CJK text won't render out of the box

font-noto-cjk

that definitely works (and i support that 100%), but i'm also curious, what other cjk fonts do you know of? there's font-sarasa-gothic (extremely huge like iosevka) and then i'm not aware of many 'full cjk' sets that aren't one-language.

if there's something else workable though, might be able to pick something that everyone likes the look of :p

By psykose on 2023-04-13T18:25:56

added status::confirmed type::bug labels

Not sure where to draw the line of fonts included by default. After font-noto and font-noto-emoji (installed by default), the rest of the noto fonts (font-noto-all) is 421 MiB so not really feasible to have preinstalled everywhere. Sure, I doubt anyone would really need font-noto-mayan-numerals or many others. Funnily font-noto-all doesn't even seem to include font-noto-cjk which is another 88 MiB

By Luca Weiss on 2023-04-14T08:41:38

i feel like you misread- this is just -cjk, not -all et al at all.

By psykose on 2023-04-14T08:45:23

that said, it /is/ 88mb, so that's why i'm curious what cjk fonts exist.

By psykose on 2023-04-14T08:46:10

Sure, I know this issue is only about cjk but why preinstall cjk and not e.g. Arabic or any other script people use? I know preinstalling just latin1 is bad for UX that's why I said I'm not sure what fonts for various scripts to preinstall and which ones not. Number of speakers worldwide?

By Luca Weiss on 2023-04-14T08:49:53

ah, that makes sense.

i think cjk in itself is pretty huge, because that's china-japan-korea which is a pretty massive part of the internet. it's quite big compared to any other individual language.. i would say it's the second biggest after "a latin script" as a quick 'shorthand of language spec'?

of course then we get to all the other scripts, so i'm not really sure how we want to decide.

By psykose on 2023-04-14T09:12:17

TBH, I'm not sure if we shouldn't include Arabic script.

I can imagine it would suck for an Arabic speaker to download pmOS and notice that all their family's and friend's contact names are just □□□□□ (or do they use latin for their names??).

I'm also shocked at the size of font-noto-cjk. font-twemoji is 3MiB and font-noto-emoji is 9MiB. Then again, there are a lot more Chinese characters than there are emoji. I guess the main issue is that fonts include all possible characters, instead of just radicals and the rules to combine them. However, fixing that would probably require tinkering with font rendering engines.

By WhyNotHugo on 2023-04-14T09:15:35

Then again, there are a lot more Chinese characters than there are emoji. I guess the main issue is that fonts include all possible characters, instead of just radicals and the rules to combine them. However, fixing that would probably require tinkering with font rendering engines.

To elaborate on this a bit: the current approach font CJK fonts is the equivalent of having a single glyph for each possible syllable in English -- instead of having just individual letters (which are a couple of orders of magnitude less) and rules on how to combine them .

By WhyNotHugo on 2023-04-14T09:15:35

Is this something that could be dealt through pmbootstrap or the ondev installer? So that if the user chooses a specific locale, they're automatically installed?

By Pablo Correa Gomez on 2023-04-14T09:31:38

The locale is unrelated; my locale is en_GB.UTF-8, but I have contacts with names in Chinese.

Is this something that could be dealt through pmbootstrap or the ondev installer?

Indeed, a step offering to install extra fonts could be an alternative solution. I'm not sure about the benefits of it though; _pmb_recommends can be overriden too, right?

By WhyNotHugo on 2023-04-14T09:31:38

Maybe it would be good to check for alternative font(families) that support cjk beside the noto-family, but with a smaller foot-print.

Alpine package serach for "cjk" only returns font-noto-cjk as of now, so it might be a good idea to upstream to alpine?

By Robert Eckelmann on 2023-04-15T15:52:44

https://www.freejapanesefont.com/hanazono-mincho-%E8%8A%B1%E5%9C%92%E6%98%8E%E6%9C%9D/ is a 40MB one, and looks ok

By psykose on 2023-04-15T18:58:58

Does this include the C in CJK? https://cn.lipsum.com/

What about K? https://ko.wikipedia.org/wiki/로렘_입숨

By WhyNotHugo on 2023-04-15T18:58:58

yes, it does, though the korean one is i guess not very good

By psykose on 2023-04-15T19:10:06

I gathered a bit of feedback from native speakers (since I'm in no way qualified to say anything beyond "it renders"):

Noto CJK seems to be a rather neutral font and look okay for both Chinese and Japanese.
Hanazono looks a lot "better" for Japanese, but looks "weird" for Chinese.

The "weird" is something like what we'd think if the default font for English text is an exotic-looking one and italic used for regular text. Readable, but somehow looks wrong.

By WhyNotHugo on 2023-04-18T14:19:52

I've come across a few other languages that don't render. E.g.: the author's name in https://www.atbrakhi.dev/

I think that installing font-noto-all might be the best solution; that meta-package pulls all the other font-noto-* packages which each contain fonts for a separate languages. I guess that pmbootstrap could allow disabling them when doing manual builds? But it seems sane to include them in upstream builds.

By WhyNotHugo on 2024-07-26T20:58:43

Hi, I just noticed this issue, while I had filled a relevant issue last week.

A possible solution would be somewhat opposite to !5469 (merged): include font-droid-nonlatin by default and make it opt-out in pmbootstrap. Its footprint is 13.4 MB and it contains Arabic and some other scripts too.

Using Noto variable fonts is another reasonable choice. I opened https://gitlab.alpinelinux.org/alpine/aports/-/issues/16329 for that.

By Masanori Ogino on 2024-08-08T13:30:49

We'll probably want pmboostrap to prompt whether somebody wants to include these fonts or not when building an image (e.g.: in the same way that it prompts for a DE, etc).

I haven't had time to try and hack this, so if someone else has time, feel free.

For the official images, we should enable these by default. People should be able to read their own name (or names of family and friends) and their own language(s) out of the box.

By WhyNotHugo on 2024-08-08T13:30:49

I will try it.

People should be able to read their own name (or names of family and friends) and their own language(s) out of the box.

Agreed. I feel that I am really welcomed when a software handles my primary language without hurdles.

By Masanori Ogino on 2024-08-08T13:49:07

For the official images, we should enable these by default. People should be able to read their own name (or names of family and friends) and their own language(s) out of the box.

Agreed. 15MiB seems like a very reasonable price to pay too

By Pablo Correa Gomez on 2024-08-08T17:16:03

mentioned in merge request !5469 (merged)

By Masanori Ogino on 2024-08-09T10:15:10

mentioned in merge request !5540 (merged)

By Masanori Ogino on 2024-08-29T07:57:12

mentioned in merge request !5558 (closed)

By Masanori Ogino on 2024-09-05T03:01:38

added category::ui label

closed

CJK text won't render out of the box

Describe your issue

What's the expected behaviour?

What's the current behaviour?

How to reproduce your issue?

What device are you using?

On what postmarketOS version did you encounter the issue?

On what environment did you encounter the issue?

Environments

How did you get postmarketOS image?

What's the build date of the image? (in yyyy-mm-dd format)

Additional information

Child items 0

Activity