apps/www/_blog/2021-07-30-supabase-functions-updates.mdx
The question on everyone's mind - are we launching Supabase Functions? Well, it's complicated.
Today we're announcing part of Functions - Supabase Hooks - in Alpha, for all new projects.
We're also releasing support for Postgres Functions and Triggers in our Dashboard, and some timelines for the rest of Supabase Functions. Let's cover the features we're launching today before the item that everyone is waiting for: Supabase Functions.
(Not to be confused with Supabase Functions!)
Postgres has built-in support for SQL functions. Today we're making it even easier for developers to build PostgreSQL Functions by releasing a native Functions editor. Soon we'll release some handy templates!
You can call PostgreSQL Functions with supabase-js using your project API [Docs]:
const { data, error } = await supabase.rpc('best_star_wars_series', {
name: 'The Prequels',
})
Triggers are another amazing feature of Postgres, which allows you to execute any SQL code after inserting, updating, or deleting data.
While triggers are a staple of Database Administrators, they can be a bit complex and hard to use. We plan to change that with a simple interface for building and managing PostgreSQL triggers.
They say building a startup is like jumping off a cliff and assembling the plane on the way down. At Supabase it's more like assembling a 747 since, although we're still in Beta, thousands of companies depend on us to power their apps and websites.
For the past few months we've been designing Supabase Functions based on our customer feedback.
A recurring request from our customers is the ability to trigger their existing Functions. This is especially true for our Enterprise customers, but also Jamstack developers who develop API Functions directly within their stack (like Next.js API routes, or Redwood Serverless Functions).
To meet these goals, we're releasing Supabase Functions in stages:
(Note: Database Webhooks were previously called "Function Hooks")
Today we're releasing Functions Hooks in ALPHA. The ALPHA tag means that it is NOT stable, but it's available for testing and feedback. The API will change, so do not use it for anything critical. You have been warned.
Hooks? Triggers? Firestore has the concept of Function Triggers, which are very cool. Supabase Hooks are the same concept, just with a different name. Postgres already has the concept of Triggers, and we thought this would be less confusing1.
Database Webhooks allow you to "listen" to any change in your tables to trigger an asynchronous Function. You can hook into a few different events:INSERT, UPDATE, and DELETE. All events are fired after a database row is changed. Keen eyes will be able to spot the similarity to Postgres triggers, and that's because Database Webhooks are just a convenience wrapper around triggers.
Supabase will support several different targets.
If the target is a Serverless function or an HTTP POST request, the payload is automatically generated from the underlying table data. The format matches Supabase Realtime, except in this case you don't a client to "listen" to the changes. This provides yet another mechanism for responding to database changes.
type InsertPayload = {
type: 'INSERT'
table: string
schema: string
record: TableRecord<T>
old_record: null
}
type UpdatePayload = {
type: 'UPDATE'
table: string
schema: string
record: TableRecord<T>
old_record: TableRecord<T>
}
type DeletePayload = {
type: 'DELETE'
table: string
schema: string
record: null
old_record: TableRecord<T>
}
pg_net v0.1As with most of the Supabase platform, we leverage PostgreSQL's native functionality to implement Database Webhooks (previously called "Function Hooks").
To build hooks, we've released a new PostgreSQL Extension, pg_net, an asynchronous networking extension with an emphasis on scalability/throughput. In its initial (unstable) release we expose:
GET requests.POST requests with a JSON payload.The extension is (currently) capable of >300 requests per second and is the networking layer underpinning Database Webhooks. For a complete view of capabilities, check out the docs.
pg_net allows you to make asynchronous HTTP requests directly within your SQL queries.
-- Make a request
select
net.http_post(
url:='https://httpbin.org/post',
body:='{"hello": "world"}'::jsonb
);
-- Immediately returns a response ID
http_post
---------
1
After making a request, the extension will return an ID. You can use this ID to collect a response.
-- Collect the response from a request
select
*
from
net.http_collect_response(1);
-- Results (1 row)
status | message | response
--------+---------+----------
SUCCESS ok (
status_code := 200,
headers := '{"date": ...}',
body := '{"args": ...}'
)::net.http_response_result
You can cast the response to JSON within PostgreSQL:
-- Collect the response json payload from a request
select
(response).body::json
from
net.http_collect_response(request_id:=1);
Result:
{
"args": {},
"data": "{\"hello\": \"world\"}",
"files": {},
"form": {},
"headers": {
"Accept": "*/*",
"Content-Length": "18",
"Content-Type": "application/json",
"Host": "httpbin.org",
"User-Agent": "pg_net/0.1",
"X-Amzn-Trace-Id": "Root=1-61031a5c-7e1afeae69bffa8614d8e48e"
},
"json": {
"hello": "world"
},
"origin": "135.63.38.488",
"url": "https://httpbin.org/post"
}
(1 row)
To build asynchronous behavior, we use a PostgreSQL background worker with a queue. This, coupled with the libcurl multi interface, enables us to do multiple simultaneous requests in the same background worker process.
Shout out to Paul Ramsey, who gave us the implementation idea in pgsql-http. While we originally hoped to add background workers to his extension, the implementation became too cumbersome and we decided to start with a clean slate. The advantage of being async can be seen by making some requests with both extensions:
\timing on
-- using pgsql-http to fetch from "supabase.io" 10 times
select
*
from
http_get('https://supabase.com')
cross join
generate_series(1, 10) _;
-- Returns in 3.5 seconds
Time: 3501.935 ms
-- using pg_net to fetch from "supabase.io" 10 times
select
net.http_get('https://supabase.com')
from
generate_series (1,10) _;
-- Returns in 1.5 milliseconds
Time: 1.562 ms
Of course, the sync version waits until each request is completed to return the result, taking around 3.5 seconds for 10 requests; while the async version returns almost immediately in 1.5 milliseconds. This is really important for Supabase hooks, which run requests for every event fired from a SQL trigger - potentially thousands of requests per second.
This is only the beginning! First we'll thoroughly test it and make a stable release, then we expect to add support for
PUT / PATCH)Database Webhooks is enabled today on all new projects. Find it under Database > Alpha Preview > Database Webhooks.