Ruff treats every Chinese character as two and often falsely reports the "Line too long (E501)" warning

[jackjyq]

I have noticed that Ruff treats every Chinese character as two, so it often falsely reports the "Line too long (E501)" warning.

x = "一二三四五六七八九十一二三四五六七八九十一二三四五六七八九十一二三四五六七八九十一二三四五六七八九十一二三四五六七八九十一二三四五六七八九十一二三四五六七八九十"

I believe that this behavior is different than Flake8 and Black as they treat one Chinese character as one.

I use Ruff (v2023.20.0) the VSCode version with all the default settings.

Other considerations

I know that Chinese characters take more space than ASCII characters and the actual ratio depends on fonts and rendering.

On Windows 11 VSCode,

• Dejavu Sans Mono: 53 Chinese characters take 88 ASCII characters' space
• Cascadia Mono: 52 Chinese characters take 88 ASCII characters' space

As we can see, the radio is not 2.

Also, this behaviour makes Ruff not compatible with the Black formatter because more often than not, the Black does not break lines while Ruff keeps reporting the E501 warning.

[charliermarsh]

There's a bit more info on this in #3902 and #3825. Ruff is using character width (which is defined as part of the Unicode standard, I think, although I'm not an expert), rather than character count. Black made the same change, but it's still in preview (you should see the same behavior if you run with black --preview). We may consider changing this, though I'm going to close this as a duplicate of the others for now.

[charliermarsh]

(For posterity: the primary issue for tracking this is #3825.)