remove sliced

[Uzlopak]

Remove the package "sliced" in favor of native functions.

[Uzlopak]

Closing for now, as I am unsure if this has a perf impact.

[Uzlopak]

Here some benchmarks on comparing spread-operator, Array.from and sliced

var sliced = require('./')
var Bench = require('benchmark');
var s = new Bench.Suite;
var slice = [].slice;

var testArray = new Array(10).fill("a")


s.add('[...]', function () {
  [...testArray];
}).add('Array.from', function () {
  Array.from(testArray);
}).add('Array.prototype.slice.call(testArray)', function () {
  Array.prototype.slice.call(testArray);
}).add('cached slice.call', function () {
  slice.call(testArray)
}).add('sliced', function () {
  sliced(testArray)
}).on('cycle', function (evt) {
  console.log(String(evt.target));
}).on('complete', function () {
  console.log('fastest is %s', this.filter('fastest').pluck('name'));
})
.run();

with 10 elements

[...] x 33,485,060 ops/sec ±1.69% (84 runs sampled)
Array.from x 25,203,273 ops/sec ±0.55% (96 runs sampled)
Array.prototype.slice.call(testArray) x 30,466,867 ops/sec ±0.58% (97 runs sampled)
cached slice.call x 31,882,883 ops/sec ±0.64% (94 runs sampled)
sliced x 18,666,861 ops/sec ±0.94% (94 runs sampled)
fastest is [ '[...]' ]

with 1000 elements:

[...] x 510,025 ops/sec ±0.85% (94 runs sampled)
Array.from x 538,160 ops/sec ±1.06% (91 runs sampled)
Array.prototype.slice.call(testArray) x 887,796 ops/sec ±0.63% (99 runs sampled)
cached slice.call x 949,157 ops/sec ±1.06% (97 runs sampled)
sliced x 308,079 ops/sec ±0.53% (99 runs sampled)
fastest is [ 'cached slice.call' ]

So [...] and Array.from are about the same speed. slice.call is not usable because arguments is not an array but an iteratable. But in any case sliced is slower than the built-ins.

[Uzlopak]

here i used normal slice, as I think debug is not used in production?!

[vkarpov15]

Debug can be used in production. I'm sure some people use it for logging or perf. Shouldn't assume that certain features are never used in prod.

[Uzlopak]

Also

var sliced = require('./')
var Bench = require('benchmark');
var s = new Bench.Suite;
var slice = [].slice;

var testArray = new Array(3).fill("a")


s.add('direct', function () {
  [testArray[0], testArray[1], testArray[2]];
}).add('[...]', function () {
  [...testArray];
}).add('Array.from', function () {
  Array.from(testArray);
}).add('Array.prototype.slice.call(testArray)', function () {
  Array.prototype.slice.call(testArray);
}).add('cached slice.call', function () {
  slice.call(testArray)
}).add('sliced', function () {
  sliced(testArray)
}).on('cycle', function (evt) {
  console.log(String(evt.target));
}).on('complete', function () {
  console.log('fastest is %s', this.filter('fastest').pluck('name'));
})
.run();

direct x 2,804,270,988 ops/sec ±0.83% (100 runs sampled)
[...] x 47,105,492 ops/sec ±0.88% (92 runs sampled)
Array.from x 31,716,899 ops/sec ±0.92% (93 runs sampled)
Array.prototype.slice.call(testArray) x 36,296,758 ops/sec ±0.80% (94 runs sampled)
cached slice.call x 36,020,585 ops/sec ±0.71% (94 runs sampled)
sliced x 28,710,453 ops/sec ±1.06% (91 runs sampled)
fastest is [ 'direct' ]

So if we know that the length of the array is 2 or 3, than directly assigning the values to the new Array is much faster...

[Uzlopak]

I dont know why removing this line makes the unit tests fail.

[Uzlopak]

I checked mongoose sourcecode. It seems that the spread operator is not used usually, but everywhere Array.from. I can change the occurences of the spread operator to Array.from if it is desired.

[vkarpov15]

Nice work, thanks 👍 I like removing the extra dependency, especially if the benchmarks that justified using sliced are no longer accurate.

[AbdelrahmanHafez]

Nice work, thanks @Uzlopak 👍