How does this work with pipelining?

[feelingsonice]

Sorry if this is a dumb question, trying to understand how this would work. tokio_postgres states:

Pipelining happens automatically when futures are polled concurrently

So am I correct to then assume that pipelining will never naturally happen? Especially in a standard HTTP server? Because pool.get() will always return an unused connection hence there'll never be futures polled concurrently on the same connection?

[bikeshedder]

You can perform pipelining with connections from tokio-postgres regardless if you create the connection manually or get them from the pool.

There are two use cases:

• One worker issues two (or more) queries at the same time using a single connection.
  This works just as explained in the tokio-postgres documentation:

  async fn pipelined_prepare(
      client: &Client,
  ) -> Result<(Statement, Statement), Error>
  {
      future::try_join(
          client.prepare("SELECT * FROM foo"),
          client.prepare("INSERT INTO bar (id, name) VALUES ($1, $2)")
      ).await
  }

• Two (or more) workers sharing a single connection and using pipelining that way.
  This requires a different way to think about pooled objects altogether and is very tokio-postgres specific as most DB backends don't really support sharing of connections.

  There is a tracking issue for that feature: Mechanism for requesting shared Objects #196