Clever Geek Handbook

Native ne veut pas dire rapide. Dépassement de la carte, filtrer et réduire sur de grands tableaux



Il y a quelques jours, j'ai présenté un article LINQ en JavaScript pour les plus petits . Mais ma bibliothèque était bien inférieure en performances aux méthodes natives et à Lodash . En général, nous allons maintenant changer la situation.



Je dirai tout de suite: l'article ne contiendra aucune révélation, nous ne construirons pas d'algorithmes fous, etc. Nous comparerons simplement les performances de différentes constructions de langage.



, , . callback, , map, filter reduce , .









:





!



, , for.

, .

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.

À mon avis, en général, cela s'est avéré être une expérience plutôt curieuse, mais maintenant disposant de telles informations, sur un grand nombre d'éléments, je vais probablement, au moins, vérifier les performances des implémentations natives.



Merci de votre attention!



More articles:


All Articles