how is the way for cancel query?

Can I cancel a query inside query.on(‘row’…?

I want to do something like:

query.on('row',function(row, result){
    if(row.field1=='ok'){
        result.addRow(row);
    }else{
        result.cancel(row, 'bla bla bla'); 
    }
});

The need is to not continue fetching rows if I detect an error condition.

[robhaswell]

You should be able to cancel queries using pg_cancel_backend(). I think the pattern looks like this:

const pool = pg.Pool()

/* Check out a client */
pool.connect()
.then((c) => {
  /* Get my PID */
  c.query('SELECT pg_backend_pid()')
  .then((result) => {
    const pid = result.rows[0][0]

    const badResult = c.query('SELECT pg_sleep(10000000000)')

    /* Kill this after 1 second */
    setTimeout(() => {
      /* NOTE this query is executed by the pool */
      const cancelResult = pool.query('SELECT pg_cancel_backend($1)', [ pid ])
    }, 1000)
  })
})

[robhaswell]

start the governing connection before starting the cancellable query

This is VERY expensive, especially in a scenario where resources are limited.

A better approach would be to have a reserved pool for cancelling connections:

// My platform has allocated me 20 connections
const pool = new Pool({max: 19})
const adminPool = new Pool({max: 1})

const client = await pool.connect()
client.query('SELECT pg_sleep(1000000)')

const otherClient = await adminPool.connect()
otherClient.cancel(client)
// - or, preferably: -
await adminPool.cancel(client) // like Pool.query

Whether you want the user to instantiate the admin pool, or allow them to specify the number of reserved connections, I think either is fine. I expect the latter would be quite a significant refactor, to have Pool manage two logical pools.