Clever Geek Handbook

原生并不意味着快速。在大型阵列上超车地图，过滤和缩小



几天前，我为小孩子们布置了一个关于JavaScript的LINQ帖子。但是我的库在性能上不如本地方法和Lodash。总的来说，现在我们将改变这种状况。



我马上要说：本文将不包含任何启示，我们将不构建疯狂的算法，等等。我们将仅比较不同语言结构的性能。



当我第一次向公众发布我的库时，我唯一的问题是如何减少迭代本机方法的成本。我试图使回调更加容易，因为我什至从未想到过map，filter或reduce可能比光速还慢一点。



在开始之前





让我们谈谈统计



我们将测量以下方法：





阿通



如果您查看源代码，您会发现我库中的方法完全基于for构造。

但是我纯粹是为了给他们提供代码吹牛 .

Benchmark.js. 10 10,000,000 .

: Node.js.



, , .



.



: . -.







Filter



:



const filterCondition = (item: number) => !!(item % 2);




array.filter(filterCondition);


Benchmark

native filter x 3.58 ops/sec ±5.48% (13 runs sampled)




For of



const result = [];

for(let item of array) {
    if(filterCondition(item)) {
        result.push(item)
    }
}


Benchmark

for… of x 3.48 ops/sec ±3.46% (14 runs sampled)

, for of , — .





For



filterSuite.add('for', () => {
    const result = [];

    for(let i = 0; i < array.length; i++) {
        const item = array[i]
        if(filterCondition(item)) {
            result.push(item)
        }
    }
})


Benchmark

for x 6.92 ops/sec ±5.47% (20 runs sampled)

for filter 2 .





Lodash



:



lodash(array).filter(filterCondition).value();


Benchmark

lodash filter x 3.60 ops/sec ±4.13% (13 runs sampled)

Lodash filter.







ursus(array).where(filterCondition).toArray();


Benchmark

ursus where x 6.87 ops/sec ±4.86% (20 runs sampled)

± for, .







Benchmark log 
lodash filter x 3.75 ops/sec ±4.09% (13 runs sampled)
native filter x 3.35 ops/sec ±7.99% (13 runs sampled)
for ... of x 3.38 ops/sec ±3.88% (13 runs sampled)
for x 6.04 ops/sec ±3.54% (20 runs sampled)
optimized for x 5.82 ops/sec ±3.05% (20 runs sampled)
ursus where x 5.83 ops/sec ±4.62% (21 runs sampled)

lodash filter x 3.37 ops/sec ±3.40% (13 runs sampled)
native filter x 3.33 ops/sec ±4.76% (13 runs sampled)
for ... of x 3.86 ops/sec ±9.36% (14 runs sampled)
for x 6.92 ops/sec ±5.47% (20 runs sampled)
optimized for x 6.96 ops/sec ±4.22% (20 runs sampled)
ursus where x 6.71 ops/sec ±4.75% (19 runs sampled)

lodash filter x 3.51 ops/sec ±4.94% (13 runs sampled)
native filter x 3.72 ops/sec ±0.74% (13 runs sampled)
for ... of x 3.54 ops/sec ±2.14% (14 runs sampled)
for x 6.84 ops/sec ±4.60% (20 runs sampled)
optimized for x 6.85 ops/sec ±3.58% (19 runs sampled)
ursus where x 6.46 ops/sec ±11.83% (20 runs sampled)

lodash filter x 3.55 ops/sec ±6.26% (13 runs sampled)
native filter x 3.66 ops/sec ±3.06% (13 runs sampled)
for ... of x 3.41 ops/sec ±4.28% (14 runs sampled)
for x 6.95 ops/sec ±3.85% (20 runs sampled)
optimized for x 6.79 ops/sec ±4.33% (20 runs sampled)
ursus where x 7.17 ops/sec ±3.29% (21 runs sampled)

lodash filter x 3.63 ops/sec ±3.38% (13 runs sampled)
native filter x 3.63 ops/sec ±3.35% (13 runs sampled)
for ... of x 3.44 ops/sec ±3.99% (14 runs sampled)
for x 6.89 ops/sec ±4.53% (20 runs sampled)
optimized for x 6.95 ops/sec ±3.17% (20 runs sampled)
ursus where x 6.87 ops/sec ±4.86% (20 runs sampled)

lodash filter x 3.60 ops/sec ±4.13% (13 runs sampled)
native filter x 3.58 ops/sec ±5.48% (13 runs sampled)
for ... of x 3.48 ops/sec ±3.46% (14 runs sampled)
for x 6.95 ops/sec ±3.73% (20 runs sampled)
optimized for x 6.78 ops/sec ±5.62% (20 runs sampled)
ursus where x 7.17 ops/sec ±3.79% (21 runs sampled)

lodash filter x 3.64 ops/sec ±4.11% (13 runs sampled)
native filter x 3.63 ops/sec ±3.35% (13 runs sampled)
for ... of x 3.55 ops/sec ±2.36% (14 runs sampled)
for x 6.91 ops/sec ±4.51% (20 runs sampled)
optimized for x 6.89 ops/sec ±3.76% (20 runs sampled)
ursus where x 7.03 ops/sec ±4.84% (21 runs sampled)

lodash filter x 3.59 ops/sec ±4.17% (13 runs sampled)
native filter x 3.60 ops/sec ±3.14% (13 runs sampled)
for ... of x 3.45 ops/sec ±4.69% (14 runs sampled)
for x 7.09 ops/sec ±2.65% (20 runs sampled)
optimized for x 6.81 ops/sec ±2.90% (20 runs sampled)
ursus where x 7.15 ops/sec ±2.60% (21 runs sampled)

lodash filter x 3.60 ops/sec ±5.57% (13 runs sampled)
native filter x 3.60 ops/sec ±4.55% (13 runs sampled)
for ... of x 3.39 ops/sec ±7.33% (13 runs sampled)
for x 5.71 ops/sec ±2.74% (19 runs sampled)
optimized for x 5.85 ops/sec ±2.70% (20 runs sampled)
ursus where x 6.10 ops/sec ±2.43% (21 runs sampled)

lodash filter x 3.18 ops/sec ±5.88% (13 runs sampled)
native filter x 3.34 ops/sec ±4.43% (13 runs sampled)
for ... of x 3.89 ops/sec ±6.84% (14 runs sampled)
for x 7.09 ops/sec ±2.79% (21 runs sampled)
optimized for x 6.70 ops/sec ±3.32% (20 runs sampled)
ursus where x 7.07 ops/sec ±4.02% (20 runs sampled)


.
filter 3,57 ops/sec 3,60 ops/sec
for 3,48 ops/sec 3,47 ops/sec
for of 6,91 ops/sec 6,92 ops/sec
lodash 3,58 ops/sec 3,60 ops/sec
ursus 6,88 ops/sec 6,95 ops/sec


:



1) For

2) Ursus

3) Lodash

4) Filter

5) For of



Map



+1



const mapCondition = (item: number) => item + 1;




array.map(mapCondition);


Benchmark

native map x 0.68 ops/sec ±3.60% (6 runs sampled)




For of



const result = [];
for(let item of array) {
    result.push(mapCondition(item))
}


Benchmark

for… of x 2.19 ops/sec ±3.47% (10 runs sampled)

for of 3 .





For



const result = [];
for(let i = 0; i < array.length; i++) {
    result.push(mapCondition(array[i]))
}


Benchmark

for x 3.49 ops/sec ±6.82% (12 runs sampled)

for 5 .





Lodash



lodash(array).map(mapCondition).value();


Benchmark

lodash map x 5.78 ops/sec ±9.03% (18 runs sampled)

Lodash 9 !



, - , .






ursus(array).select(mapCondition).toArray();


Benchmark

ursus select x 3.54 ops/sec ±5.71% (13 runs sampled)

- for, . ¯_(ツ)_/¯





Benchmark log 
lodash map x 6.08 ops/sec ±4.84% (19 runs sampled)
native map x 0.57 ops/sec ±17.60% (6 runs sampled)
for ... of x 1.91 ops/sec ±13.65% (9 runs sampled)
for x 3.51 ops/sec ±5.25% (13 runs sampled)
optimized for x 3.62 ops/sec ±7.49% (13 runs sampled)
ursus select x 3.29 ops/sec ±9.24% (13 runs sampled)

lodash map x 5.59 ops/sec ±10.61% (19 runs sampled)
native map x 0.61 ops/sec ±11.70% (6 runs sampled)
for ... of x 2.30 ops/sec ±2.13% (10 runs sampled)
for x 3.72 ops/sec ±4.39% (13 runs sampled)
optimized for x 3.58 ops/sec ±5.24% (13 runs sampled)
ursus select x 3.58 ops/sec ±5.21% (13 runs sampled)

lodash map x 6.06 ops/sec ±5.23% (19 runs sampled)
native map x 0.68 ops/sec ±3.60% (6 runs sampled)
for ... of x 2.27 ops/sec ±3.49% (10 runs sampled)
for x 3.45 ops/sec ±10.41% (13 runs sampled)
optimized for x 3.59 ops/sec ±4.29% (13 runs sampled)
ursus select x 3.54 ops/sec ±6.08% (12 runs sampled)

lodash map x 5.81 ops/sec ±7.23% (19 runs sampled)
native map x 0.68 ops/sec ±3.63% (6 runs sampled)
for ... of x 2.31 ops/sec ±7.11% (10 runs sampled)
for x 3.62 ops/sec ±4.74% (13 runs sampled)
optimized for x 3.45 ops/sec ±6.67% (13 runs sampled)
ursus select x 3.64 ops/sec ±4.42% (13 runs sampled)

lodash map x 6.03 ops/sec ±5.26% (20 runs sampled)
native map x 0.69 ops/sec ±6.27% (6 runs sampled)
for ... of x 2.12 ops/sec ±8.87% (10 runs sampled)
for x 3.29 ops/sec ±9.33% (13 runs sampled)
optimized for x 3.53 ops/sec ±5.18% (13 runs sampled)
ursus select x 3.66 ops/sec ±4.03% (13 runs sampled)

lodash map x 5.78 ops/sec ±9.03% (18 runs sampled)
native map x 0.65 ops/sec ±6.52% (6 runs sampled)
for ... of x 2.07 ops/sec ±7.41% (10 runs sampled)
for x 3.49 ops/sec ±6.82% (12 runs sampled)
optimized for x 3.50 ops/sec ±5.93% (13 runs sampled)
ursus select x 3.54 ops/sec ±5.71% (13 runs sampled)

lodash map x 5.68 ops/sec ±8.47% (18 runs sampled)
native map x 0.67 ops/sec ±6.40% (6 runs sampled)
for ... of x 2.11 ops/sec ±5.06% (10 runs sampled)
for x 3.52 ops/sec ±5.58% (13 runs sampled)
optimized for x 3.29 ops/sec ±5.51% (13 runs sampled)
ursus select x 3.38 ops/sec ±5.31% (13 runs sampled)

lodash map x 6.37 ops/sec ±3.10% (19 runs sampled)
native map x 0.67 ops/sec ±2.43% (6 runs sampled)
for ... of x 2.19 ops/sec ±3.47% (10 runs sampled)
for x 3.41 ops/sec ±8.13% (13 runs sampled)
optimized for x 3.54 ops/sec ±5.15% (13 runs sampled)
ursus select x 3.53 ops/sec ±6.28% (13 runs sampled)

lodash map x 5.85 ops/sec ±11.04% (19 runs sampled)
native map x 0.66 ops/sec ±4.30% (6 runs sampled)
for ... of x 2.20 ops/sec ±2.97% (10 runs sampled)
for x 3.45 ops/sec ±8.03% (13 runs sampled)
optimized for x 3.48 ops/sec ±5.13% (13 runs sampled)
ursus select x 3.68 ops/sec ±3.33% (13 runs sampled)

lodash map x 5.31 ops/sec ±12.87% (18 runs sampled)
native map x 0.68 ops/sec ±4.26% (6 runs sampled)
for ... of x 2.11 ops/sec ±6.97% (10 runs sampled)
for x 3.35 ops/sec ±6.12% (13 runs sampled)
optimized for x 3.38 ops/sec ±5.55% (13 runs sampled)
ursus select x 3.54 ops/sec ±6.20% (13 runs sampled)


.
map 0.67 ops/sec 0.67 ops/sec
for 2.17 ops/sec 2.16 ops/sec
for of 3.47 ops/sec 3.47 ops/sec
lodash 5.79 ops/sec 5.80 ops/sec
ursus 3.56 ops/sec 3.54 ops/sec


:



1) Lodash

2) Ursus

3) For

4) For of

5) Map



Reduce





const sumCondition = (item1: number, item2: number) => item1 + item2;




array.reduce(sumCondition);


Benchmark

native reduce x 6.09 ops/sec ±9.13% (20 runs sampled)




For of



, . .




For



let result = array[0];
for(let i = 1; i < array.length; i++) {
    result = sumCondition(result, array[i])
}


Benchmark

for x 57.01 ops/sec ±2.53% (59 runs sampled)

For 10 !





Lodash



lodash(array).sum();


Benchmark

lodash sum x 8.30 ops/sec ±7.79% (25 runs sampled)

lodash , , reduce.







ursus(array).sum(sumCondition);


Benchmark

ursus sum x 56.12 ops/sec ±2.38% (58 runs sampled)




Benchmark log 
lodash sum x 8.60 ops/sec ±4.35% (25 runs sampled)
native reduce x 6.69 ops/sec ±3.73% (21 runs sampled)
for x 68.67 ops/sec ±3.41% (70 runs sampled)
optimized for x 70.75 ops/sec ±2.63% (72 runs sampled)
ursus sum x 67.78 ops/sec ±3.12% (70 runs sampled)

lodash sum x 9.00 ops/sec ±3.93% (26 runs sampled)
native reduce x 5.47 ops/sec ±21.31% (19 runs sampled)
for x 56.61 ops/sec ±2.70% (59 runs sampled)
optimized for x 56.85 ops/sec ±2.27% (59 runs sampled)
ursus sum x 56.08 ops/sec ±2.40% (59 runs sampled)

lodash sum x 8.69 ops/sec ±3.36% (26 runs sampled)
native reduce x 6.09 ops/sec ±9.13% (20 runs sampled)
for x 57.01 ops/sec ±2.53% (59 runs sampled)
optimized for x 57.38 ops/sec ±2.64% (60 runs sampled)
ursus sum x 56.12 ops/sec ±2.38% (58 runs sampled)

lodash sum x 8.68 ops/sec ±4.11% (26 runs sampled)
native reduce x 6.06 ops/sec ±9.39% (19 runs sampled)
for x 69.97 ops/sec ±2.82% (71 runs sampled)
optimized for x 66.55 ops/sec ±4.16% (68 runs sampled)
ursus sum x 69.29 ops/sec ±2.73% (71 runs sampled)

lodash sum x 7.86 ops/sec ±8.39% (24 runs sampled)
native reduce x 6.35 ops/sec ±4.79% (20 runs sampled)
for x 55.91 ops/sec ±5.01% (58 runs sampled)
optimized for x 56.41 ops/sec ±2.70% (59 runs sampled)
ursus sum x 57.11 ops/sec ±2.16% (58 runs sampled)

lodash sum x 8.11 ops/sec ±4.72% (24 runs sampled)
native reduce x 5.97 ops/sec ±7.80% (20 runs sampled)
for x 56.43 ops/sec ±3.62% (59 runs sampled)
optimized for x 56.87 ops/sec ±3.75% (59 runs sampled)
ursus sum x 55.37 ops/sec ±3.60% (58 runs sampled)

lodash sum x 8.52 ops/sec ±6.70% (25 runs sampled)
native reduce x 6.12 ops/sec ±7.39% (20 runs sampled)
for x 57.96 ops/sec ±3.50% (58 runs sampled)
optimized for x 55.19 ops/sec ±5.32% (59 runs sampled)
ursus sum x 56.75 ops/sec ±3.33% (58 runs sampled)

lodash sum x 8.00 ops/sec ±8.94% (25 runs sampled)
native reduce x 5.75 ops/sec ±6.95% (19 runs sampled)
for x 56.78 ops/sec ±4.21% (57 runs sampled)
optimized for x 56.89 ops/sec ±2.32% (60 runs sampled)
ursus sum x 54.61 ops/sec ±7.04% (57 runs sampled)

lodash sum x 8.11 ops/sec ±8.83% (24 runs sampled)
native reduce x 5.97 ops/sec ±7.84% (19 runs sampled)
for x 57.32 ops/sec ±4.17% (59 runs sampled)
optimized for x 55.97 ops/sec ±4.18% (59 runs sampled)
ursus sum x 55.76 ops/sec ±3.90% (58 runs sampled)

lodash sum x 8.30 ops/sec ±7.79% (25 runs sampled)
native reduce x 6.31 ops/sec ±5.42% (20 runs sampled)
for x 55.45 ops/sec ±5.56% (58 runs sampled)
optimized for x 57.54 ops/sec ±3.52% (59 runs sampled)
ursus sum x 55.22 ops/sec ±4.34% (57 runs sampled)


.
reduce 6.09 ops/sec 6.08 ops/sec
for 57.02 ops/sec 56.90 ops/sec
lodash 8.39 ops/sec 8.41 ops/sec
ursus 56.20 ops/sec 56.10 ops/sec


:



1) For

2) Ursus

3) Lodash

4) Reduce





, 10 , for .



-, lodash 50k .



25k for.

在我看来，总的来说，这是一个相当奇怪的实验，但是现在有了关于大量元素的此类信息，我可能至少会检查本机实现的性能。



感谢您的关注！



More articles:


All Articles