Using Promise.all inside a transaction / task vs using t.batch in pg-promise

Question

The documentation here clearly outlines that it is a bad practice to write

Promise.all(data.map(d => db.none('insert into...', d)))

which is querying against root database protocol. Hence it is better to write

db.task(t => t.batch(data.map(d => t.none('insert into...', d))));

However, is it wrong to write this if I do not intend to use BatchError or query duration (ref)?

db.task(async t => {
  await Promise.all(data.map(d => t.none('insert into...', d)));
});

By the looks of it, it still uses shared connection protocol instead of root connection.

Answer 1

The coming of ES7 with async/await to the scene, dissipated the need for putting queries in a batch.

You can simply do this instead:

await db.task(async t => {
    await forEachAsync(data, d => t.none('insert into...', d));
});

or just return the result:

await db.task(t => forEachAsync(data, d => t.none('insert into...', d)));

That would need async for-each function like this:

async function forEachAsync(data, cb) {
    for (let i = 0; i < data.length; i ++) {
        await cb(data[i], i, data);
    }
}

But normally, you do not need any of these when executing an array of inserts, because for that you should be using multi-row inserts instead.