perf: use map for sting-based stores

[gurgunday]

When an object is used as a map and has string-based keys, Map's performance will be superior

For this repo, not sure how to test it but here's some related discussion and benchmarks in kibertoad/toad-cache#40

And a V8 engineer explaining why: https://stackoverflow.com/questions/62350146/why-map-manipulation-is-much-slower-than-object-in-javascript-v8-for-integ/62351925#62351925

[mcollina]

Did you run the benchmarks? Where there any improvements?

[gurgunday]

I will post a benchmark if I can

[gurgunday]

Yeah no significant improvement here

[gurgunday]

I don't believe this didn't improve anything, I'll take another look 🤣

[gurgunday]

@mcollina so I ran some tests again

For keys that aren't integers, Map is still consistently faster, it also doesn't force V8 to reconsider the hidden class a bunch of times

We had the same results in toad-cache, so it's perfectly safe to land

We don't have a benchmark that tests these because they are only used in Constrainer, but here is a generic benchmark:

Map time for 1000 iterations: 0.90 ms
Object time for 1000 iterations: 1.60 ms

function bench(iterations) {
    const map = new Map();
    const obj = {};

    for (let i = 1; i <= 10; i++) {
        map.set(`${i}.x.x`, `host${i}`);
        obj[`${i}.x.x`] = `host${i}`;
    }

    function benchmarkMap() {
        let sum = 0;
        for (let i = 0; i < iterations; i++) {
            for (let j = 1; j <= 10; j++) {
                sum += map.get(`${j}.x.x`).length;
            }
        }
        return sum;
    }

    function benchmarkObject() {
        let sum = 0;
        for (let i = 0; i < iterations; i++) {
            for (let j = 1; j <= 10; j++) {
                sum += obj[`${j}.x.x`].length;
            }
        }
        return sum;
    }

    const startMap = performance.now();
    benchmarkMap();
    const endMap = performance.now();

    const startObject = performance.now();
    benchmarkObject();
    const endObject = performance.now();

    console.log(`Map time for ${iterations} iterations:`, (endMap - startMap).toFixed(2), "ms");
    console.log(`Object time for ${iterations} iterations:`, (endObject - startObject).toFixed(2), "ms");
}

bench(1000);

[mcollina]

lgtm