Fix histogram scrape performance

[nowells]

When profiling histogram scraping I discovered that the major bottleneck is creating a clone of the accumulator in reduce. We should not actually need to skip mutation.

You can see the benchmark of concat vs push here: https://codeburst.io/jsnoob-push-vs-concat-basics-and-performance-comparison-7a4b55242fa9

This should address #216

I used this test script to profile.

const prom = require('prom-client');
const _ = require('lodash');

const histogram = new prom.Histogram({
    name: 'histogram',
    help: 'histogram',
    labelNames: ['a', 'b', 'c', 'd'],
});

_.times(2, a => {
    _.times(3, b => {
        _.times(10, c => {
            _.times(100, d => {
                const end = histogram.startTimer();

                end({a,b,c,d});
            });
        });
    });
});

console.log('Starting metrics');
const now = Date.now();
const metrics = prom.register.getMetricsAsJSON();

console.log(`${(Date.now() - now) / 1000}s`);

Which resulted in before this patch of 7.021s and after of 0.501s

[SimenB]

why not just acc.push(d.buckets);?

[SimenB]

oh, it's an array. makes sense :)

[SimenB]

it can be acc.push(...d.buckets);, we have a minimum node version of 6

[nowells]

yeah, I didn't look up minimum node version, I had it that way initially but then decided to start off with more support. I'll update to the ES6 syntax.

[SimenB]

Thanks for the PR! Mind updating the changelog as well?

[nowells]

Thanks for the PR! Mind updating the changelog as well?

Done

[nowells]

Thanks for the merge @SimenB! I did not package bump because I assume that you do that as part of your release process. Thanks for the great package, and for such a quick response!

[SimenB]

It's released as 11.1.2 now 🙂

[nowells]

It's released as 11.1.2 now 🙂

❤️ thanks!!!

[siimon]

So nice!! Thank you! ❤️