perf: faster escape html & unescape html #421

SukkaW · 2025-05-31T11:03:50Z

check list

Add test cases for the changes.
Passed the CI test.

Description

Recently, I have created the fastest HTMl escaping with plain JavaScript: fast-escape-html. It is even faster than the Rust counterpart with real-world HTML.

hexo-util can't use fast-escape-url directly since hexo-utils needs to escape more characters (to work with swig/mustache.js/pug template languages), and also needs to avoid double escape.

I have managed to modify fast-escape-url for hexo-util, and it still results in significant performance improvements. I have also slightly modified unescapeHTML to use some performance techniques from fast-escape-url as well.

The escapeTML benchmark results are as follows.

TL; DR: due to extra checks and look-ahead, the new escapeHTML will never beat my fast-escape-url, but it is 145% faster than we currently have, and also beats lodash.escape and escape-goat (that don't even check double escape, and they escape fewer symbols).

clk: ~3.11 GHz
cpu: Apple M2 Max
runtime: node 22.15.1 (arm64-darwin)

benchmark                   avg (min … max) p75 / p99    (min … top 1%)
------------------------------------------- -------------------------------
• skk.moe
------------------------------------------- -------------------------------
hexo-util (old)              712.23 µs/iter 707.13 µs  ▃█
                      (639.04 µs … 2.73 ms)   1.15 ms  ██▃
                    (  1.09 mb …   1.24 mb)   1.19 mb ▆███▅▂▁▁▁▁▁▁▁▁▂▁▂▁▁▁▁
                  4.79 ipc (  2.19% stalls)  98.94% L1 data cache
          2.43M cycles  11.63M instructions  35.83% retired LD/ST (  4.17M)

hexo-util (new)              303.18 µs/iter 305.17 µs ▂█▃
                    (278.54 µs … 563.63 µs) 502.21 µs ███▂
                    (396.88 kb … 872.90 kb) 568.87 kb ████▅▂▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁
                  4.45 ipc (  1.18% stalls)  98.66% L1 data cache
          1.05M cycles   4.65M instructions  19.20% retired LD/ST (893.48k)

fast-escape-html             226.87 µs/iter 227.71 µs ▄█▃
                      (208.75 µs … 1.03 ms) 375.92 µs ███
                    (484.91 kb … 532.95 kb) 484.95 kb ████▅▂▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁
                  4.61 ipc (  0.74% stalls)  99.53% L1 data cache
        777.10k cycles   3.58M instructions  21.25% retired LD/ST (761.79k)

escape-html                  228.08 µs/iter 230.00 µs ▃█
                    (211.75 µs … 537.58 µs) 366.29 µs ███
                    ( 75.42 kb … 769.93 kb) 484.94 kb ████▅▂▂▁▁▁▁▁▁▁▁▁▁▁▁▁▁
                  4.57 ipc (  0.77% stalls)  99.45% L1 data cache
        789.15k cycles   3.60M instructions  21.07% retired LD/ST (759.37k)

@napi-rs/escape              249.93 µs/iter 251.88 µs  █
                      (226.96 µs … 1.08 ms) 566.04 µs ▃█
                    (207.75 kb … 207.77 kb) 207.77 kb ███▂▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁
                  4.50 ipc ( 17.16% stalls)  83.34% L1 data cache
        863.18k cycles   3.88M instructions  23.64% retired LD/ST (917.79k)

html-escaper                 382.64 µs/iter 380.04 µs  █
                      (347.13 µs … 1.68 ms) 842.04 µs ▄█
                    (540.52 kb … 576.24 kb) 574.07 kb ██▇▂▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁
                  4.21 ipc (  2.96% stalls)  97.82% L1 data cache
          1.31M cycles   5.51M instructions  34.04% retired LD/ST (  1.88M)

lodash.escape                391.07 µs/iter 391.00 µs  █
                      (353.50 µs … 1.16 ms) 862.46 µs  █
                    (540.52 kb … 576.36 kb) 574.15 kb ███▂▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁
                  4.18 ipc (  2.89% stalls)  97.87% L1 data cache
          1.35M cycles   5.64M instructions  34.12% retired LD/ST (  1.92M)

escape-goat                  513.57 µs/iter 503.92 µs  █
                      (452.79 µs … 1.38 ms)   1.12 ms  █
                    (  1.22 mb …   1.22 mb)   1.22 mb ███▂▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁
                  3.53 ipc (  2.86% stalls)  97.52% L1 data cache
          1.76M cycles   6.21M instructions  30.85% retired LD/ST (  1.92M)

• github.com (incognito)
------------------------------------------- -------------------------------
hexo-util (old)                2.00 ms/iter   1.94 ms  █▅
                        (1.77 ms … 4.01 ms)   3.04 ms  ██▂
                    (  2.90 mb …   2.97 mb)   2.90 mb ▇███▂▁▁▁▁▁▁▁▁▁▁▁▂▂▄▂▂
                  4.59 ipc (  1.96% stalls)  99.02% L1 data cache
          6.74M cycles  30.95M instructions  35.47% retired LD/ST ( 10.98M)

hexo-util (new)              981.66 µs/iter   1.00 ms  ▄█▇▆
                      (915.50 µs … 1.37 ms)   1.20 ms ▂████▅▂
                    (  1.20 mb …   1.49 mb)   1.20 mb ████████▇▄▂▃▄▄▂▃▁▂▁▁▁
                  4.30 ipc (  0.91% stalls)  98.74% L1 data cache
          3.36M cycles  14.41M instructions  17.00% retired LD/ST (  2.45M)

fast-escape-html             762.76 µs/iter 765.58 µs   █▅
                      (708.46 µs … 1.06 ms)   1.01 ms  ███▆
                    (  1.19 mb …   1.19 mb)   1.19 mb █████▆▂▂▁▁▁▁▁▁▂▂▃▂▁▂▁
                  4.58 ipc (  0.89% stalls)  99.17% L1 data cache
          2.63M cycles  12.06M instructions  20.31% retired LD/ST (  2.45M)

escape-html                  771.41 µs/iter 772.92 µs  ▅█▅
                      (717.83 µs … 2.70 ms)   1.01 ms ▄███▄
                    (  1.19 mb …   1.19 mb)   1.19 mb █████▇▃▂▁▁▁▁▁▁▂▂▂▂▂▂▁
                  4.55 ipc (  0.87% stalls)  99.20% L1 data cache
          2.65M cycles  12.07M instructions  20.17% retired LD/ST (  2.44M)

@napi-rs/escape              886.43 µs/iter 903.08 µs    ▃█▇
                      (799.21 µs … 1.40 ms)   1.16 ms   ▇███▇▂
                    (697.81 kb … 697.81 kb) 697.81 kb ▂▆██████▄▄▃▂▁▂▁▁▁▁▁▁▁
                  4.64 ipc ( 18.92% stalls)  82.14% L1 data cache
          3.01M cycles  13.97M instructions  23.41% retired LD/ST (  3.27M)

html-escaper                   1.13 ms/iter   1.14 ms  ▆▇█▄
                        (1.04 ms … 1.64 ms)   1.52 ms  ████
                    (  1.37 mb …   1.43 mb)   1.39 mb ▇█████▃▂▂▂▃▂▂▂▁▂▂▂▂▂▁
                  3.98 ipc (  2.54% stalls)  98.00% L1 data cache
          3.88M cycles  15.44M instructions  33.38% retired LD/ST (  5.15M)

lodash.escape                  1.15 ms/iter   1.15 ms   ▂█
                        (1.04 ms … 1.88 ms)   1.57 ms  ███▅
                    (  1.37 mb …   1.43 mb)   1.39 mb ▃████▇▃▂▁▁▁▁▁▁▁▂▂▂▂▃▁
                  3.98 ipc (  2.49% stalls)  98.05% L1 data cache
          3.94M cycles  15.66M instructions  33.62% retired LD/ST (  5.26M)

escape-goat                    1.35 ms/iter   1.33 ms   █
                        (1.17 ms … 2.15 ms)   2.00 ms  ▅██
                    (  3.82 mb …   3.82 mb)   3.82 mb ▂████▂▁▁▁▁▂▂▂▂▂▂▃▂▂▂▁
                  3.51 ipc (  3.34% stalls)  97.22% L1 data cache
          4.64M cycles  16.31M instructions  30.47% retired LD/ST (  4.97M)

• stackoverflow.com (incognito)
------------------------------------------- -------------------------------
hexo-util (old)                2.12 ms/iter   2.12 ms   █▆
                        (1.93 ms … 3.43 ms)   2.96 ms  ▅██▂
                    (  3.35 mb …   3.46 mb)   3.46 mb ▅████▃▁▁▁▁▂▃▃▂▂▂▁▁▁▁▁
                  4.45 ipc (  2.45% stalls)  98.75% L1 data cache
          7.22M cycles  32.18M instructions  35.53% retired LD/ST ( 11.43M)

hexo-util (new)              849.81 µs/iter 843.25 µs  ▂█
                      (783.04 µs … 1.26 ms)   1.19 ms  ██▆
                    (  1.00 mb …   1.44 mb)   1.39 mb ████▆▂▁▁▁▁▁▁▁▁▁▂▂▃▂▁▁
                  3.85 ipc (  0.65% stalls)  99.44% L1 data cache
          2.93M cycles  11.31M instructions  19.18% retired LD/ST (  2.17M)

fast-escape-html             668.95 µs/iter 664.88 µs  ▇█
                      (611.96 µs … 1.29 ms) 975.38 µs ▃███
                    (  1.18 mb …   1.18 mb)   1.18 mb ████▆▂▁▁▁▁▁▁▁▁▁▂▂▂▂▂▁
                  4.43 ipc (  0.78% stalls)  99.43% L1 data cache
          2.29M cycles  10.16M instructions  21.21% retired LD/ST (  2.16M)

escape-html                  683.08 µs/iter 673.71 µs  █▆
                      (622.29 µs … 2.98 ms)   1.00 ms ▃██▅
                    (  1.18 mb …   1.18 mb)   1.18 mb ████▅▂▁▁▁▁▁▁▁▁▂▂▂▂▂▁▁
                  4.35 ipc (  0.73% stalls)  99.48% L1 data cache
          2.34M cycles  10.18M instructions  21.03% retired LD/ST (  2.14M)

@napi-rs/escape              749.64 µs/iter 766.83 µs       ▇█
                      (666.42 µs … 1.36 ms) 903.63 µs     ▄████▇▃
                    (586.23 kb … 586.23 kb) 586.23 kb ▂▄▅████████▅▃▂▁▁▁▁▁▁▁
                  4.39 ipc ( 17.59% stalls)  82.46% L1 data cache
          2.53M cycles  11.11M instructions  23.38% retired LD/ST (  2.60M)

html-escaper                   1.11 ms/iter   1.12 ms   ▇█
                        (1.01 ms … 1.59 ms)   1.51 ms  ████
                    (  1.26 mb …   1.32 mb)   1.28 mb ▃█████▄▂▂▃▃▂▂▂▂▂▁▂▂▁▁
                  3.83 ipc (  2.44% stalls)  98.05% L1 data cache
          3.79M cycles  14.50M instructions  33.52% retired LD/ST (  4.86M)

lodash.escape                  1.12 ms/iter   1.12 ms  ▃█▃
                        (1.02 ms … 1.87 ms)   1.62 ms  ███▃
                    (  1.26 mb …   1.32 mb)   1.28 mb ▃████▄▂▁▁▁▁▁▁▂▂▂▂▁▁▁▁
                  3.85 ipc (  2.39% stalls)  98.12% L1 data cache
          3.83M cycles  14.74M instructions  33.74% retired LD/ST (  4.97M)

escape-goat                    1.30 ms/iter   1.26 ms  ▅█
                        (1.11 ms … 2.31 ms)   2.10 ms  ██▃
                    (  3.30 mb …   3.30 mb)   3.30 mb ▄███▅▂▁▁▁▁▁▁▁▂▂▂▂▂▂▂▁
                  3.45 ipc (  2.75% stalls)  97.68% L1 data cache
          4.42M cycles  15.25M instructions  31.28% retired LD/ST (  4.77M)

• www.google.com (incognito)
------------------------------------------- -------------------------------
hexo-util (old)                1.05 ms/iter   1.04 ms  ▇█
                      (945.83 µs … 1.88 ms)   1.80 ms  ██
                    (  1.75 mb …   1.83 mb)   1.76 mb ▅██▅▁▁▁▁▁▁▁▁▁▁▁▁▁▁▂▁▂
                  4.38 ipc (  2.74% stalls)  98.44% L1 data cache
          3.59M cycles  15.73M instructions  35.03% retired LD/ST (  5.51M)

hexo-util (new)              627.14 µs/iter 625.46 µs  █▄
                      (579.17 µs … 1.42 ms) 993.42 µs ▆██
                    (676.91 kb … 676.97 kb) 676.93 kb ████▂▁▁▁▁▁▁▁▁▁▁▁▁▁▂▂▁
                  3.44 ipc (  0.75% stalls)  98.67% L1 data cache
          2.16M cycles   7.44M instructions  17.23% retired LD/ST (  1.28M)

fast-escape-html             397.64 µs/iter 401.04 µs  █▇
                    (370.92 µs … 743.79 µs) 596.79 µs ███▆
                    (513.24 kb … 513.29 kb) 513.27 kb ████▆▂▂▁▁▁▁▁▁▁▁▁▁▂▂▁▁
                  4.76 ipc (  0.85% stalls)  99.01% L1 data cache
          1.37M cycles   6.53M instructions  19.20% retired LD/ST (  1.25M)

escape-html                  400.46 µs/iter 403.54 µs  ▆█
                    (375.17 µs … 901.50 µs) 591.17 µs ███▅
                    (513.24 kb … 513.29 kb) 513.27 kb ████▆▃▂▁▁▁▁▁▁▁▁▁▁▁▁▂▁
                  4.74 ipc (  0.84% stalls)  99.02% L1 data cache
          1.38M cycles   6.53M instructions  19.10% retired LD/ST (  1.25M)

@napi-rs/escape              443.95 µs/iter 464.38 µs  ▄█▂
                      (390.88 µs … 1.42 ms) 658.25 µs  ███▅▇▇
                    (390.05 kb … 390.05 kb) 390.05 kb ▇███████▄▃▂▁▁▁▁▁▁▁▁▁▁
                  4.58 ipc ( 16.59% stalls)  82.85% L1 data cache
          1.51M cycles   6.91M instructions  22.02% retired LD/ST (  1.52M)

html-escaper                 511.33 µs/iter 512.13 µs  ▄█▂
                      (458.33 µs … 1.10 ms) 779.25 µs  ███
                    (732.50 kb … 761.35 kb) 758.76 kb ▄████▅▂▂▁▁▂▁▂▂▂▂▂▁▁▁▁
                  4.05 ipc (  3.08% stalls)  97.58% L1 data cache
          1.74M cycles   7.06M instructions  33.19% retired LD/ST (  2.34M)

lodash.escape                515.14 µs/iter 514.71 µs  █
                      (465.71 µs … 1.19 ms) 952.42 µs  ██
                    (733.70 kb … 825.25 kb) 759.80 kb ▇██▅▂▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁
                  4.08 ipc (  3.00% stalls)  97.68% L1 data cache
          1.76M cycles   7.18M instructions  33.41% retired LD/ST (  2.40M)

escape-goat                  660.67 µs/iter 652.08 µs  █▇
                      (563.75 µs … 2.13 ms)   1.20 ms  ██▅
                    (  1.84 mb …   2.10 mb)   1.97 mb ▄███▅▁▁▁▁▁▂▂▂▂▁▂▂▁▁▁▁
                  3.30 ipc (  4.70% stalls)  95.00% L1 data cache
          2.26M cycles   7.46M instructions  29.01% retired LD/ST (  2.16M)

• about.gitlab.com
------------------------------------------- -------------------------------
hexo-util (old)                1.79 ms/iter   1.73 ms  ▄█
                        (1.59 ms … 2.91 ms)   2.81 ms  ██
                    (  2.76 mb …   2.80 mb)   2.80 mb ▃██▆▁▁▁▁▁▁▁▁▁▁▁▁▁▃▂▂▁
                  4.67 ipc (  2.30% stalls)  98.86% L1 data cache
          6.00M cycles  28.05M instructions  35.44% retired LD/ST (  9.94M)

hexo-util (new)              984.53 µs/iter 984.50 µs  ▄█▄
                      (908.79 µs … 1.60 ms)   1.35 ms  ███▂
                    (  1.09 mb …   1.09 mb)   1.09 mb ▆████▄▂▁▁▁▁▁▁▁▁▂▂▃▂▁▁
                  4.00 ipc (  0.79% stalls)  98.71% L1 data cache
          3.38M cycles  13.54M instructions  16.85% retired LD/ST (  2.28M)

fast-escape-html             745.81 µs/iter 745.46 µs  ▅█
                      (693.33 µs … 1.27 ms)   1.08 ms ▇██▇
                    (917.91 kb … 917.98 kb) 917.93 kb ████▅▂▁▁▁▁▁▁▁▁▁▁▂▂▂▂▁
                  4.80 ipc (  0.74% stalls)  99.27% L1 data cache
          2.57M cycles  12.31M instructions  19.35% retired LD/ST (  2.38M)

escape-html                  757.80 µs/iter 757.46 µs  ▇█
                      (701.75 µs … 1.18 ms)   1.10 ms ▄██▇
                    (917.91 kb … 917.98 kb) 917.93 kb ████▆▂▁▁▁▁▁▁▁▁▁▁▂▂▂▂▁
                  4.74 ipc (  0.71% stalls)  99.31% L1 data cache
          2.60M cycles  12.33M instructions  19.25% retired LD/ST (  2.37M)

@napi-rs/escape              881.45 µs/iter 895.46 µs    █▆
                      (791.96 µs … 2.35 ms)   1.20 ms   ▇███▃
                    (732.54 kb … 732.54 kb) 732.54 kb ▄▆█████▄▃▂▂▁▁▁▁▁▁▁▁▁▁
                  4.95 ipc ( 22.49% stalls)  79.85% L1 data cache
          2.98M cycles  14.72M instructions  23.03% retired LD/ST (  3.39M)

html-escaper                 953.99 µs/iter 957.13 µs  ▆█▂
                      (843.92 µs … 1.75 ms)   1.48 ms  ███
                    (  1.36 mb …   1.42 mb)   1.37 mb ▄████▄▂▃▃▂▂▂▁▂▂▁▁▁▁▁▁
                  4.04 ipc (  3.39% stalls)  97.46% L1 data cache
          3.23M cycles  13.06M instructions  32.83% retired LD/ST (  4.29M)

lodash.escape                944.43 µs/iter 941.96 µs   █▃
                      (856.21 µs … 1.48 ms)   1.35 ms  ███▂
                    (  1.36 mb …   1.42 mb)   1.37 mb ▄████▃▂▁▁▁▁▁▁▁▂▂▂▂▂▁▁
                  4.09 ipc (  3.38% stalls)  97.53% L1 data cache
          3.23M cycles  13.21M instructions  33.10% retired LD/ST (  4.37M)

escape-goat                    1.13 ms/iter   1.15 ms  ▂██
                      (995.13 µs … 3.06 ms)   1.54 ms  ████▃
                    (  3.65 mb …   3.65 mb)   3.65 mb ▄██████▄▃▃▃▅▄▃▂▃▂▂▁▁▁
                  3.43 ipc (  4.99% stalls)  95.23% L1 data cache
          3.84M cycles  13.17M instructions  29.18% retired LD/ST (  3.84M)

coveralls · 2025-05-31T11:06:01Z

coverage: 97.09% (+0.2%) from 96.875%
when pulling dfcfcdc on SukkaW:escape-html
into d497bc7 on hexojs:master.

D-Sketon · 2025-05-31T14:46:01Z

before PR

escapeHTML('&0') // => &amp;0

after PR

escapeHTML('&0') // => &0

Is this intentional?

SukkaW · 2025-06-01T08:36:58Z

Is this intentional?

Hmmmm, extra check again.

SukkaW · 2025-06-14T08:43:32Z

@D-Sketon Check again? Added extra check for ;.

D-Sketon · 2025-06-14T12:27:26Z

consider malformed entity patterns
before PR

escapeHTML('&#12345678;') // => &amp;#12345678;   (#\d{1,7} ×)
escapeHTML('&#x1234567;') // => &amp;#x1234567;   (#[Xx][a-fA-F0-9]{1,6} ×)

after PR

escapeHTML('&#12345678;') // => &#12345678;
escapeHTML('&#x1234567;') // => &#x1234567;

SukkaW requested review from D-Sketon, curbengh and stevenjoezhang May 31, 2025 11:04

SukkaW mentioned this pull request May 31, 2025

perf(escapeHTML): use multiple replace #178

Closed

SukkaW force-pushed the escape-html branch from c6d671d to db49159 Compare May 31, 2025 11:05

SukkaW requested a review from yoshinorin May 31, 2025 11:05

perf: faster escape html & unescape html

058e98b

SukkaW force-pushed the escape-html branch from db49159 to 058e98b Compare May 31, 2025 11:11

SukkaW added 3 commits May 31, 2025 19:15

test(escape_html): improve coverage

7164b01

chore: make eslint happy

a8893ef

fix(escape_html): an edge case

0e16484

SukkaW marked this pull request as draft June 1, 2025 08:37

SukkaW added 2 commits June 10, 2025 23:35

fix(escape_html): more edge cases

29f074f

chore: make eslint happy

dfcfcdc

SukkaW marked this pull request as ready for review June 14, 2025 08:43

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

perf: faster escape html & unescape html #421

perf: faster escape html & unescape html #421

Uh oh!

SukkaW commented May 31, 2025 •

edited

Loading

Uh oh!

coveralls commented May 31, 2025 •

edited

Loading

Uh oh!

D-Sketon commented May 31, 2025

Uh oh!

SukkaW commented Jun 1, 2025

Uh oh!

SukkaW commented Jun 14, 2025

Uh oh!

D-Sketon commented Jun 14, 2025 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

perf: faster escape html & unescape html #421

Are you sure you want to change the base?

perf: faster escape html & unescape html #421

Uh oh!

Conversation

SukkaW commented May 31, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

check list

Description

Uh oh!

coveralls commented May 31, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

D-Sketon commented May 31, 2025

Uh oh!

SukkaW commented Jun 1, 2025

Uh oh!

SukkaW commented Jun 14, 2025

Uh oh!

D-Sketon commented Jun 14, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

SukkaW commented May 31, 2025 •

edited

Loading

coveralls commented May 31, 2025 •

edited

Loading

D-Sketon commented Jun 14, 2025 •

edited

Loading