Supabase has reached General Availability

Learn more

What's New in pg_graphql v1.2

2023-04-21

6 minute read

It's been 4 months since the 1.0.0 release of pg_graphql. Since then, we’ve pushed several features to improve the APIs that pg_graphql produces.

In this article, we’ll walk through those features and show examples of each.

📢 These features are only available on projects with Postgres version 15.1.0.63 or higher. For help with upgrading, please review the migrating and upgrading projects guide.

View Support

Prior to v1.1, pg_graphql would only reflect standard tables. Since then, views, materialized views, and foreign tables are now also reflected in the GraphQL schema.

For example:


_10
create view "ProjectOwner" as
_10
select
_10
acc.id,
_10
acc.name
_10
from
_10
account as acc
_10
join role as r on r.id = acc.role_id
_10
where acc.role = 'project_owner';

Since all entities exposed by pg_graphql require primary keys, we must define that constraint for the view. We do that using a comment directive:


_10
comment on view "ProjectOwner"
_10
is '@graphql({"primary_key_columns": ["id"]})';

Which yields the GraphQL type:


_10
type ProjectOwner implements Node {
_10
nodeId: ID!
_10
id: UUID!
_10
name: String
_10
}

With associated Edge and Connection types. That enables querying via:


_10
{
_10
projectOwnerCollection(first: 2) {
_10
edges {
_10
node {
_10
nodeId
_10
name
_10
}
_10
}
_10
}
_10
}

Additionally, simple views automatically support mutation events like inserts and updates. You might use these to migrate underlying tables while maintaining backwards compatibility with previous API versions.

Filtering

Filtering in SQL is endlessly flexible. We’ve taken two incremental steps to bring more of that flexibility to the GraphQL interface.

is null and is not null

Handling null values can be tricky in both SQL and GraphQL. However, there are similarities we can take advantage of. In pg_graphql, every scalar data type has its own filter type, such as IntFilter and StringFilter. Each of these filter types now includes an is argument, which allows you to filter based on whether a value is null or not null. You can do this by using {is: NULL} for null values and {is: NOT_NULL} for non-null values.


_10
enum FilterIs {
_10
NULL
_10
NOT_NULL
_10
}
_10
_10
type IntFilter {
_10
...
_10
is: FilterIs
_10
}

For example:


_10
{
_10
blogCollection(filter: { name: {is: NULL}}) {
_10
...
_10
}
_10
}

to return all blogs where the name is null.

like, ilike, and startsWith

Text filtering options in pg_graphql have historically been restricted to equality checks. The hesitation was due to concerns about exposing a default filter that is difficult to index. The combination of citext and PGroonga available on the platform solves those scalability risks and enabled us to expand the StringFilter with options for like ilike and startsWith.


_10
input StringFilter {
_10
eq: String
_10
...
_10
startsWith: String
_10
like: String
_10
ilike: String
_10
}

Note that startsWith filters should be preferred where appropriate because they can leverage simple B-Tree indexes to improve performance.


_11
{
_11
generalLedgerCollection(filter: { identifierCode: { startsWith: "BX1:" } }) {
_11
edges {
_11
node {
_11
nodeId
_11
identifierCode
_11
amount
_11
}
_11
}
_11
}
_11
}

GraphQL directives @skip and @include

The GraphQL spec has evolved over time. Although the spec is clear, it is common for GraphQL servers to selectively omit some chunks of functionality. For example, some frameworks intentionally do not expose an introspection schema as a form of security through obscurity.

pg_graphql aims to be unopinionated and adhere exactly to the spec. The @skip and @include directives are part of the GraphQL core specification and are now functional.

The @skip directive in GraphQL is used to conditionally skip a field or fragment during query execution based on a Boolean variable. It can be used to make the query more efficient by reducing the amount of data retrieved from the server.

The @include directive is the mirror of @skip where a field or fragment is conditionally included depending on the value of a Boolean variable.

Here's an example of how the @skip directive can be used in a GraphQL query:


_11
query getBooks($includeDetails: Boolean!) {
_11
booksCollection {
_11
edges {
_11
node {
_11
id
_11
title
_11
description @skip(if: $includeDetails)
_11
}
_11
}
_11
}
_11
}

User Defined Descriptions

Users can now use the comment directive system to assign descriptions to tables, views and columns.


_10
create table public.book(
_10
id int primary key,
_10
title text not null
_10
);
_10
_10
comment on table public.book
_10
is e'@graphql({"description": "a library book"})';
_10
_10
comment on column public.book.title
_10
is e'@graphql({"description": "the title of the book"})';

GraphQL IDEs, such as GraphiQL render those descriptions, allowing developers to provide clearer API documentation.

Roadmap

The headline features we aim to launch in coming releases of pg_graphql include:

  1. Support for user-defined functions: GitHub issue
  2. Support for nested inserts: GitHub issue
  3. An alternative approach to computed relationships based on SQL functions returning SET OF rather than comment directives (compatible with PostgREST)

More pg_graphql

Share this article

Build in a weekend, scale to millions