Introduce Bun.stringWidth

[Jarred-Sumner]

What does this PR do?

This fixes an issue where console.table would not align text properly when the input contains ANSII escape sequences.

This also introduces Bun.stringWidth, which has nearly the same API as the extremely popular string-width npm package (> 100 million weekly downloads) except this implementation is roughly 490x faster compared to running string-width in node.

Fixes #8155

❯ bun string-width.mjs
cpu: Apple M1 Max
runtime: bun 1.0.24 (arm64-darwin)

benchmark                                    time (avg)             (min … max)       p75       p99      p995
------------------------------------------------------------------------------- -----------------------------
npm/string-width (ansi + emoji + ascii)   10.68 µs/iter   (8.17 µs … 832.46 µs)   9.92 µs  16.83 µs  23.13 µs
npm/string-width (ansi + emoji)            8.67 µs/iter     (7.88 µs … 9.59 µs)   8.92 µs   9.59 µs   9.59 µs
npm/string-width (ansi + ascii)             9.9 µs/iter   (7.75 µs … 734.63 µs)   8.75 µs   14.5 µs  28.04 µs
Bun.stringWidth (ansi + emoji + ascii)    82.29 ns/iter   (78.89 ns … 88.89 ns)  83.53 ns  86.75 ns  87.52 ns
Bun.stringWidth (ansi + emoji)            82.48 ns/iter   (76.63 ns … 94.76 ns)  84.05 ns  87.02 ns   88.3 ns
Bun.stringWidth (ansi + ascii)            24.71 ns/iter   (20.75 ns … 35.67 ns)  25.37 ns  26.84 ns  27.58 ns


❯ node string-width.mjs
cpu: Apple M1 Max
runtime: node v21.4.0 (arm64-darwin)

benchmark                                    time (avg)             (min … max)       p75       p99      p995
------------------------------------------------------------------------------- -----------------------------
npm/string-width (ansi + emoji + ascii)   13.85 µs/iter  (11.54 µs … 770.92 µs)  12.71 µs  18.29 µs  22.58 µs
npm/string-width (ansi + emoji)             9.5 µs/iter     (8.94 µs … 9.98 µs)   9.66 µs   9.98 µs   9.98 µs
npm/string-width (ansi + ascii)           12.31 µs/iter     (9.96 µs … 2.28 ms)  11.25 µs   15.5 µs  19.54 µs

How did you verify your code works?

There are some tests

[Jarred-Sumner]

[github-actions]

❌ @Jarred-Sumner 3 files with test failures on bun-darwin-aarch64:

• test/integration/next/default-pages-dir/test/dev-server.test.ts 1 failing
• test/js/node/watch/fs.watchFile.test.ts 1 failing
• test/js/third_party/webpack/webpack.test.ts 1 failing

View test output

^{#a53d35eeeb22474bea3da334e210027081ca4b78}

[github-actions]

❌ @Jarred-Sumner 4 files with test failures on linux-x64:

• test/cli/test/bun-test.test.ts 1 failing
• test/integration/next/default-pages-dir/test/dev-server.test.ts 2 failing
• test/integration/next/default-pages-dir/test/next-build.test.ts 1 failing
• test/js/third_party/webpack/webpack.test.ts 1 failing

View test output

^{#a53d35eeeb22474bea3da334e210027081ca4b78}

[github-actions]

❌ @Jarred-Sumner 4 files with test failures on linux-x64-baseline:

• test/cli/test/bun-test.test.ts 1 failing
• test/integration/next/default-pages-dir/test/dev-server.test.ts 1 failing
• test/integration/next/default-pages-dir/test/next-build.test.ts 1 failing
• test/js/third_party/webpack/webpack.test.ts 1 failing

View test output

^{#a53d35eeeb22474bea3da334e210027081ca4b78}

[github-actions]

❌ @Jarred-Sumner 5 files with test failures on bun-darwin-x64:

• test/cli/test/bun-test.test.ts 1 failing
• test/js/bun/test/test-test.test.ts 1 failing
• test/js/bun/util/which.test.ts 1 failing
• test/js/node/fs/fs.test.ts 1 failing
• test/js/third_party/webpack/webpack.test.ts 1 failing

View test output

^{#a53d35eeeb22474bea3da334e210027081ca4b78}

[tiptenbrink]

This is not a trivial problem, I would be careful taking on the maintenance burden. Looking at the implementation, I think the most important things are correct (also very nice fast implementation), although I'm not entirely sure how a character knows to be width zero after following a zero-width join. I'm sure you've looked at a lot of resources already, but https://www.jeffquast.com/post/ucs-detect-test-results/ is a good article on the difficulty of the problem. It includes some specific examples that I recommend adding to the tests. Of course there is also the problem of width changing over time for Unicode versions, so what is actually displayed might differ by device.

[Jarred-Sumner]

This is not a trivial problem, I would be careful taking on the maintenance burden. Looking at the implementation, I think the most important things are correct (also very nice fast implementation), although I'm not entirely sure how a character knows to be width zero after following a zero-width join. I'm sure you've looked at a lot of resources already, but https://www.jeffquast.com/post/ucs-detect-test-results/ is a good article on the difficulty of the problem. It includes some specific examples that I recommend adding to the tests. Of course there is also the problem of width changing over time for Unicode versions, so what is actually displayed might differ by device.

We're mostly aiming to match the behavior of the string-width library and/or node's readline. @paperdave has a branch that makes it more directly use ICU's tables instead of our copied version. The ZWJ issue will need to be fixed and I had a branch that does it but it really should use icu's functions instead of our copied version.

[Jarred-Sumner]

Merging because I'd like to get this in the release and it feels pretty safe to do so.

[github-actions]

❌🪟 @Jarred-Sumner, there are 10 test regressions on Windows x86_64

• test\cli\inspect\inspect.test.ts
• test\js\bun\shell\bunshell-instance.test.ts
• test\js\bun\shell\bunshell.test.ts
• test\js\bun\shell\commands\rm.test.ts
• test\js\bun\shell\lazy.test.ts
• test\js\bun\shell\leak.test.ts
• test\js\bun\shell\lex.test.ts
• test\js\bun\util\stringWidth.test.ts
• test\js\node\events\events-cjs.test.js
• test\regression\issue\08095.test.ts

Full Test Output