v3 site: error 500 in translated docs slugs

[AlexxNB]

I'm translating the docs right now. Site can't be build when I use cyrillic in *** headers, because this line:

.replace(/[^a-zA-Z0-9]+/g, '-')

at https://github.com/sveltejs/svelte/blob/master/site/src/routes/docs/_sections.js#L105

... deletes all signs from the slug and causes 500 error.

a-zA-Z is a pattern for english letters only. The \p{Letter} could be used instead, but not all possible letters are allowed to be in URL.

[thgh]

Would latinize js work?

[AlexxNB]

Yes, it works for russian language. Should be checked with ex. chineese or arabian, I think...
BTW, latinization in URL is a bad practice. Usually, we use russian slugs in URL or leave english ones, if possible.

[Ryuno-Ki]

The other day I did some research about maintained slug libraries. Maybe you can check them out locally with different locales.
Would be great if you could share your findings (I'd like to inform 11ty about them).

[AlexxNB]

I can speak about Russian language only, maybe it will be applicable for most of alphabets based on non-latin chracters, like Arabian, or Greek(maybe)... Most of slug-libraries use latinization. This way is great for most of latin-characters languages like Deutsch, French, Swedish and so on.

Russian also can be latinized and used in slugs same way. There are many sites which use russian latinized slugs. But it is a bad practice:

www.site.ru/uchebnik/urok1/#Vvedenie

Good practices are:

1. slugs should be in English only:

www.site.ru/tutorial/lesson1#Introduction

2. slugs are in russian

www.site.ru/учебник/урок1#Введение

3. mixed 1 and 2

www.site.ru/tutorial/lesson1#Введение

Russian, Arabian, Chinese and many other languages can be used in URL(as I found). But I don't know, is it good practice to use native characters in URL in every language.

So, as for me there are two good ways:

1. Should be a way to write slugs manually
2. Slugs should be autogenerated in english or native language.

[AlexxNB]

Now, translated 4-level headings **** headers are also cause 500 error

[AlexxNB]

Just for competent example:
Link to MDN with Russian slugs: https://developer.mozilla.org/ru/docs/Learn/JavaScript/Building_blocks/События#Всплытие_и_перехват_событий
They use mixed English and Russian slugs based on headers in text...

[mindrones]

MDN is based on kuma, a massive Django app with i18n support that does a lot of things related to languages (@Rich-Harris probably a good repo to look at if/when the Sapper i18n/l10n project will actually start): I had a look there and I got lost honestly as I don't have much time :)

I added some support to unicode slugs, see #2404 (comment), but I'm no i18n expert so I'm not sure this would be the right way to go.