Blog post

PostgREST 11 pre-release

2022-12-16

5 minute read

PostgREST 11 pre-release

PostgREST 11 is not wrapped up yet, however a pre-release with the latest features and fixes is available on the Supabase CLI.

In this blog post we'll cover some of the improved querying capabilities: spreading related tables, related orders and anti-joins.

Very often the way we structure a database is not the way we want to present it to the frontend application. For example, let's assume we have a films and technical_specs tables and they form a one-to-one relationship.

Using PostgREST resource embedding, we can query them in one request like so

From HTTP:

_10
GET /films?select=title,technical_specs(camera,laboratory,sound_mix)

or JavaScript:

_10
const { data, error } = await supabase.from('films').select(`
_10
title,
_10
technical_specs (
_10
camera, laboratory, duration
_10
)
_10
`)

Response:

_11
[
_11
{
_11
"title": "Pulp Fiction",
_11
"technical_specs": {
_11
"camera": "Arriflex 35-III",
_11
"laboratory": "DeLuxe, Hollywood (CA), USA (color)",
_11
"duration": "02:34:00"
_11
}
_11
},
_11
"..."
_11
]

But we'd like to present a “flattened” result to the frontend, without the technical_specs object. For this we could create a new database view or function that shapes the json the way we want, but creating extra database objects is not always convenient.

Using the new “spread” operator(syntax borrowed from JS), we can expand a related table columns and remove the nested object.

From HTTP:

_10
GET /films?select=title,...technical_specs(camera,laboratory,duration)

or JavaScript:

_10
const { data, error } = await supabase.from('films').select(`
_10
title,
_10
...technical_specs (
_10
camera, laboratory, duration
_10
)
_10
`)

Response:

_10
[
_10
{
_10
"title": "Pulp Fiction",
_10
"camera": "Arriflex 35-III",
_10
"laboratory": "DeLuxe, Hollywood (CA), USA (color)",
_10
"duration": "02:34:00"
_10
},
_10
"..."
_10
]

This only works for one-to-one and many-to-one relationships for now but we're looking at ways to remove this restriction.

It's also a common use case to order a table by a related table column. For example, suppose you'd like to order films based on the technical_specs.duration column.

You can now do it like so:

From HTTP:

_10
GET /films?select=title,...technical_specs(duration)&order=technical_specs(duration).desc

or JavaScript:

_10
const { data, error } = await supabase
_10
.from('films')
_10
.select(`
_10
title,
_10
...technical_specs (
_10
duration
_10
)
_10
`)
_10
.order('technical_specs(duration)', { descending: true }))

Response:

_11
[
_11
{
_11
"title": "Amra Ekta Cinema Banabo",
_11
"duration": "21:05:00"
_11
},
_11
{
_11
"title": "Resan",
_11
"duration": "14:33:00"
_11
},
_11
"..."
_11
]

Similarly to spreading related tables, this only works for one-to-one and many-to-one relationships.

Anti-Joins

To do the equivalent of a left anti-join, you can now filter the rows where the related table is null.

From HTTP:

_10
GET /films?select=title,nominations()&nominations=is.null

or JavaScript:

_10
const { data, error } = await supabase
_10
.from('films')
_10
.select(`
_10
title,
_10
nominations()
_10
`)
_10
.is('nominations', null))

Response:

_12
[
_12
{
_12
"title": "Memories of Murder"
_12
},
_12
{
_12
"title": "Rush"
_12
},
_12
{
_12
"title": "Groundhog Day"
_12
},
_12
"..."
_12
]

Note that nominations doesn't select any columns so they don't show on the resulting response.

The equivalent of an inner join can be done by filtering the rows where the related table is not null.


_10
GET /films?select=title,nominations(rank,...competitions(name))&nominations=not.is.null


_10
const { data, error } = await supabase
_10
.from('films')
_10
.select(
_10
`
_10
title,
_10
nominations(rank,...competitions(name))
_10
`
_10
)
_10
.not('nominations', 'is', null)

Response:

_11
[
_11
{
_11
"title": "Pulp Fiction"
_11
"nominations": [
_11
{"rank": 1, "name": "Palme d'Or"},
_11
{"rank": 1, "name": "BAFTA Film Award"},
_11
{"..."}
_11
]
_11
},
_11
"..."
_11
]

This was already possible with the !inner modifier(introduced on PostgREST 9) but the not null filter is more flexible and can be used with an or filter to combine related tables' conditions.

Try it out

This pre-release is not deployed to Supabase cloud but you can try it out locally with the Supabase CLI.


_10
$ supabase start

Please try it and report any bugs, suggestions or ideas!

More Launch Week 6

Share this article

Last post

Supabase Vault is now in Beta

16 December 2022

Next post

Point in Time Recovery is now available for Pro projects

16 December 2022

Related articles

Supabase Beta May 2023

Supabase Vecs: a vector client for Postgres

Flutter Hackathon Winners

ChatGPT plugins now support Postgres & Supabase

Building ChatGPT Plugins with Supabase Edge Runtime

Build in a weekend, scale to millions