[ROADMAP] 2024年, weapp-tailwindcss 发展规划

[sonofmagic]

转眼这个项目已经 2 年多了，算是我这么多开源里面，目前唯一火起来，有人用的吧（笑～），秉着不抛弃不放弃用户的原则，和目前所存在的问题，我大概制定了一下未来的技术规划以及立下一个 Flag:

- 编写 swc 插件，使用 swc 来替换掉 babel ，因为 babel 在项目大的时候，性能比较低，可能一个大js文件 parse 就要 200-300ms，这个是缓存也无法解决的问题，所以需要对 js 方面的转义，以及相关的包 “锈化”(具体原因见下)

• - 添加 ast-grep (rust编写) 支持, 通过基准测试，在项目中，平均速度大约比 babel 快 2 倍左右 (3.1.0 版本支持 2024/03/23) Feature/ast grep #283
• - 发展到 10 个贡献者 (6/10)
• - 至少编写 4 篇文章给 weapp-tailwindcss (4/4)
• - used by 开源项目超过 500
• - 支持未来将要发布的 [email protected], 这个是新的 rust 引擎
• - 探究和支持 taro rn，这个也是很多用户都在提的 taro rn模式适配 #262

目前想到的就这些，希望我们能达成目标！

贡献者们快来啊，有啥不懂的，我都事无巨细的解释给你们听！

贡献指南: https://weapp-tw.icebreaker.top/docs/how-to-contribute

---

2024年12月31日完成情况:

• - 添加 ast-grep (rust编写) 支持
• - 发展到 10 个贡献者 (6/10) #544
• - 至少编写 4 篇文章给 weapp-tailwindcss (4/4)
• - used by 开源项目超过 500
• - 支持未来将要发布的 [email protected] #543
• - 探究和支持 taro rn #542

工作是真的忙啊！开源是真的不赚钱啊！
2025年继续！

[sonofmagic]

去除编写 swc 插件目标 (2024/03/13)

我把编写 swc 插件这个目标给去除了，原因是我基于 @ast-grep/napi 实现了一个 beta 版本，然后发布了。
结果跑 e2e 测试的时候，发现比原先的 babel 还要慢？？？！！！

然后，为了找到原因，我去看了 ast-grep 的源代码，并跑了一下它自带的 bench 来做基准测试

测试脚本不是我写的，是直接跑 https://github.com/ast-grep/ast-grep/tree/main/benches 这里的测试脚本

测试设均为 Mac Book Pro m1 ， 测试分为对 同步 和 异步 API 进行测试，

测试解析的 js 文件对象为 超小,小, 中等,巨大 四种体积，下方结果的 ops 数据越大越快。

测试结果如下:

同步API测试结果（点击查看）

Running "parseSyncBench: Parse One Line" suite...
Progress: 100%

  ast-grep sync parse:
    50 164 ops/s, ±5.38%    | 50.44% slower

  babel sync parse:
    3 642 ops/s, ±3.17%     | slowest, 96.4% slower

  oxc sync parse:
    39 646 ops/s, ±0.61%    | 60.83% slower

  swc sync parse:
    11 111 ops/s, ±0.64%    | 89.02% slower

  TypeScript sync parse:
    101 218 ops/s, ±2.51%   | fastest

Finished 5 cases!
  Fastest: TypeScript sync parse
  Slowest: babel sync parse
Running "parseSyncBench: Parse Small File" suite...
Progress: 100%

  ast-grep sync parse:
    3 516 ops/s, ±0.66%   | 58.09% slower

  babel sync parse:
    2 669 ops/s, ±4.13%   | 68.18% slower

  oxc sync parse:
    3 502 ops/s, ±2.77%   | 58.25% slower

  swc sync parse:
    2 004 ops/s, ±0.75%   | slowest, 76.11% slower

  TypeScript sync parse:
    8 389 ops/s, ±2.09%   | fastest

Finished 5 cases!
  Fastest: TypeScript sync parse
  Slowest: swc sync parse
Running "parseSyncBench: Parse Medium File" suite...
Progress: 100%

  ast-grep sync parse:
    200 ops/s, ±0.79%   | 48.05% slower

  babel sync parse:
    385 ops/s, ±3.51%   | fastest

  oxc sync parse:
    204 ops/s, ±0.92%   | 47.01% slower

  swc sync parse:
    137 ops/s, ±1.43%   | slowest, 64.42% slower

  TypeScript sync parse:
    340 ops/s, ±3.45%   | 11.69% slower

Finished 5 cases!
  Fastest: babel sync parse
  Slowest: swc sync parse
Running "parseSyncBench: Parse Huge File" suite...
Progress: 100%

  ast-grep sync parse:
    1.1 ops/s, ±1.24%   | 42.11% slower

  babel sync parse:
    1.5 ops/s, ±4.69%   | 21.05% slower

  oxc sync parse:
    1.2 ops/s, ±1.59%   | 36.84% slower

  swc sync parse:
    0.8 ops/s, ±2.14%   | slowest, 57.89% slower

  TypeScript sync parse:
    1.9 ops/s, ±3.52%   | fastest

Finished 5 cases!
  Fastest: TypeScript sync parse
  Slowest: swc sync parse

异步API测试结果（点击查看）

Running "parseAsyncBench: Parse One Line" suite...
Progress: 100%

  ast-grep async parse:
    21 930 ops/s, ±5.93%   | 77.25% slower

  babel async parse:
    1 087 ops/s, ±1.97%    | slowest, 98.87% slower

  oxc async parse:
    19 234 ops/s, ±4.30%   | 80.05% slower

  swc async parse:
    9 414 ops/s, ±1.59%    | 90.23% slower

  TypeScript parse(not async):
    96 388 ops/s, ±1.69%   | fastest

Finished 5 cases!
  Fastest: TypeScript parse(not async)
  Slowest: babel async parse
Running "parseAsyncBench: Parse Small File" suite...
Progress: 100%

  ast-grep async parse:
    4 749 ops/s, ±22.01%   | 42.87% slower

  babel async parse:
    921 ops/s, ±4.62%     | slowest, 88.92% slower

  oxc async parse:
    3 796 ops/s, ±0.56%   | 54.34% slower

  swc async parse:
    2 529 ops/s, ±2.46%   | 69.58% slower

  TypeScript parse(not async):
    8 313 ops/s, ±1.49%   | fastest

Finished 5 cases!
  Fastest: TypeScript parse(not async)
  Slowest: babel async parse
Running "parseAsyncBench: Parse Medium File" suite...
Progress: 100%

  ast-grep async parse:
    475 ops/s, ±4.92%   | fastest

  babel async parse:
    219 ops/s, ±40.10%   | 53.89% slower

  oxc async parse:
    246 ops/s, ±0.76%   | 48.21% slower

  swc async parse:
    190 ops/s, ±0.34%   | slowest, 60% slower

  TypeScript parse(not async):
    310 ops/s, ±5.35%   | 34.74% slower

Finished 5 cases!
  Fastest: ast-grep async parse
  Slowest: swc async parse
Running "parseAsyncBench: Parse Huge File" suite...
Progress: 100%

  ast-grep async parse:
    1.8 ops/s, ±64.07%   | fastest

  babel async parse:
    1.2 ops/s, ±8.97%   | 33.33% slower

  oxc async parse:
    1.5 ops/s, ±11.31%   | 16.67% slower

  swc async parse:
    1 ops/s, ±3.43%     | slowest, 44.44% slower

  TypeScript parse(not async):
    1.6 ops/s, ±9.93%   | 11.11% slower

Finished 5 cases!
  Fastest: ast-grep async parse
  Slowest: swc async parse

结论

用 rust 写的工具，异步的性能要比同步好一点，同时处理的文件越小，性能越快。
但是，小文件的解析本身也不怎么耗时间呀？
而且，看到 swc 这样的测试结果，我不由得对 rust 是前端新基建的说法产生了怀疑。

当然，如果你测试的结果有所不同，或者是有其他的异议，欢迎和我进行交流。

[sonofmagic]

找了另外一台 win64 机器做测试，配置比 mac book pro稍次，测试结果如下

同步API测试结果

Running "parseSyncBench: Parse One Line" suite...
Progress: 100%

  ast-grep sync parse:
    39 794 ops/s, ±1.73%   | 54.14% slower

  babel sync parse:
    2 688 ops/s, ±6.92%    | slowest, 96.9% slower

  oxc sync parse:
    36 530 ops/s, ±0.49%   | 57.9% slower

  swc sync parse:
    6 420 ops/s, ±2.84%    | 92.6% slower

  TypeScript sync parse:
    86 778 ops/s, ±0.84%   | fastest

Finished 5 cases!
  Fastest: TypeScript sync parse
  Slowest: babel sync parse
Running "parseSyncBench: Parse Small File" suite...
Progress: 100%

  ast-grep sync parse:
    2 710 ops/s, ±0.90%   | 65.74% slower

  babel sync parse:
    1 842 ops/s, ±2.91%   | 76.71% slower

  oxc sync parse:
    3 608 ops/s, ±1.11%   | 54.38% slower

  swc sync parse:
    1 580 ops/s, ±1.14%   | slowest, 80.02% slower

  TypeScript sync parse:
    7 909 ops/s, ±0.45%   | fastest

Finished 5 cases!
  Fastest: TypeScript sync parse
  Slowest: swc sync parse
Running "parseSyncBench: Parse Medium File" suite...
Progress: 100%

  ast-grep sync parse:
    147 ops/s, ±0.94%   | 43.68% slower

  babel sync parse:
    243 ops/s, ±11.63%   | 6.9% slower

  oxc sync parse:
    148 ops/s, ±2.25%   | 43.3% slower

  swc sync parse:
    100 ops/s, ±2.95%   | slowest, 61.69% slower

  TypeScript sync parse:
    261 ops/s, ±3.16%   | fastest

Finished 5 cases!
  Fastest: TypeScript sync parse
  Slowest: swc sync parse
Running "parseSyncBench: Parse Huge File" suite...
Progress: 100%

  ast-grep sync parse:
    0.8 ops/s, ±9.34%   | 52.94% slower

  babel sync parse:
    1.3 ops/s, ±7.97%   | 23.53% slower

  oxc sync parse:
    1 ops/s, ±5.11%     | 41.18% slower

  swc sync parse:
    0.7 ops/s, ±5.34%   | slowest, 58.82% slower

  TypeScript sync parse:
    1.7 ops/s, ±3.38%   | fastest

Finished 5 cases!
  Fastest: TypeScript sync parse
  Slowest: swc sync parse

异步API测试结果

Running "parseAsyncBench: Parse One Line" suite...
Progress: 100%

  ast-grep async parse:
    34 357 ops/s, ±5.95%   | 56.62% slower

  babel async parse:
    154 ops/s, ±2.47%      | slowest, 99.81% slower

  oxc async parse:
    26 347 ops/s, ±2.03%   | 66.73% slower

  swc async parse:
    7 503 ops/s, ±1.23%    | 90.53% slower

  TypeScript parse(not async):
    79 194 ops/s, ±3.63%   | fastest

Finished 5 cases!
  Fastest: TypeScript parse(not async)
  Slowest: babel async parse
Running "parseAsyncBench: Parse Small File" suite...
Progress: 100%

  ast-grep async parse:
    4 515 ops/s, ±13.88%   | 39.82% slower

  babel async parse:
    165 ops/s, ±0.70%     | slowest, 97.8% slower

  oxc async parse:
    3 560 ops/s, ±1.69%   | 52.55% slower

  swc async parse:
    2 271 ops/s, ±1.57%   | 69.73% slower

  TypeScript parse(not async):
    7 503 ops/s, ±1.00%   | fastest

Finished 5 cases!
  Fastest: TypeScript parse(not async)
  Slowest: babel async parse
Running "parseAsyncBench: Parse Medium File" suite...
Progress: 100%

  ast-grep async parse:
    283 ops/s, ±25.56%   | 5.03% slower

  babel async parse:
    94 ops/s, ±4.11%    | slowest, 68.46% slower

  oxc async parse:
    213 ops/s, ±0.59%   | 28.52% slower

  swc async parse:
    176 ops/s, ±1.04%   | 40.94% slower

  TypeScript parse(not async):
    298 ops/s, ±1.75%   | fastest

Finished 5 cases!
  Fastest: TypeScript parse(not async)
  Slowest: babel async parse
Running "parseAsyncBench: Parse Huge File" suite...
Progress: 100%

  ast-grep async parse:
    1.9 ops/s, ±4.50%   | fastest

  babel async parse:
    1.1 ops/s, ±35.09%   | 42.11% slower

  oxc async parse:
    1.3 ops/s, ±5.87%   | 31.58% slower

  swc async parse:
    1 ops/s, ±4.05%     | slowest, 47.37% slower

  TypeScript parse(not async):
    1.7 ops/s, ±5.60%   | 10.53% slower

Finished 5 cases!
  Fastest: ast-grep async parse
  Slowest: swc async parse

[sonofmagic]

ast-grep 作者 HerringtonDarkholme 大佬给出的测试结果:
https://docs.google.com/spreadsheets/d/1oIRXDaJ-EnjKz8GKpmUjVwh_FNw4Nsf6mbA0QPCiTh0/edit#gid=0
也同样应征了这个结果

文章在此: https://medium.com/@hchan_nvim/benchmark-typescript-parsers-demystify-rust-tooling-performance-025ebfd391a3