Thank you 🙂

I think they'll be saved at a project level to begin with, and then we'll work in other scopes for user (private) and organization over time

@User nice - yep saved scopes would be great too. Not sure if you've seen BigQuery on GCP but they have a similar setup which is quite nice.

nice! some of us will be familiar with it, I personally can't remember - we'll check it out again though

Looking for some inspiration on project ideas, what are y’all building with supabase ⚡️

Hi, I would like to suggest the option of having Anonymous login for authentication guest user. It will be really helpful to have this for my project.

Already asked here https://github.com/supabase/gotrue/issues/170

disable email/password login when auth providers are in use

Allow users to "link" accounts, for example, log in with google, link discord, sign in with discord, same account.

NoSQL

CORS support - details in thread

Link Accounts

A read-only URL and Key for writing tests for packages. Similar to Stripe, which has a test key and URL you can use to initiate a test checkout. This would be useful for making more standard tests for Supabase libraries without having to create a new project for each one, and worry about security etc for keys that may be public on GitHub.

NoSQL

I wish there was a way to easily upgrade a project to the latest set up. It feels a bit as if I had to have a new iPhone every time there’s an iOS update.

Supabase is still in beta and this should be expected at times, I'm sure the team is working on making this process easier in the future.

I'm using a Java backend. Looks like the easier way to do auth is to use the vanilla JS example as a starting point for auth and then set cookies to send back to the Java server.

Here's a gist of my adapted vanilla JS script: https://gist.github.com/wiverson/615a31b0ac9c85ee698ba164f0b30eb1

If I'm reading it all right, the supabase auth JS will refresh the auth tokens as the user browses, and then I can use the cookie-sent tokens to validate on the server as needed...?

Idea: I'd like to see a simpler API for accessing table relationships. Currently I have a query that looks like this:

const { data: pack, error: packError } = await supabase
      .from('trip_packs')
      .select(
        `
          pack:pack_id(*, categories(*, category_items(*), items(*))) 
        `
      )
      .match({ trip_id: tripId, user_id: me.id });

As you can see, it's a fairly complicated select method, trying to access data several relationships deep. It could be cool to do something like Prisma, which looks like this:

const getUser = await prisma.user.findUnique({
  where: {
    id: 19,
  },
  include: {
    posts: true,
  },
})

^ a user's posts are included. Or even better (I come from a Rails background), the ability to set up eager loading relationships. So in my first example, I would just query for a Pack, and all of the nested categories and items would automatically be included in the query. Thanks for coming to my Ted Talk!

Supabase uses postgrest underneath, you could use any other PostgreSQL client that could serve you better. For me personally, I went with procedure functions to do complex queries and return only data I need.

ah interesting, so the api for a

.select()

actually comes from Postgrest and not Supabase?

All of the query requests sent by the library are routed through Postgrest.

.subscribe()

is a little different, but essentially the initial query routes through Postgrest, and then connects to the realtime server to handle listening for changes. Auth ones are routed through GoTrue. In all cases, they all go through the Kong server.

so theoretically, making the .select() api more "user friendly" on the supabase side of things is possible?

Another idea: a "duplicate row" button could be helpful for the table editor dashboard UI. As I'm developing a new application, I keep having to copy values to a text document, create a new row, then insert all of those values with minor changes to the new row

Is there any reason behind the decision to use shortened words for the modifier syntax? Eg: why use

.eq

when

.equal

would be more user friendly and easy to read and pick up. If there isn’t, I’d ask for adding the long names alongside the short ones

I use DataGrip to query and manipulate the database. It does a brilliant job. So for example if a query is complex I’ll build it as a view in DataGrip and then have a simple query in Supabase-JS. Everything comes out flat in a single table row but if you rename your columns with strings like this: users.name AS “user.name” then use unflatten https://www.npmjs.com/package/flat to create nested JS objects

Sounds cool! Is this in reference to my other post about nested relationship queries?

I use merge and groupBy from Lodash to deal with many to many join-table queries which usually return many rows which should be grouped into single JS object

Yeah and just in general because this is an ideas thread and I went through a period of thinking the Supabase team need to be building a better way to manipulate the database. But now I think actually there are really mature tools for that already so maybe we would prefer them to focus on Supabase specific tooling.