pref: use new.target to replace instanceof

[xtx1130]

as titled

[mcollina]

lgtm

[mcollina]

how much does this improve things?

[mcollina]

For the record:

const max = 100000000

function InstanceOfClass () {
  if (!(this instanceof InstanceOfClass)) {
    return new InstanceOfClass()
  }
}

function NewTargetClass () {
  if (!new.target) {
    return new NewTargetClass()
  }
}

console.time('InstanceOfClass')
for (let i = 0; i < max; i++) {
  InstanceOfClass()
}
console.timeEnd('InstanceOfClass')

console.time('NewTargetClass')
for (let i = 0; i < max; i++) {
  NewTargetClass()
}
console.timeEnd('NewTargetClass')

results in:

InstanceOfClass: 3.741s
NewTargetClass: 1.154s

Impressive, I didn't know about this!

[kurtextrem]

If we're talking about perf, we should maybe use the fastest pattern, explicit === undefined: https://esbench.com/bench/620f77c86c89f600a570162f

[mcollina]

Would you like to send a PR to update this?

[xtx1130]

The result is almost the same between === undefined and !

[xtx1130]

I just saw the mentioned above. It is really faster.

const max = 100000000

function InstanceOfClass () {
  if (!(this instanceof InstanceOfClass)) {
    return new InstanceOfClass()
  }
}

function NewTargetClass () {
  if (!new.target) {
    return new NewTargetClass()
  }
}

function NewTargetClassUndefined () {
  if (new.target === undefined) {
    return new NewTargetClass()
  }
}

console.time('NewTargetClass')
for (let i = 0; i < max; i++) {
  NewTargetClass()
}
console.timeEnd('NewTargetClass')

console.time('InstanceOfClass')
for (let i = 0; i < max; i++) {
  InstanceOfClass()
}
console.timeEnd('InstanceOfClass')

console.time('NewTargetClassWithUndefined')
for (let i = 0; i < max; i++) {
  NewTargetClassUndefined()
}
console.timeEnd('NewTargetClassWithUndefined')

The result is :

NewTargetClass: 1.169s
InstanceOfClass: 3.983s
NewTargetClassWithUndefined: 52.08ms

=== undefined is more faster at nodejs v16, @kurtextrem would you mind to send a PR to update this?

[kurtextrem]

At this point we probably shouldn't care or not put any further time into this, see #61