`Bun.sql` tracking issue (Postgres client)

[Jarred-Sumner]

Bun.sql is Bun's builtin postgres client, currently only available on canary builds of Bun.

import { sql } from 'bun';

const [{x}] = await sql`select 1 as x`;
console.log({x});

sqlite and other database protocols will be added sometime after it ships.

Minimum necessary to unflag for Bun v1.2:

• [x] Pass primitive types back and forth
• [x] Pass objects back and forth
• [x] Support sql(array) in INSERT INTO queries
• [x] Query pipelining
• [x] SCRAM-SHA-256 authentication
• [ ] MD5 authentication
• [ ] TLS support
• [ ] sslmode=require
• [ ] Connection timeouts
• [ ] Make sure it works on RDS
• [ ] Make sure it works on Neon
• [ ] Make sure it works on Supabase

Feature complete:

• [ ] Support COPY protocol
• [ ] Round-robin database connections, automatic graceful connection closing after timeout
• [ ] Support async iterators
• [ ] Support Point & other geo-related types

Nice to have:

• [ ] --sql-preconnect CLI flag to start connecting to DB before code loads
• [ ] sql.begin(callback) API for transactions
• [ ] sql.array(array) for various operations involving arrays
• [ ] Query instrumentation / logging

[aquapi]

Can this be an import from bun:postgres to be more clear?

import { sql } from 'bun:postgres';

const [{ x }] = await sql`select 1 as x`;
console.log({ x });

[truongan07]

Use .iterate() to run a query and incrementally return results. This is useful for large result sets that you want to process one row at a time without loading all the results into memory.

Support this method in Postgres client, please.

[versecafe]

What's the reason for a import { sql } from ... instead of matching the SQLite setup and just swapping it to bun:postgres

[rafaell-lycan]

I'd like to see some features such as flexible prepared statements and Pool as both exist on libs such as Postgres.js and Node-PG.

This is personal but the prepared statements on pg lib feel quite easy to use, but no hard feelings or anything really:

sql(
  'INSERT INTO photos(user_id, photo_url) VALUES ($1, $2)',
  ['1', 'example.com']
)

/**
 * Alternatively you could also accept an object as a param
 * which could have more metadata to improve tracing later on:
 */

sql({
  name: 'add-photo',
  text: 'INSERT INTO photos(user_id, photo_url) VALUES ($1, $2)',
  values: ['1', 'example.com'],
})

It can be a future enhancement tho.