Does bun:sql support PostgreSQL COPY FROM STDIN for bulk inserts?

[bylly1]

I'm using bun:sql (via Drizzle ORM / drizzle-orm/bun-sql) and need to
bulk insert large amounts of rows into a partitioned PostgreSQL table.

PostgreSQL's COPY FROM STDIN is significantly faster than multi-row
INSERT for this use case.

Questions:

1. Does bun:sql currently support COPY FROM STDIN?
2. If not, is it on the roadmap?
3. Is there a workaround to stream data directly using the PostgreSQL
   COPY protocol through bun:sql?

[sueun-dev]

No, bun:sql doesn't do COPY FROM STDIN yet (tested on 1.3.14). There's no copy/stream method on the sql object, and if you just send COPY t FROM STDIN, the server enters copy mode and the call hangs, since bun:sql has no way to push the CopyData frames. It is being added though, there's an open PR: #23350.

For now the fastest thing that actually works is bun:sql's multi-row INSERT helper:

await sql`INSERT INTO t ${sql(rows)}`;  // rows = array of objects

That sends a single multi-row INSERT (checked the wire log, 1000 rows in one statement). Since drizzle-orm/bun-sql rides on top of bun:sql you hit the same limit, so use one db.insert(t).values(bigArray), which also compiles to a single multi-row INSERT. If you genuinely need COPY streaming before that PR lands, you'd have to drop to a client that implements the COPY protocol.