reduce calls to Node.prototype.toString()

[romainmenke]

see : #1847

[ai]

What are https://github.com/postcss/benchmark changes?

[romainmenke]

Apparently there is something not happy on M1/M2.
Everything is consistently faster on my machine than on an older Intel, but these benchmarks run a lot slower than expected.
So the exact times are way of, but the delta seems to a good indication.

Running suite Preprocessors 
   LibSass x 1.62 ops/sec ±0.85% (13 runs sampled)
   LibSass sync x 1.63 ops/sec ±1.12% (9 runs sampled)
   Dart Sass x 0.53 ops/sec ±6.31% (7 runs sampled)
   Dart Sass sync x 0.83 ops/sec ±3.14% (7 runs sampled)
   PostCSS x 2.50 ops/sec ±4.95% (17 runs sampled)
   PostCSS sync x 2.39 ops/sec ±3.67% (16 runs sampled)
   Next PostCSS x 2.43 ops/sec ±10.33% (16 runs sampled)
   Less x 1.21 ops/sec ±5.01% (11 runs sampled)
   Next PostCSS sync x 2.21 ops/sec ±5.47% (15 runs sampled)
Fastest tests are PostCSS,Next PostCSS at 1.03x faster than Next PostCSS

PostCSS:           400 ms
Next PostCSS:      412 ms  (1.0 times slower)
PostCSS sync:      418 ms  (1.0 times slower)
Next PostCSS sync: 453 ms  (1.1 times slower)
LibSass sync:      615 ms  (1.5 times slower)
LibSass:           615 ms  (1.5 times slower)
Less:              824 ms  (2.1 times slower)
Dart Sass sync:    1207 ms (3.0 times slower)
Dart Sass:         1900 ms (4.7 times slower)


Running suite Parsers
   Rework x 2.64 ops/sec ±7.73% (11 runs sampled)
   Next PostCSS x 3.67 ops/sec ±6.53% (22 runs sampled)
   PostCSS x 3.40 ops/sec ±7.82% (21 runs sampled)
   PostCSS Full x 1.30 ops/sec ±12.56% (10 runs sampled)
   CSSOM x 0.38 ops/sec ±0.87% (5 runs sampled)
   Mensch x 0.06 ops/sec ±6.67% (5 runs sampled)
   Gonzales x 0.23 ops/sec ±15.73% (5 runs sampled)
   CSSTree x 3.54 ops/sec ±9.83% (14 runs sampled)
   ParserLib x 0.83 ops/sec ±5.68% (7 runs sampled)
   Stylecow x 1.81 ops/sec ±3.01% (9 runs sampled)
   Stylis x 7.75 ops/sec ±16.33% (26 runs sampled)
Fastest test is Stylis at 2.2x faster than CSSTree and Next PostCSS

Stylis:       129 ms   (2.3 times faster)
Next PostCSS: 272 ms   (1.1 times faster)
CSSTree:      282 ms   (1.0 times faster)
PostCSS:      294 ms
Rework:       380 ms   (1.3 times slower)
Stylecow:     553 ms   (1.9 times slower)
PostCSS Full: 772 ms   (2.6 times slower)
ParserLib:    1207 ms  (4.1 times slower)
CSSOM:        2658 ms  (9.0 times slower)
Gonzales:     4269 ms  (14.5 times slower)
Mensch:       15479 ms (52.6 times slower)


Running suite Prefixers 
   Autoprefixer dev x 1.17 ops/sec ±9.44% (11 runs sampled)
   Stylis x 6.71 ops/sec ±13.86% (22 runs sampled)
   Autoprefixer x 1.21 ops/sec ±6.53% (11 runs sampled)
   Stylecow x 0.02 ops/sec ±2.47% (5 runs sampled)
Fastest test is Stylis at 5.7x faster than Autoprefixer dev and Autoprefixer

Stylis:           149 ms   (5.6 times faster)
Autoprefixer:     829 ms
Autoprefixer dev: 857 ms   (1.0 times slower)
Stylecow:         66070 ms (79.7 times slower)


Running suite Tokenizers
   PostCSS x 15.37 ops/sec ±11.82% (77 runs sampled)
   Next PostCSS x 15.25 ops/sec ±13.91% (75 runs sampled)
Fastest tests are PostCSS,Next PostCSS

PostCSS:      65 ms
Next PostCSS: 66 ms (1.0 times slower)


Running suite Linters
   PostCSS x 0.16 ops/sec ±5.04% (5 runs sampled)
   Next PostCSS x 0.17 ops/sec ±9.38% (5 runs sampled)
Fastest tests are Next PostCSS,PostCSS

Next PostCSS: 5980 ms (1.1 times faster)
PostCSS:      6332 ms

A run with only the linters test after letting my machine cool a bit :

   PostCSS x 0.20 ops/sec ±2.80% (5 runs sampled)
   Next PostCSS x 0.22 ops/sec ±1.34% (6 runs sampled)
Fastest test is Next PostCSS at 1.10x faster than PostCSS

Next PostCSS: 4649 ms (1.1 times faster)
PostCSS:      5099 ms

[ai]

Thanks. Released at 8.4.25.

[romainmenke]

Thank you @ai 🙇