Initial Implementation

[Uzlopak]

@mcollina
@kibertoad

This is the promised faster implementation.

I tried to extract the function processMatch but it is slightly slower. Or I had not the right idea.

identity;q=1 and gzip,identity x 7,380,422 ops/sec ±0.99% (88 runs sampled)
gzip;q=1, identity;q=0.5 and gzip,deflate x 11,526,066 ops/sec ±0.89% (89 runs sampled)
deflate;q=0.5,identity; q=0.5 and gzip,deflate x 2,521,171 ops/sec ±0.90% (87 runs sampled)
deflate;q=0.5, gzip;q=0.5 and gzip,deflate x 2,846,303 ops/sec ±0.88% (88 runs sampled)
deflate;q=0.5, gzip;q=0.5 and deflate,gzip x 2,920,872 ops/sec ±0.97% (89 runs sampled)
* and gzip,deflate x 734,274,924 ops/sec ±1.77% (82 runs sampled)
deflate;q=1.0, * and gzip x 5,058,241 ops/sec ±0.87% (89 runs sampled)
test,br and br x 10,555,576 ops/sec ±1.07% (85 runs sampled)
gzip;q=0 and  x 88,404,780 ops/sec ±0.85% (94 runs sampled)
gzip;q=0 and gzip,identity x 12,899,899 ops/sec ±1.21% (87 runs sampled)
white rabbit and gzip,identity x 6,975,091 ops/sec ±0.97% (90 runs sampled)
undefined and gzip,identity x 85,799,975 ops/sec ±0.69% (92 runs sampled)
compress;q=0.5, gzip;q=1.0 and gzip,compress x 3,131,648 ops/sec ±1.04% (89 runs sampled)
gzip;q=1.0, compress;q=0.5 and compress,gzip x 3,237,159 ops/sec ±1.15% (88 runs sampled)
compress;q=0.5, gzip;q=1.0 and compress x 2,976,764 ops/sec ±0.66% (93 runs sampled)
gzip, deflate, br and br,gzip,deflate x 4,861,383 ops/sec ±0.65% (89 runs sampled)
* and br,gzip,deflate x 399,423,765 ops/sec ±19.45% (59 runs sampled)
*;q=0, identity;q=1 and gzip,identity x 5,312,961 ops/sec ±0.90% (78 runs sampled)
identity;q=0 and identity x 7,171,469 ops/sec ±1.33% (87 runs sampled)
gzip, compress;q=0 and compress,gzip x 4,881,432 ops/sec ±0.87% (89 runs sampled)
gzip;q=0.8, deflate and gzip,deflate x 3,711,530 ops/sec ±1.81% (89 runs sampled)
gzip;q=0.8, identity;q=0.5, *;q=0.3 and deflate,gzip,br x 2,252,691 ops/sec ±0.77% (94 runs sampled)

[Uzlopak]

We could theoretically write the negotiate function like a function where you set the supportedValues and then it has a fixed supportedValues array. Then we could apply caching, which could do some performance gain. But I am kind of unsure if this is actually a smart idea or not, as encodings could be alot variations.

[Uzlopak]

@mcollina
Any feedback?

[mcollina]

Does this passes the same tests of the more popular module?

[Uzlopak]

Yes, I used the same tests + some tests from encoding package from dougwilson.

[mcollina]

Code looks good.
How is the performance against the original?

Could you fill in the README too?

[Uzlopak]

I implemented also a Negotiator-Class, which seems to be slightly faster. You can also pass a Cache-Store, like mnemonist LRU-Cache, which makes it then incredible fast of course.

Here the latest benchmarks:

node benchmarks/negotiate.js 
identity;q=1 and gzip,identity x 8,324,480 ops/sec ±1.12% (88 runs sampled)
gzip;q=1, identity;q=0.5 and gzip,deflate x 12,881,974 ops/sec ±1.07% (89 runs sampled)
deflate;q=0.5,identity; q=0.5 and gzip,deflate x 2,640,400 ops/sec ±1.15% (90 runs sampled)
deflate;q=0.5, gzip;q=0.5 and gzip,deflate x 2,976,596 ops/sec ±1.10% (91 runs sampled)
deflate;q=0.5, gzip;q=0.5 and deflate,gzip x 3,030,612 ops/sec ±1.26% (86 runs sampled)
* and gzip,deflate x 721,577,308 ops/sec ±1.88% (84 runs sampled)
deflate;q=1.0, * and gzip x 5,104,490 ops/sec ±1.02% (87 runs sampled)
test,br and br x 9,533,048 ops/sec ±0.86% (88 runs sampled)
gzip;q=0 and  x 75,220,077 ops/sec ±0.73% (92 runs sampled)
gzip;q=0 and gzip,identity x 13,089,768 ops/sec ±1.08% (88 runs sampled)
white rabbit and gzip,identity x 7,194,967 ops/sec ±1.17% (85 runs sampled)
undefined and gzip,identity x 75,556,391 ops/sec ±0.85% (93 runs sampled)
compress;q=0.5, gzip;q=1.0 and gzip,compress x 2,930,521 ops/sec ±1.76% (90 runs sampled)
gzip;q=1.0, compress;q=0.5 and compress,gzip x 3,070,647 ops/sec ±0.83% (87 runs sampled)
compress;q=0.5, gzip;q=1.0 and compress x 2,932,494 ops/sec ±1.02% (88 runs sampled)
gzip, deflate, br and br,gzip,deflate x 4,868,715 ops/sec ±1.37% (90 runs sampled)
* and br,gzip,deflate x 120,858,353 ops/sec ±3.32% (81 runs sampled)
*;q=0, identity;q=1 and gzip,identity x 5,266,389 ops/sec ±1.17% (85 runs sampled)
identity;q=0 and identity x 7,097,012 ops/sec ±1.14% (85 runs sampled)
gzip, compress;q=0 and compress,gzip x 4,879,960 ops/sec ±0.65% (90 runs sampled)
gzip;q=0.8, deflate and gzip,deflate x 3,872,093 ops/sec ±0.69% (93 runs sampled)
gzip;q=0.8, identity;q=0.5, *;q=0.3 and deflate,gzip,br x 2,275,268 ops/sec ±1.96% (89 runs sampled)

The benchmarks of Negotiator-class

node benchmarks/Negotiator.js 
identity;q=1 and gzip,identity x 9,056,308 ops/sec ±1.10% (91 runs sampled)
gzip;q=1, identity;q=0.5 and gzip,deflate x 13,893,625 ops/sec ±1.08% (93 runs sampled)
deflate;q=0.5,identity; q=0.5 and gzip,deflate x 2,819,782 ops/sec ±1.21% (92 runs sampled)
deflate;q=0.5, gzip;q=0.5 and gzip,deflate x 3,354,313 ops/sec ±0.68% (95 runs sampled)
deflate;q=0.5, gzip;q=0.5 and deflate,gzip x 3,302,053 ops/sec ±1.42% (93 runs sampled)
* and gzip,deflate x 701,541,302 ops/sec ±0.43% (98 runs sampled)
deflate;q=1.0, * and gzip x 5,385,857 ops/sec ±0.83% (92 runs sampled)
test,br and br x 10,469,662 ops/sec ±1.94% (94 runs sampled)
gzip;q=0 and  x 82,215,189 ops/sec ±0.71% (94 runs sampled)
gzip;q=0 and gzip,identity x 13,703,082 ops/sec ±0.97% (89 runs sampled)
white rabbit and gzip,identity x 7,522,050 ops/sec ±0.99% (93 runs sampled)
undefined and gzip,identity x 82,851,862 ops/sec ±0.70% (95 runs sampled)
compress;q=0.5, gzip;q=1.0 and gzip,compress x 3,173,137 ops/sec ±0.75% (94 runs sampled)
gzip;q=1.0, compress;q=0.5 and compress,gzip x 3,285,739 ops/sec ±0.88% (92 runs sampled)
compress;q=0.5, gzip;q=1.0 and compress x 3,239,147 ops/sec ±0.69% (95 runs sampled)
gzip, deflate, br and br,gzip,deflate x 5,111,553 ops/sec ±1.34% (96 runs sampled)
* and br,gzip,deflate x 665,710,687 ops/sec ±1.45% (93 runs sampled)
*;q=0, identity;q=1 and gzip,identity x 5,685,409 ops/sec ±1.30% (93 runs sampled)
identity;q=0 and identity x 8,424,843 ops/sec ±0.86% (92 runs sampled)
gzip, compress;q=0 and compress,gzip x 5,223,453 ops/sec ±0.67% (94 runs sampled)
gzip;q=0.8, deflate and gzip,deflate x 4,018,832 ops/sec ±1.23% (94 runs sampled)
gzip;q=0.8, identity;q=0.5, *;q=0.3 and deflate,gzip,br x 2,524,941 ops/sec ±1.32% (93 runs sampled)

The benchmarks of Negotiator-class with LRUCache from mnemonist with a size of 1000

identity;q=1 and gzip,identity x 1,276,276,316 ops/sec ±0.11% (100 runs sampled)
gzip;q=1, identity;q=0.5 and gzip,deflate x 83,448,701 ops/sec ±0.12% (100 runs sampled)
deflate;q=0.5,identity; q=0.5 and gzip,deflate x 83,267,662 ops/sec ±0.16% (100 runs sampled)
deflate;q=0.5, gzip;q=0.5 and gzip,deflate x 82,609,401 ops/sec ±0.68% (94 runs sampled)
deflate;q=0.5, gzip;q=0.5 and deflate,gzip x 83,379,855 ops/sec ±0.15% (96 runs sampled)
* and gzip,deflate x 81,986,292 ops/sec ±1.22% (97 runs sampled)
deflate;q=1.0, * and gzip x 82,118,327 ops/sec ±0.28% (99 runs sampled)
test,br and br x 69,695,704 ops/sec ±4.09% (84 runs sampled)
gzip;q=0 and  x 56,319,830 ops/sec ±1.09% (94 runs sampled)
gzip;q=0 and gzip,identity x 56,102,778 ops/sec ±1.16% (92 runs sampled)
white rabbit and gzip,identity x 82,132,157 ops/sec ±0.39% (91 runs sampled)
undefined and gzip,identity x 159,007,665 ops/sec ±0.14% (98 runs sampled)
compress;q=0.5, gzip;q=1.0 and gzip,compress x 79,755,680 ops/sec ±0.38% (96 runs sampled)
gzip;q=1.0, compress;q=0.5 and compress,gzip x 65,173,032 ops/sec ±4.15% (80 runs sampled)
compress;q=0.5, gzip;q=1.0 and compress x 54,334,682 ops/sec ±0.93% (89 runs sampled)
gzip, deflate, br and br,gzip,deflate x 78,978,702 ops/sec ±1.38% (95 runs sampled)
* and br,gzip,deflate x 71,324,729 ops/sec ±4.07% (83 runs sampled)
*;q=0, identity;q=1 and gzip,identity x 53,901,367 ops/sec ±1.06% (93 runs sampled)
identity;q=0 and identity x 75,140,280 ops/sec ±3.05% (93 runs sampled)
gzip, compress;q=0 and compress,gzip x 54,929,535 ops/sec ±0.94% (94 runs sampled)
gzip;q=0.8, deflate and gzip,deflate x 53,847,423 ops/sec ±1.40% (87 runs sampled)
gzip;q=0.8, identity;q=0.5, *;q=0.3 and deflate,gzip,br x 64,022,160 ops/sec ±3.90% (78 runs sampled)

The same benchmarks against the encoding-negotiator-package

identity;q=1 and gzip,identity x 2,471,479 ops/sec ±1.23% (87 runs sampled)
gzip;q=1, identity;q=0.5 and gzip,deflate x 895,704 ops/sec ±0.94% (90 runs sampled)
deflate;q=0.5,identity; q=0.5 and gzip,deflate x 944,281 ops/sec ±1.75% (89 runs sampled)
deflate;q=0.5, gzip;q=0.5 and gzip,deflate x 788,710 ops/sec ±1.60% (87 runs sampled)
deflate;q=0.5, gzip;q=0.5 and deflate,gzip x 777,065 ops/sec ±2.07% (88 runs sampled)
* and gzip,deflate x 3,098,768 ops/sec ±1.07% (88 runs sampled)
deflate;q=1.0, * and gzip x 1,318,978 ops/sec ±0.99% (86 runs sampled)
test,br and br x 1,696,693 ops/sec ±0.89% (91 runs sampled)
gzip;q=0 and  x 2,896,883 ops/sec ±0.96% (90 runs sampled)
gzip;q=0 and gzip,identity x 2,466,758 ops/sec ±1.07% (92 runs sampled)
white rabbit and gzip,identity x 2,998,611 ops/sec ±1.18% (89 runs sampled)
undefined and gzip,identity x 110,211,761 ops/sec ±1.13% (88 runs sampled)
compress;q=0.5, gzip;q=1.0 and gzip,compress x 804,550 ops/sec ±1.02% (90 runs sampled)
gzip;q=1.0, compress;q=0.5 and compress,gzip x 848,624 ops/sec ±1.02% (87 runs sampled)
compress;q=0.5, gzip;q=1.0 and compress x 804,496 ops/sec ±1.04% (90 runs sampled)
gzip, deflate, br and br,gzip,deflate x 852,218 ops/sec ±0.97% (87 runs sampled)
* and br,gzip,deflate x 2,893,678 ops/sec ±0.89% (85 runs sampled)
*;q=0, identity;q=1 and gzip,identity x 886,444 ops/sec ±1.10% (88 runs sampled)
identity;q=0 and identity x 2,659,673 ops/sec ±0.90% (91 runs sampled)
gzip, compress;q=0 and compress,gzip x 1,083,041 ops/sec ±1.43% (88 runs sampled)
gzip;q=0.8, deflate and gzip,deflate x 1,061,551 ops/sec ±1.15% (92 runs sampled)
gzip;q=0.8, identity;q=0.5, *;q=0.3 and deflate,gzip,br x 578,058 ops/sec ±0.79% (90 runs sampled)

The same benchmarks against the negotiator-package, which is used by express:

identity;q=1 and gzip,identity x 1,196,783 ops/sec ±0.54% (95 runs sampled)
gzip;q=1, identity;q=0.5 and gzip,deflate x 678,071 ops/sec ±0.47% (96 runs sampled)
deflate;q=0.5,identity; q=0.5 and gzip,deflate x 645,024 ops/sec ±0.60% (92 runs sampled)
deflate;q=0.5, gzip;q=0.5 and gzip,deflate x 560,488 ops/sec ±0.54% (97 runs sampled)
deflate;q=0.5, gzip;q=0.5 and deflate,gzip x 574,648 ops/sec ±0.96% (91 runs sampled)
* and gzip,deflate x 1,771,614 ops/sec ±0.55% (97 runs sampled)
deflate;q=1.0, * and gzip x 983,923 ops/sec ±0.98% (94 runs sampled)
test,br and br x 1,894,286 ops/sec ±0.78% (91 runs sampled)
gzip;q=0 and  x 1,345,700 ops/sec ±2.14% (95 runs sampled)
gzip;q=0 and gzip,identity x 1,113,240 ops/sec ±0.55% (90 runs sampled)
white rabbit and gzip,identity x 1,866,715 ops/sec ±0.76% (91 runs sampled)
undefined and gzip,identity x 2,074,228 ops/sec ±1.18% (97 runs sampled)
compress;q=0.5, gzip;q=1.0 and gzip,compress x 497,194 ops/sec ±0.89% (95 runs sampled)
gzip;q=1.0, compress;q=0.5 and compress,gzip x 492,783 ops/sec ±0.53% (93 runs sampled)
compress;q=0.5, gzip;q=1.0 and compress x 571,516 ops/sec ±0.76% (93 runs sampled)
gzip, deflate, br and br,gzip,deflate x 773,635 ops/sec ±0.85% (86 runs sampled)
* and br,gzip,deflate x 1,508,239 ops/sec ±1.17% (88 runs sampled)
*;q=0, identity;q=1 and gzip,identity x 710,647 ops/sec ±1.76% (97 runs sampled)
identity;q=0 and identity x 1,185,037 ops/sec ±1.42% (92 runs sampled)
gzip, compress;q=0 and compress,gzip x 782,730 ops/sec ±0.48% (89 runs sampled)
gzip;q=0.8, deflate and gzip,deflate x 647,693 ops/sec ±0.43% (88 runs sampled)
gzip;q=0.8, identity;q=0.5, *;q=0.3 and deflate,gzip,br x 414,305 ops/sec ±0.61% (91 runs sampled)

[Uzlopak]

I guess we can also use this for fastify-accepts

[mcollina]

lgtm