "createdAt": "2023-07-24T04:01:09.882781",
28
            "description": "a desc3"
29
          },
30
          "cursor": "WzNd"
31
        }
32
      ]
33
    }
34
  }
35
}

Example: empty

Empty filters are ignored, i.e. they behave as if the operator was not specified at all.

1
{
2
  blogCollection(
3
    filter: {
4
      and: [], or: [], not: {}
5
    }
6
  ) {
7
    edges {
8
      cursor
9
      node {
10
        id
11
        name
12
        description
13
        createdAt
14
      }
15
    }
16
  }
17
}

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 1,
8
            "name": "A: Blog 1",
9
            "createdAt": "2023-07-24T04:01:09.882781",
10
            "description": "a desc1"
11
          },
12
          "cursor": "WzFd"
13
        },
14
        {
15
          "node": {
16
            "id": 2,
17
            "name": "A: Blog 2",
18
            "createdAt": "2023-07-24T04:01:09.882781",
19
            "description": "a desc2"
20
          },
21
          "cursor": "WzJd"
22
        },
23
        {
24
          "node": {
25
            "id": 3,
26
            "name": "A: Blog 3",
27
            "createdAt": "2023-07-24T04:01:09.882781",
28
            "description": "a desc3"
29
          },
30
          "cursor": "WzNd"
31
        },
32
        {
33
          "node": {
34
            "id": 4,
35
            "name": "B: Blog 3",
36
            "createdAt": "2023-07-24T04:01:09.882781",
37
            "description": "b desc1"
38
          },
39
          "cursor": "WzRd"
40
        }
41
      ]
42
    }
43
  }
44
}

Example: implicit and

Multiple column filters at the same level will be implicitly combined with boolean and. In the following example the id: {eq: 1} and name: {eq: "A: Blog 1"} will be anded.

1
{
2
  blogCollection(
3
    filter: {
4
      # Equivalent to not: { and: [{id: {eq: 1}}, {name: {eq: "A: Blog 1"}}]}
5
      not: {
6
        id: {eq: 1}
7
        name: {eq: "A: Blog 1"}
8
      }
9
    }
10
  ) {
11
    edges {
12
      cursor
13
      node {
14
        id
15
        name
16
        description
17
        createdAt
18
      }
19
    }
20
  }
21
}

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 2,
8
            "name": "A: Blog 2",
9
            "createdAt": "2023-07-24T04:01:09.882781",
10
            "description": "a desc2"
11
          },
12
          "cursor": "WzJd"
13
        },
14
        {
15
          "node": {
16
            "id": 3,
17
            "name": "A: Blog 3",
18
            "createdAt": "2023-07-24T04:01:09.882781",
19
            "description": "a desc3"
20
          },
21
          "cursor": "WzNd"
22
        },
23
        {
24
          "node": {
25
            "id": 4,
26
            "name": "B: Blog 3",
27
            "createdAt": "2023-07-24T04:01:09.882781",
28
            "description": "b desc1"
29
          },
30
          "cursor": "WzRd"
31
        }
32
      ]
33
    }
34
  }
35
}

This means that an and filter can be often be simplified. In the following example all queries are equivalent and produce the same result.

and

Query"

1
{
2
  blogCollection(
3
    filter: {
4
      and: [
5
        {id: {gt: 0}}
6
        {id: {lt: 2}}
7
        {name: {eq: "A: Blog 1"}}
8
      ]
9
    }
10
  ) {
11
    edges {
12
      cursor
13
      node {
14
        id
15
        name
16
        description
17
        createdAt
18
      }
19
    }
20
  }
21
}

and

Query"

1
{
2
  blogCollection(
3
    filter: {
4
        id: {gt: 0}
5
        id: {lt: 2}
6
        name: {eq: "A: Blog 1"}
7
    }
8
  ) {
9
    edges {
10
      cursor
11
      node {
12
        id
13
        name
14
        description
15
        createdAt
16
      }
17
    }
18
  }
19
}

1
{
2
  blogCollection(
3
    filter: {
4
        id: {gt: 0, lt: 2}
5
        name: {eq: "A: Blog 1"}
6
    }
7
  ) {
8
    edges {
9
      cursor
10
      node {
11
        id
12
        name
13
        description
14
        createdAt
15
      }
16
    }
17
  }
18
}

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 2,
8
            "name": "A: Blog 2",
9
            "createdAt": "2023-07-24T04:01:09.882781",
10
            "description": "a desc2"
11
          },
12
          "cursor": "WzJd"
13
        },
14
        {
15
          "node": {
16
            "id": 3,
17
            "name": "A: Blog 3",
18
            "createdAt": "2023-07-24T04:01:09.882781",
19
            "description": "a desc3"
20
          },
21
          "cursor": "WzNd"
22
        },
23
        {
24
          "node": {
25
            "id": 4,
26
            "name": "B: Blog 3",
27
            "createdAt": "2023-07-24T04:01:09.882781",
28
            "description": "b desc1"
29
          },
30
          "cursor": "WzRd"
31
        }
32
      ]
33
    }
34
  }
35
}

Be aware that the above simplification only works for the and operator. If you try it with an or operator it will behave like an and.

1
{
2
  blogCollection(
3
    filter: {
4
      # This is really an `and` in `or`'s clothing
5
      or: {
6
        id: {eq: 1}
7
        name: {eq: "A: Blog 2"}
8
      }
9
    }
10
  ) {
11
    edges {
12
      cursor
13
      node {
14
        id
15
        name
16
        description
17
        createdAt
18
      }
19
    }
20
  }
21
}

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": []
5
    }
6
  }
7
}

This is because according to the rules of GraphQL list input coercion, if a value passed to an input of list type is not a list, then it is coerced to a list of a single item. So in the above example or: {id: {eq: 1}, name: {eq: "A: Blog 2}} will be coerced into or: [{id: {eq: 1}, name: {eq: "A: Blog 2}}] which is equivalent to or: [and: [{id: {eq: 1}}, {name: {eq: "A: Blog 2}}}] due to implicit anding.

Avoid naming your columns and, or or not. If you do, the corresponding filter operator will not be available for use.

The and, or and not operators also work with update and delete mutations.

Ordering

The default order of results is defined by the underlying table's primary key column in ascending order. That default can be overridden by passing an array of <Table>OrderBy to the collection's orderBy argument.

1
type Query {
2
3
  blogCollection(
4
5
    """Sort order to apply to the collection"""
6
    orderBy: [BlogOrderBy!]
7
8
    ...truncated...
9
10
  ): BlogConnection!
11
}

1
input BlogOrderBy {
2
  id: OrderByDirection
3
  name: OrderByDirection
4
  description: OrderByDirection
5
  createdAt: OrderByDirection
6
  updatedAt: OrderByDirection
7
}

1
"""Defines a per-field sorting order"""
2
enum OrderByDirection {
3
  """Ascending order, nulls first"""
4
  AscNullsFirst
5
6
  """Ascending order, nulls last"""
7
  AscNullsLast
8
9
  """Descending order, nulls first"""
10
  DescNullsFirst
11
12
  """Descending order, nulls last"""
13
  DescNullsLast
14
}

Example

1
{
2
  blogCollection(
3
    orderBy: [{id: DescNullsLast}]
4
  ) {
5
    edges {
6
      node {
7
        id
8
      }
9
    }
10
  }
11
}

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 4
8
          }
9
        },
10
        {
11
          "node": {
12
            "id": 3
13
          }
14
        },
15
        {
16
          "node": {
17
            "id": 2
18
          }
19
        },
20
        {
21
          "node": {
22
            "id": 1
23
          }
24
        }
25
      ]
26
    }
27
  }
28
}

Note, only one key value pair may be provided to each element of the input array. For example, [{name: AscNullsLast}, {id: AscNullFirst}] is valid. Passing multiple key value pairs in a single element of the input array e.g. [{name: AscNullsLast, id: AscNullFirst}], is invalid.

Primary Key Queries

Each table has a top level field in the Query type for selecting a single record by primary key from that table. The field is named <table>ByPk

SQL Setup

1
create table "Blog"(
2
  id serial primary key,
3
  name varchar(255) not null,
4
  description varchar(255),
5
  "createdAt" timestamp not null,
6
  "updatedAt" timestamp not null
7
);

GraphQL Types

1
"""The root type for querying data"""
2
type Query {
3
4
  """Retrieve a blog by its id"""
5
  blogByPk(id: Int!): Blog
6
7
}

To query the table by primary key, pass the value of the primary key field to the field:

Example

1
{
2
  blogByPk(
3
    id: 1
4
  ) {
5
    id
6
    name
7
    description
8
  }
9
}

1
{
2
  "data": {
3
    "blogByPk": {
4
      "id": 1,
5
      "name": "Some Blog",
6
      "description": "Description of Some Blog"
7
    }
8
  }
9
}

If a record with the give id doesn't exist, the field will return null:

Example

1
{
2
  blogByPk(
3
    id: 999
4
  ) {
5
    id
6
    name
7
    description
8
  }
9
}

1
{
2
  "data": {
3
    "blogByPk": null
4
  }
5
}

If the key is a composite primary key, all the columns of the primary key should be sent in the query:

SQL Setup

1
create table item(
2
    item_id int,
3
    product_id int,
4
    quantity int,
5
    price numeric(10,2),
6
    primary key(item_id, product_id)
7
);

GraphQL Types

1
"""The root type for querying data"""
2
type Query {
3
4
  """Retrieve an item by its item and product ids"""
5
  itemByPk(itemId: Int!, productId: Int!): Item
6
7
}

1
{
2
  itemByPk(
3
    itemId: 1, productId: 2
4
  ) {
5
    itemId
6
    productId
7
    quantity
8
    price
9
  }
10
}

Example

1
{
2
  "data": {
3
    "itemByPk": {
4
      "itemId": 1,
5
      "productId": 2,
6
      "quantity": 1,
7
      "price": "24.99"
8
    }
9
  }
10
}

Otherwise an error will be returned:

Example

1
{
2
  itemByPk(
3
    itemId: 1
4
  ) {
5
    itemId
6
    productId
7
    quantity
8
    price
9
  }
10
}

1
{
2
  "data": null,
3
  "errors": [
4
    {
5
      "message": "Missing primary key column(s): product_id"
6
    }
7
  ]
8
}

MutationType

The Mutation type is the entrypoint for mutations/edits.

Each table has top level entry in the Mutation type for inserting insertInto<Table>Collection, updating update<Table>Collection and deleting deleteFrom<Table>Collection.

SQL Setup

1
create table "Blog"(
2
  id serial primary key,
3
  name varchar(255) not null,
4
  description varchar(255),
5
  "createdAt" timestamp not null default now(),
6
  "updatedAt" timestamp
7
);

1
"""The root type for creating and mutating data"""
2
type Mutation {
3
4
  """Adds one or more `BlogInsertResponse` records to the collection"""
5
  insertIntoBlogCollection(
6
7
    """Records to add to the Blog collection"""
8
    objects: [BlogInsertInput!]!
9
10
  ): BlogInsertResponse
11
12
  """Updates zero or more records in the collection"""
13
  updateBlogCollection(
14
    """
15
    Fields that are set will be updated for all records matching the `filter`
16
    """
17
    set: BlogUpdateInput!
18
19
    """Restricts the mutation's impact to records matching the critera"""
20
    filter: BlogFilter
21
22
    """
23
    The maximum number of records in the collection permitted to be affected
24
    """
25
    atMost: Int! = 1
26
27
  ): BlogUpdateResponse!
28
29
  """Deletes zero or more records from the collection"""
30
  deleteFromBlogCollection(
31
    """Restricts the mutation's impact to records matching the critera"""
32
    filter: BlogFilter
33
34
    """
35
    The maximum number of records in the collection permitted to be affected
36
    """
37
    atMost: Int! = 1
38
39
  ): BlogDeleteResponse!
40
41
}

Insert

To add records to a collection, use the insertInto<Table>Collection field on the Mutation type.

SQL Setup

1
create table "Blog"(
2
  id serial primary key,
3
  name varchar(255) not null,
4
  description varchar(255),
5
  "createdAt" timestamp not null default now(),
6
  "updatedAt" timestamp
7
);

GraphQL Types

1
"""The root type for creating and mutating data"""
2
type Mutation {
3
4
  """Adds one or more `BlogInsertResponse` records to the collection"""
5
  insertIntoBlogCollection(
6
7
    """Records to add to the Blog collection"""
8
    objects: [BlogInsertInput!]!
9
10
  ): BlogInsertResponse
11
12
}

1
input BlogInsertInput {
2
  name: String
3
  description: String
4
  createdAt: Datetime
5
  updatedAt: Datetime
6
}

1
type BlogInsertResponse {
2
  """Count of the records impacted by the mutation"""
3
  affectedCount: Int!
4
5
  """Array of records impacted by the mutation"""
6
  records: [Blog!]!
7
}

Where elements in the objects array are inserted into the underlying table.

Example

1
mutation {
2
  insertIntoBlogCollection(
3
    objects: [
4
      {name: "foo"},
5
      {name: "bar"},
6
    ]
7
  ) {
8
    affectedCount
9
    records {
10
      id
11
      name
12
    }
13
  }
14
}

1
{
2
  "data": {
3
    "insertIntoBlogCollection": {
4
      "records": [
5
        {
6
          "id": 1,
7
          "name": "foo"
8
        },
9
        {
10
          "id": 2,
11
          "name": "bar"
12
        }
13
      ],
14
      "affectedCount": 2
15
    }
16
  }
17
}

Update

To update records in a collection, use the update<Table>Collection field on the Mutation type.

SQL Setup

1
create table "Blog"(
2
  id serial primary key,
3
  name varchar(255) not null,
4
  description varchar(255),
5
  "createdAt" timestamp not null default now(),
6
  "updatedAt" timestamp
7
);

GraphQL Types

1
"""The root type for creating and mutating data"""
2
type Mutation {
3
4
  """Updates zero or more records in the collection"""
5
  updateBlogCollection(
6
    """
7
    Fields that are set will be updated for all records matching the `filter`
8
    """
9
    set: BlogUpdateInput!
10
11
    """Restricts the mutation's impact to records matching the critera"""
12
    filter: BlogFilter
13
14
    """
15
    The maximum number of records in the collection permitted to be affected
16
    """
17
    atMost: Int! = 1
18
19
  ): BlogUpdateResponse!
20
21
}

1
input BlogUpdateInput {
2
  name: String
3
  description: String
4
  createdAt: Datetime
5
  updatedAt: Datetime
6
}

1
type BlogUpdateResponse {
2
3
  """Count of the records impacted by the mutation"""
4
  affectedCount: Int!
5
6
  """Array of records impacted by the mutation"""
7
  records: [Blog!]!
8
9
}

Where the set argument is a key value pair describing the values to update, filter controls which records should be updated, and atMost restricts the maximum number of records that may be impacted. If the number of records impacted by the mutation exceeds the atMost parameter the operation will return an error.

Example

1
mutation {
2
  updateBlogCollection(
3
    set: {name: "baz"}
4
    filter: {id: {eq: 1}}
5
  ) {
6
    affectedCount
7
    records {
8
      id
9
      name
10
    }
11
  }
12
}

1
{
2
  "data": {
3
    "updateBlogCollection": {
4
      "records": [
5
        {
6
          "id": 1,
7
          "name": "baz"
8
        }
9
      ],
10
      "affectedCount": 1
11
    }
12
  }
13
}

Delete

To remove records from a collection, use the deleteFrom<Table>Collection field on the Mutation type.

SQL Setup

1
create table "Blog"(
2
  id serial primary key,
3
  name varchar(255) not null,
4
  description varchar(255),
5
  "createdAt" timestamp not null default now(),
6
  "updatedAt" timestamp
7
);

GraphQL Types

1
"""The root type for creating and mutating data"""
2
type Mutation {
3
4
  """Deletes zero or more records from the collection"""
5
  deleteFromBlogCollection(
6
    """Restricts the mutation's impact to records matching the critera"""
7
    filter: BlogFilter
8
9
    """
10
    The maximum number of records in the collection permitted to be affected
11
    """
12
    atMost: Int! = 1
13
14
  ): BlogDeleteResponse!
15
16
}

1
input BlogFilter {
2
  id: IntFilter
3
  name: StringFilter
4
  description: StringFilter
5
  createdAt: DatetimeFilter
6
  updatedAt: DatetimeFilter
7
  and: [BlogFilter!]
8
  or: [BlogFilter!]
9
  not: BlogFilter
10
}

1
type BlogDeleteResponse {
2
  """Count of the records impacted by the mutation"""
3
  affectedCount: Int!
4
5
  """Array of records impacted by the mutation"""
6
  records: [Blog!]!
7
}

Where filter controls which records should be deleted and atMost restricts the maximum number of records that may be deleted. If the number of records impacted by the mutation exceeds the atMost parameter the operation will return an error.

Example

1
mutation {
2
  deleteFromBlogCollection(
3
    filter: {id: {eq: 1}}
4
  ) {
5
    affectedCount
6
    records {
7
      id
8
      name
9
    }
10
  }
11
}

1
{
2
  "data": {
3
    "deleteFromBlogCollection": {
4
      "records": [
5
        {
6
          "id": 1,
7
          "name": "baz"
8
        }
9
      ],
10
      "affectedCount": 1
11
    }
12
  }
13
}

Concepts

nodeId

The base GraphQL type for every table with a primary key is automatically assigned a nodeId: ID! field. That value, can be passed to the node entrypoint of the Query type to retrieve its other fields. nodeId may also be used as a caching key.

relay support

By default relay expects the ID field for types to have the name id. pg_graphql uses nodeId by default to avoid conflicting with user defined id columns. You can configure relay to work with pg_graphql's nodeId field with relay's nodeInterfaceIdField option. More info available here.

SQL Setup

1
create table "Blog"(
2
    id serial primary key,
3
    name varchar(255) not null
4
);

GraphQL Types

1
type Blog {
2
  nodeId: ID! # this field
3
  id: Int!
4
  name: String!
5
}

Relationships

Relationships between collections in the Graph are derived from foreign keys.

One-to-Many

A foreign key on table A referencing table B defines a one-to-many relationship from table A to table B.

SQL Setup

1
create table "Blog"(
2
    id serial primary key,
3
    name varchar(255) not null
4
);
5
6
create table "BlogPost"(
7
    id serial primary key,
8
    "blogId" integer not null references "Blog"(id),
9
    title varchar(255) not null,
10
    body varchar(10000)
11
);

GraphQL Types

1
type Blog {
2
3
  # globally unique identifier
4
  nodeId: ID!
5
6
  id: Int!
7
  name: String!
8
  description: String
9
10
  blogPostCollection(
11
    """Query the first `n` records in the collection"""
12
    first: Int
13
14
    """Query the last `n` records in the collection"""
15
    last: Int
16
17
    """Query values in the collection before the provided cursor"""
18
    before: Cursor
19
20
    """Query values in the collection after the provided cursor"""
21
    after: Cursor
22
23
    """
24
    Skip n values from the after cursor. Alternative to cursor pagination. Backward pagination not supported.
25
    """
26
    offset: Int
27
28
    """Filters to apply to the results set when querying from the collection"""
29
    filter: BlogPostFilter
30
31
    """Sort order to apply to the collection"""
32
    orderBy: [BlogPostOrderBy!]
33
  ): BlogPostConnection!
34
35
}

Where blogPostCollection exposes the full Query interface to BlogPosts.

Example

1
{
2
  blogCollection {
3
    edges {
4
      node {
5
        name
6
        blogPostCollection {
7
          edges {
8
            node {
9
              id
10
              title
11
            }
12
          }
13
        }
14
      }
15
    }
16
  }
17
}

1
{
2
  "data": {
3
    "blogCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "name": "pg_graphql blog",
8
            "blogPostCollection": {
9
              "edges": [
10
                {
11
                  "node": {
12
                    "id": 2,
13
                    "title": "fIr3t p0sT"
14
                  }
15
                },
16
                {
17
                  "node": {
18
                    "id": 3,
19
                    "title": "graphql with postgres"
20
                  }
21
                }
22
              ]
23
            }
24
          }
25
        }
26
      ]
27
    }
28
  }
29
}

Many-to-One

A foreign key on table A referencing table B defines a many-to-one relationship from table B to table A.

SQL Setup

1
create table "Blog"(
2
    id serial primary key,
3
    name varchar(255) not null
4
);
5
6
create table "BlogPost"(
7
    id serial primary key,
8
    "blogId" integer not null references "Blog"(id),
9
    title varchar(255) not null,
10
    body varchar(10000)
11
);

GraphQL Types

1
type BlogPost {
2
  nodeId: ID!
3
  id: Int!
4
  blogId: Int!
5
  title: String!
6
  body: String
7
8
  blog: Blog
9
}

Where blog exposes the Blog record associated with the BlogPost.

1
{
2
  blogPostCollection {
3
    edges {
4
      node {
5
        title
6
        blog {
7
          name
8
        }
9
      }
10
    }
11
  }
12
}

1
{
2
  "data": {
3
    "blogPostCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "blog": {
8
              "name": "pg_graphql blog"
9
            },
10
            "title": "fIr3t p0sT"
11
          }
12
        },
13
        {
14
          "node": {
15
            "blog": {
16
              "name": "pg_graphql blog"
17
            },
18
            "title": "graphql with postgres"
19
          }
20
        }
21
      ]
22
    }
23
  }
24
}

One-to-One

A one-to-one relationship is defined by a foreign key on table A referencing table B where the columns making up the foreign key on table A are unique.

SQL Setup

1
create table "EmailAddress"(
2
    id serial primary key,
3
    address text unique not null
4
);
5
6
create table "Employee"(
7
    id serial primary key,
8
    name text not null,
9
    email_address_id int unique references "EmailAddress"(id)
10
);

GraphQL Types

1
type Employee {
2
  nodeId: ID!
3
  id: Int!
4
  name: String!
5
  emailAddressId: Int
6
  emailAddress: EmailAddress
7
}

1
type EmailAddress {
2
  nodeId: ID!
3
  id: Int!
4
  address: String!
5
  employee: Employee
6
}

1
{
2
  employeeCollection {
3
    edges {
4
      node {
5
        name
6
        emailAddress {
7
          address
8
          employee {
9
            name
10
          }
11
        }
12
      }
13
    }
14
  }
15
}

Example

1
{
2
  "data": {
3
    "employeeCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "name": "Foo Barington",
8
            "emailAddress": {
9
              "address": "foo@bar.com",
10
              "employee": {
11
                "name": "Foo Barington"
12
              }
13
            }
14
          }
15
        }
16
      ]
17
    }
18
  }
19
}

Custom Scalars

Due to differences among the types supported by PostgreSQL, JSON, and GraphQL, pg_graphql adds several new Scalar types to handle PostgreSQL builtins that require special handling.

JSON

pg_graphql serializes json and jsonb data types as String under the custom scalar name JSON.

1
scalar JSON

Example

Given the setup

1
create table "User"(
2
    id bigserial primary key,
3
    config jsonb
4
);
5
6
insert into "User"(config)
7
values (jsonb_build_object('palette', 'dark-mode'));

1
type User {
2
  nodeId: ID!
3
  id: BigInt!
4
  config: JSON
5
}

The query

1
{
2
  userCollection {
3
    edges {
4
      node {
5
        config
6
      }
7
    }
8
  }
9
}

The returns the following data. Note that config is serialized as a string

1
{
2
  "data": {
3
    "userCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "config": "{\"palette\": \"dark-mode\"}"
8
          }
9
        }
10
      ]
11
    }
12
  }
13
}

Use serialized JSON strings when updating or inserting JSON fields via the GraphQL API.

JSON does not currently support filtering.

BigInt

PostgreSQL bigint and bigserial types are 64 bit integers. In contrast, JSON supports 32 bit integers.

Since PostgreSQL bigint values may be outside the min/max range allowed by JSON, they are represented in the GraphQL schema as BigInts and values are serialized as strings.

1
scalar BigInt
2
3
input BigIntFilter {
4
  eq: BigInt
5
  gt: BigInt
6
  gte: BigInt
7
  in: [BigInt!]
8
  lt: BigInt
9
  lte: BigInt
10
  neq: BigInt
11
  is: FilterIs
12
}

Example

Given the setup

1
create table "Person"(
2
    id bigserial primary key,
3
    name text
4
);
5
6
insert into "Person"(name)
7
values ('J. Bazworth');

1
type Person {
2
  nodeId: ID!
3
  id: BigInt!
4
  name: String
5
}

The query

1
{
2
  personCollection {
3
    edges {
4
      node {
5
        id
6
        name
7
      }
8
    }
9
  }
10
}

The returns the following data. Note that id is serialized as a string

1
{
2
  "data": {
3
    "personCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": "1",
8
            "name": "Foo Barington",
9
          }
10
        }
11
      ]
12
    }
13
  }
14
}

BigFloat

PostgreSQL's numeric type supports arbitrary precision floating point values. JSON's float is limited to 64-bit precision.

Since a PostgreSQL numeric may require more precision than can be handled by JSON, numeric types are represented in the GraphQL schema as BigFloat and values are serialized as strings.

1
scalar BigFloat
2
3
input BigFloatFilter {
4
  eq: BigFloat
5
  gt: BigFloat
6
  gte: BigFloat
7
  in: [BigFloat!]
8
  lt: BigFloat
9
  lte: BigFloat
10
  neq: BigFloat
11
  is: FilterIs
12
}

Example

Given the SQL setup

1
create table "GeneralLedger"(
2
    id serial primary key,
3
    amount numeric(10,2)
4
);
5
6
insert into "GeneralLedger"(amount)
7
values (22.15);

The query

1
{
2
  generalLedgerCollection {
3
    edges {
4
      node {
5
        id
6
        amount
7
      }
8
    }
9
  }
10
}

The returns the following data. Note that amount is serialized as a string

1
{
2
  "data": {
3
    "generalLedgerCollection": {
4
      "edges": [
5
        {
6
          "node": {
7
            "id": 1,
8
            "amount": "22.15",
9
          }
10
        }
11
      ]
12
    }
13
  }
14
}

Opaque

PostgreSQL's type system is extensible and not all types handle all operations e.g. filtering with like. To account for these, pg_graphql introduces a scalar Opaque type. The Opaque type uses PostgreSQL's to_json method to serialize values. That allows complex or unknown types to be included in the schema by delegating handling to the client.

1
scalar Opaque
2
3
input OpaqueFilter {
4
  eq: Opaque
5
  is: FilterIs
6
}

GraphQL API

Primary Keys (Required)

QueryType

Node

Collections

Aggregates

Keyset Pagination

Offset Pagination

Filtering

Ordering

Primary Key Queries

MutationType

Insert

Update

Delete

Concepts

nodeId

Relationships

One-to-Many

Many-to-One

One-to-One

Custom Scalars

JSON

BigInt

BigFloat

Opaque

Is this helpful?

AI tools