Can you add benchmarking for cloning?

[matthew-dean]

I'm not trying to micro-optimize, but at the same time, I've been leaning more and more on this library in my current project just to not have to think about performance in most of my data structure.

My question is around cloning performance. I notice that, for example, in the HashMap clone() method, it does a for / of over the iterator. (What's also interesting / weird, is that cloning in many structures creates a specific class instead of creating an instance of this.constructor, which means that it's somewhat an unreliable clone, if classes are extended. Can you explain that choice?)

A while back, I set up some tests for Map / Set because I was kind of baffled to learn that, in Chromium, creating a Map by passing in a Map was slower than iterating over each element. I passed that issue over to the Chromium team, but in the process of doing those tests and some experimentation, I found that, currently, there's even a fastest-path iteration in Chrome, which is a while loop, like:

let clone = new Map()
let s
let entries = map.entries()
while (s = entries.next(), s.done !== true) {
	let entry = s.value
	clone.set(entry[0], entry[1])
}

This is currently decently faster than for / of in Chromium. In Firefox, as of today, for / of and while are roughly equal, and in Safari, they actually seem to have figured out that cloning a map from another map is as simple as an internal memory copy, so new Map(mapToClone) is orders of magnitude faster than both other engines.

So, anyways, seeing cloning in the performance tests would be great. Thanks!